ETHIOPIA S Productive Safety Net Programme (PSNP):

Transcription

1 ETHIOPIA S Productive Safety Net Programme (PSNP): 2008 Assessment Report December 2008 Study Team Stephen Devereux and Rachel Sabates-Wheeler Centre for Social Protection, Institute of Development Studies Rachel Slater Overseas Development Institute Mulugeta Tefera Dadimos Development Consultants Taylor Brown The IDL Group Amdissa Teshome A-Z Consult

2 Contents EXECUTIVE SUMMARY...1 Chapter 1 INTRODUCTION Overview of the Productive Safety Net Programme Background to this study Rationale for the study Structure of this report...6 Chapter 2 METHODOLOGY Quantitative methods Household survey Community questionnaire Market survey Sample size and sampling frame Sampling procedure Region and woreda selection Kebele selection Village selection Household selection Sampling for the panel survey Qualitative methods Data collection Data entry and analysis Limitations of fieldwork Characteristics of surveyed households...14 Chapter 3 UNDERSTANDINGS OF THE PSNP Programme objectives and principles Targeting principles and procedures Payment schedules Graduation...21 Chapter 4 PSNP IMPLEMENTATION Targeting General targeting process Full family targeting Benefit incidence Exclusion Work norms Timing of public works Caps on working days per month Unified lists and Direct Support quotas Treatment of pregnant and lactating women under the PSNP Transfers Types of transfers Beneficiary preferences Value of PSNP transfers Reliability and predictability of PSNP transfer payments...41 i

3 4.5 Payment collection Transport and travel time Risks associated with collection of PSNP payments Livelihood Packages Implementation issues at woreda level Fedis, Oromia Chiro, Oromia Kilte awlelo, Tigray Enderta, Tigray Boricha, SNNPR Derashe, SNNPR Kalu, Amhara Lasta, Amhara...51 Chapter 5 IMPACTS OF PSNP TRANSFERS Consumption uses of PSNP transfers Investment uses of PSNP transfers Household incomes Household assets Livelihood activities Household food security Coping strategies Household self-assessment Graduation Gender impacts of PSNP transfers Social impacts Impacts on HIV and AIDS Impacts on vulnerability and emergencies PSNP impacts on markets and food prices Food price inflation and the real value of PSNP cash transfers Markets Food prices...78 Chapter 6 IMPACT OF THE PSNP ON BENEFICIARIES: RESULTS FROM THE ANALYSIS OF PANEL DATA 2006/ Analysis Results Income Assets Food security...91 Chapter 7 CONCLUSIONS AND RECOMMENDATIONS Implementation Targeting and graduation guidelines Delivery Cash and food transfers Impacts Livelihoods and food security Graduation Recommendations...96 REFERENCES...98 ii

4 Tables Table 1 Sampling frame for the household survey...9 Table 2 Sample stratification by PSNP beneficiary status...12 Table 3 Qualitative data collection methods...13 Table 4 Household demographic characteristics, by PSNP status and region...15 Table 5 Beneficiaries understanding of why they were selected for the PSNP...19 Table 6 Knowledge of the term graduation from the PSNP...25 Table 7 Beneficiaries confidence of graduating from the PSNP, by region...26 Table 8 Perspectives on graduation at different levels...27 Table 9 PSNP participation by income quintile, 2006 and Table 10 Type of transfers received by PSNP households, 2006 and Table 11 Total value of cash and food transfers to PSNP households, Table 12 Total value of cash and food transfers to PSNP households, per capita, Table 13 Beneficiary perceptions on reliability of PSNP payments...44 Table 14 Implications of delayed payments for PSNP households...45 Table 15 Current beneficiaries receiving livelihood packages, by income quintile...46 Table 16 Livelihood packages taken by PSNP beneficiaries, by year...47 Table 17 Self-reported benefits from livelihood packages...47 Table 18 Use of PSNP food transfers (last 12 months)...53 Table 19 Consumption uses of PSNP cash transfers (last 12 months)...54 Table 20 Investment uses of PSNP cash transfers (last 12 months)...55 Table 21 Annual household incomes, by PSNP status, 2006 and 2008 (birr)...56 Table 22 Changes in mean per capita income, , by region (birr/year)...58 Table 23 Changes in incomes and assets, by PSNP implementing agency...59 Table 24 Household asset values, by PSNP status, 2006 and 2008 (birr)...59 Table 25 Assets owned, by PSNP status, 2006 and Table 26 Land ownership by PSNP status, 2006 and Table 27 Household livelihood activities...62 Table 28 Households suffering food shortage in last year...64 Table 29 Households experiencing food shortages by PSNP status, 2006 and 2007/ Table 30 Household food security, incomes and assets, Table 31 Coping strategies adopted in 2007/08, by PSNP status...67 Table 32 Asset disposal behaviour in , by PSNP status (% of households)...68 Table 33 Subjective self-assessment, , by PSNP status (% of households)...69 Table 34 Livelihood packages by beneficiary status, Table 35 Responsibility for collecting PSNP cash and food payments...72 Table 36 What PSNP participants would be doing if they were not doing Public Works...72 Table 37 Beneficiary status in Table 38 Beneficiary status and livelihood packages in Table 39 Descriptive statistics of income in 2008 and 2006, by beneficiary status...85 Table 40 Estimated parameter (s.e.) of programme effects on income Table 41 Estimated parameter (s.e.) of type of support on income 2008 (only current Table 42 beneficiaries and non-beneficiaries)...88 Estimated parameter (s.e.) of interaction of livelihood package with beneficiary status on income Table 43 Descriptive statistics of asset values in 2008 and 2006, by PSNP status...91 Table 44 Descriptive statistics of food shortages in 2008 and Table 45 Descriptive statistics of food shortage in 2008 and 2006, by beneficiary status...92 iii

5 Figures Figure 1 Mean days worked on PSNP per household, 2007/ Figure 2 Type of transfers actually received by PSNP households, 2006 and Figure 3 Households receiving cash or food from the PSNP, by month...39 Figure 4 Preferences for types of assistance from PSNP...39 Figure 5 Household income, by PSNP status, 2006 to Figure 6 Changes in real household income, by PSNP status, 2006 to Figure 7 Mean per capita real income, by region and household type (birr/year)...58 Figure 8 Change in real asset values, by PSNP status, 2006 to Figure 9 Agricultural and non-agricultural wage rates, 2007 (birr/day)...63 Figure 10 Food shortages by PSNP status and month, 2007/ Figure 11 Households experiencing food shortages by region, 2007/ Figure 12 Households experiencing food shortages in 2005/06 and 2007/ Figure 13 Coping strategies adopted in 2007/08, by PSNP status...68 Figure 14 Self-assessment, current beneficiaries, 2006 to Figure 15 Self-assessment, past beneficiaries, 2006 to Figure 16 Self-assessment, non-beneficiaries, 2006 to Figure 17 Consumer Price Index (CPI) in Ethiopia, Figure 18 Food price inflation in Ethiopia, Figure 19 Average price of food crops in Ethiopia, 2007/08 (birr/kg)...78 Figure 20 Price ranges for staple food crops in Ethiopia, 2007/08 (birr/kg)...79 Figure 21 Food price seasonality in Ethiopia, 2007 (month when prices are highest)...80 Figure 22 Household income growth by PSNP status, bottom two and top quintiles...87 Figure 23 Predicted income in 2008 by type of support based on estimated model (bottom two and top quintiles)...89 Figure 24 Marginal effects for impact of programme on reducing food shortage from 2006 to Boxes Box 1 Targeting recommendations from the 2006 study... 4 Box 2 Experiences of targeting in Tigray Box 3 Public works implementation in Chiro Box 4 Improvements in unified lists and Public Works : Direct Support ratios in Lasta, Amhara Box 5 PSNP payment and public works arrangements in TIgray iv

6 Acronyms AIDS BoARD BOFED CFSTF CPI CSA DA DFID DHS DPPA DS EU FSCB FSCO FSP GoE HIV IDS IFPRI KFSTF ODI OFSP PA PIM PSNP PWP REST RRT SNNPR WFSTF Acquired Immune Deficiency Syndrome Bureau of Agriculture and Rural Development Bureau of Finance and Economic Development Community Food Security Task Force Consumer Price Index Central Statistical Agency Development Agents Department for International Development (UK) Demographic and Health Survey Disaster Prevention and Preparedness Agency Direct Support European Union Food Security Coordination Bureau Food Security Coordination Office Food Security Programme Government of Ethiopia Human Immunodeficiency Virus Institute of Development Studies International Food Policy Research Institute Kebele Food Security Task Force Overseas Development Institute Other Food Security Programmes Peasant Association Programme Implementation Manual Productive Safety Net Programme Public Works Programme Relief Society of Tigray Rapid Response Team Southern Nations, Nationalities and Peoples Region Woreda Food Security Task Force v

7 EXECUTIVE SUMMARY 1

8 CHAPTER 1 INTRODUCTION This introductory chapter provides an overview of the Productive Safety Net Programme, explains the purpose and objectives of this study, and outlines the structure of this report. 1.1 Overview of the Productive Safety Net Programme The Productive Safety Net Programme (PSNP) was launched by the Government of Ethiopia, with the support of a group of development partners, in January The PSNP targeted approximately five million chronic food insecure people living in 262 chronically food insecure woredas in 2005, but the number of PSNP beneficiaries was increased to eight million in The PSNP s objectives include smoothing household food consumption and protecting assets, strengthening household and community resilience to shocks, and breaking Ethiopia s chronic dependence on food aid. At a conceptual level, the PSNP recognised that food insecurity in Ethiopia has two dimensions: chronic and transitory. Responses to food insecurity have been dominated by emergency food aid, which is mobilised for millions of Ethiopians every year in response to localised or regional food crises, usually triggered by drought and harvest failure. In reality, a high proportion of families that receive food aid every year are not at risk of famine, but are chronically food insecure i.e. permanently unable to meet their food needs because of inadequate productive assets. They are forced to respond to shocks such as drought by selling their limited assets to buy food, which further impoverishes them, makes them more vulnerable to future shocks, and deepens their dependency on food aid. At an operational level, the PSNP aims to provide predictable transfers to meet predictable needs. In other words, chronically food insecure households will receive support for several months of the year for up to five years, bridging their annual food consumption gap and building their resilience, until they are no longer chronically food insecure and are better able to cope with moderate shocks. At this point, the household is considered to be food sufficient and is ready to graduate from the PSNP. Although emergency relief would continue to be required in years of severe shocks, if the PSNP is successful then large numbers of people would be taken off the annual emergency appeal process, and there would be a gradual shift towards a flexible multi-year safety net or integrated disaster risk management system that expands and contracts according to need. Two principles of the PSNP are: predictability ( A safety net delayed is a safety net denied ); and avoiding dependency ( by requiring able-bodied beneficiaries to provide labour in exchange for program transfers ). 1 Predictability was to be achieved by delivering cash or food transfers every month to eligible households for 5-6 months each year until the household graduates or the programme ends. (The PSNP is programmed to run for five years, from 2005 to 2009.) Dependency was minimised by requiring households with adult labour capacity to work on public works projects in return for cash or food transfers, while only a minority of households that had no labour capacity receive free transfers ( direct support ). 1 Government of Ethiopia (2004), The Ethiopian Productive Safety Net Program, Addis Ababa, pages

9 Another ambition of the PSNP was to shift the financing of the programme from food aid to cash. 2 This reflected the growing evidence base concerning the advantages of cash transfers (meeting food and non-food needs, being invested as well as consumed and stimulating rural markets) compared to food aid (which has been accused of undermining incentives for farmers and traders). One objective of the 2006 study, therefore, was to compare the impacts of food and cash transfers on recipients and local markets, to assess their respective strengths and weaknesses. Since 2006, however, there has been a significant shift against cash transfers and towards food (Pelham and Assegid, 2006), partly because woreda administrations expressed reservations about their capacity to handle cash, and partly because rising food prices have eroded the purchasing power of cash transfers and contributed to a shift in beneficiary preferences towards food (as discussed later in this report). 1.2 Background to this study In 2006, three independent teams carried out a comprehensive assessment of Ethiopia s Productive Safety Net Programme (PSNP) for the Government of Ethiopia and partner donor agencies. The three studies focused on trends in transfers (uses of cash and food transfers), targeting and institutional and programme linkages. 3 The purpose of the three linked studies was to assess the first phase of PSNP implementation (January 2005 to December 2006) and to make recommendations for the second phase, which was launched in In 2008, a repeat study was commissioned to update the understanding of key aspects of the PSNP. These include: understanding trends related to the use of cash and in-kind transfers determining progress towards making the transfers more predictable the effectiveness of changes made to targeting guidelines since the 2006 study progress made towards graduation from the PSNP beneficiary perceptions of the effectiveness and transparency of the programme HIV/AIDS-related risks and opportunities The fieldwork and analysis were divided into two parts. A team from IDS Sussex, Dadimos Development Consultants and Ethiopia s Central Statistical Agency designed and implemented a quantitative resurvey, while a second team from ODI, IDL and A-Z Consulting carried out qualitative research focusing mainly on PSNP governance and targeting issues. All findings are combined in this report, which is a synthesised product of both sets of fieldwork. 2 3 Government of Ethiopia (December 2004), Productive Safety Net Programme: Programme Implementation Manual, Ministry of Agriculture and Rural Development, Addis Ababa, page 1. These three studies were: (1) Stephen Devereux, Rachel Sabates-Wheeler, Mulugeta Tefera and Hailemichael Taye (2006), Ethiopia s Productive Safety Net Programme: Trends in PSNP Transfers within Targeted Households. Brighton and Addis Ababa: IDS Sussex and Indak International; (2) Kay Sharp, Taylor Brown and Amdissa Teshome (2006), Targeting Ethiopia s Productive Safety Net Programme (PSNP); London and Addis Ababa: ODI, the IDL Group and A-Z Consult; (3) Rachel Slater, Steve Ashley, Mulugeta Tefera, Mengistu Buta and Delelegne Esubalew (2006). Ethiopia Productive Safety Net Programme (PSNP): Study on Policy, Programme and Institutional Linkages, London and Addis Ababa: ODI, the IDL Group and Indak International. 3

10 1.2 Rationale for the study In 2006, the targeting study drew the conclusion that significant progress had been made between 2005 and 2006 in ensuring that the PSNP was reaching poor households and that institutional structures for combined administrative and community targeting were in place in most areas. Misinterpretations of targeting procedures in the first year had been corrected and no evidence of systematic corruption or large-scale abuse of the targeting system were found. The report identified some problems and areas for improvement, and made recommendations in the nine specific areas (see Box 1). The linkages study found that linkages are critical for graduation but they are inhibited by a number of conceptual, design and implementation issues within the PSNP. The study also suggested that the contribution of the other food security programmes (OFSP) could be significantly enhanced. The recommendations focused on better understanding and monitoring graduation; enhancing the performance of the PSNP; and ensuring the contribution of the OFSP to graduation. Recommendations, some of which echoed those in the targeting study included: establishing indicators or benchmarks of sustainable graduation, dissemination of graduation (operational) guidelines, setting a limit on the number of days of public works that any one person has to complete, revising the rules to allow households a mix of public works and direct support beneficiaries, A large number of the recommendations made in the two studies were adopted and have been implemented, to lesser or greater extent, since In 2007, following a series of joint donor group missions, the decision was made to repeat some aspects of the 2006 studies in order to assess PSNP progress. In particular, the donor missions threw up questions about: i. trends related to the use of cash and in-kind transfers ii. progress made towards making the transfers more predictable iii. the effectiveness of changes made since the last study to targeting guidelines iv. progress made towards graduation from the PSNP v. beneficiaries perceptions of the effectiveness and transparency of the programme vi. HIV/AIDS-related risks and opportunities These questions formed the TORs for a team of researchers from IDS, ODI, IDL and AtoZConsult, all of whom had worked on the studies in Box 1. Recommendations from the 2006 PSNP targeting study 1. Public Works and Direct Support targeting Preset quotas for DS should be dropped. The number of working and non-working beneficiaries in any community should be decided by needs assessment. A ceiling should be set on the number of days per month an individual is required to work on PW. Labour-poor households should receive transfers for the full family, even if they cannot cover the full work allocation. Community task forces should have authority to grant temporary maternity and sick leave for PW beneficiaries. 2. Temporal targeting of public works and payments PSNP works should be planned so as to minimise disruption to other activities conducive to the selfsufficiency and welfare of beneficiaries. Three levels of temporal targeting should be considered when planning the work: seasonality (i.e. 4

11 months of the year), days per month and hours per day. The PW work-day should be reduced, in areas where people are working eight hours or more. Continued efforts are needed to achieve regular, predictable, timely payment of transfers. The timing of transfer payments for DS beneficiaries should be de-linked from the public works. 3. Child labour and schooling Continued monitoring and supervision are needed to ensure that children are not employed on the PW. Community task forces and monitoring teams should check that the timing of PW does not conflict with participants school attendance, regardless of their age. 4. Retargeting and the registration period The requirement to re-target every six months (currently in the PIM) should be dropped. Individual cases of appeal should be heard periodically (without waiting for a major targeting exercise), and adjustments should be made to the beneficiary list accordingly. Once registered, beneficiaries should be guaranteed regular transfers for a minimum of 1 year. Beneficiaries should not be deregistered because of assets on credit, until debt is cleared. Woreda beneficiary databases should keep track of how long each household has been registered. 5. Appeals and grievance processes Efforts should continue to ensure that community targeting processes are transparent, participatory and well managed. Greater attention should be paid to raising awareness of the appeals process. The appeals process needs to be faster in almost all places. Bodies that hear appeals or complaints about targeting should have separate membership from the Food Security Task Forces. Reporting and appeals follow-up system is needed at the kebele, woreda and regional levels. 6. Monitoring Federal and regional RRTs should be strengthened; zonal RRTs should be established. The RRM should provide more substantive horizontal and vertical linkages between and across administrative layers. RRTs should contribute to improved reporting and record keeping on targeting and appeals, in order to strengthen transparency and accountability. The targeting section of the RRT field assessment checklist should be developed to encourage more systematic information-gathering, analysis and follow-up of targeting and appeals. 7. Geographical targeting Technical guidelines on geographical targeting should be developed for (and in consultation with) the woredas. 8. Gender Working hours on the PW should be reduced to enable women to combine participation in the programme with their domestic and other work. The number of elected women representatives on kebele and community FSTFs should be increased, and capacity-building budgets should include training and support for them. Gender awareness, and the specific gender issues encountered in the PSNP, should form part of ongoing training for all PSNP implementers. The requirement for evaluation and monitoring teams to consult women as well as men should be strengthened, and the collection and use of gender-disaggregated beneficiary information should be improved. Improvements to the appeals system (outlined above) should explicitly consider gender to ensure that women have fair access. 5

12 The head of each Woreda Women s Affairs Department, should be given a mandate and resources to monitor gender equity and women s interests within the PSNP. The Women s Affairs Department in the federal MoARD could provide advice and oversight. 9. National and Regional Targeting Guidelines A operational Revised Targeting Note should be produced to supplement the revised PIM. Regional governments with local stakeholders should detailed practical manuals adapted to their local conditions, social and cultural contexts, and capacities. Training should continue to local decision-makers to counteract staff turnover. Concise, accessible local-language versions of the key documents should be provided and kept in woreda and kebele offices. Source: Sharp et al Structure of this report Following this introduction, Chapter 2 describes the methodology that was devised for this study survey instruments, sampling, data collection and analysis, basic characteristics of surveyed households. Chapter 3 presents mainly qualitative findings about understandings of the PSNP programme objectives, targeting procedures, payment schedule, and graduation by PSNP participants, non-beneficiaries and programme staff. Chapter 4 explores various aspects of PSNP implementation, from targeting to work norms on Public Works, to types of transfers and beneficiary preferences for cash and food, to payment delivery (reliability and predictability) and collection (travel time and security risks), and concluding with a review of implementation issues in each of the eight study woredas. Chapter 5 presents survey findings on impacts of the PSNP. Chapter 6 provides findings from a panel survey comparing PSNP impacts in 2006 and 2007/08. Chapter 7 offers conclusions and recommendations based on the preceding analysis of survey findings. Finally, several Annexes to this report include the Terms of Reference, the household and community questionnaires, the qualitative methods, and additional data tables not included in the main text. 6

13 CHAPTER 2 METHODOLOGY The findings presented in this report are derived from fieldwork conducted during April 2008, in eight woredas in four regions where the PSNP is operational. The quantitative data come from a questionnaire survey of 960 PSNP beneficiary and non-beneficiary households, complemented by community questionnaires and market monitoring in 16 kebeles where the household survey was conducted. The qualitative data come from a number of key informant interviews, focus group discussions and household case studies. Three points should be noted at the outset. First, the household survey was a panel: by tracking and re-interviewing respondents who were interviewed two years ago for the first Trends in Transfers study, inferences can be drawn about changes over time and the impacts of the PSNP on a range of outcome indicators between 2006 and Second, to maximise the comparability of findings with the 2006 survey, the design of the questionnaires followed the 2006 design closely, with some modifications to capture additional issues of current interest. Third, the Central Statistical Agency was centrally involved in the 2008 survey, working in partnership with the Dadimos and IDS teams in recruiting and training enumerators, data collection, and data entry and analysis. This chapter describes the design of the quantitative and qualitative survey instruments, the sampling used in fieldwork, the process of data analysis, the timeframe for this assignment, and the limitations of the survey. 2.1 Quantitative methods Three methods of quantitative data collection were designed, pre-tested, and administered in the field: a household survey, a community questionnaire, and a market observation survey. These three survey instruments were also administered during the 2006 Trends in Transfers survey in the same sample enumeration areas. For this follow-up survey the IDS/ Dadimos team amended some questions in the earlier instruments, based on learning from the 2006 field experience, as well as requests for additional information from the PSNP donor coordination group and the Food Security Coordination Bureau (FSCB) Household survey A structured questionnaire was designed that was 26 pages in length. Although this seems long, not all questions were asked of every respondent. Some modules applied to all interviewees, but other modules applied only to PSNP beneficiaries interviewed while others were relevant only for non-beneficiaries. On average, each questionnaire took minutes to complete, and no interview lasted longer than one hour. One reason for this was to minimise respondent fatigue. A more practical concern is that the time allocated to fieldwork was limited, and a total of 960 households had to be interviewed, such that each enumerator was required to complete three to four questionnaires every day. The household questionnaire included the following modules: A. Household roster (age, sex, labour capacity, parental status of children, education level) B. Livelihood activities and income 7

14 C1. Land ownership and access C2. Crop farming D. Self-assessment (compared to last year, and to an average local household) E. Saving, lending and borrowing F1. Informal transfers F2. Formal transfers G. Asset inventory (livestock, productive assets, household goods, consumer durables) H. Food security (food shortage months, meals per day) I. Coping strategies over the past year (rationing, asset sales, borrowing, etc.) J1. Targeting and participation in PSNP (days worked on public works, transfers received) J2. Excluded (non-beneficiary) households (reasons for exclusion, etc.) K. Use of PSNP cash and food transfers L. Asset protection and asset building. The household survey questionnaire is appended as Annex 2 to this report Community questionnaire Complementing the household questionnaire, basic descriptive information was collected at the community level in each survey site. A standard form was drafted for the 2006 survey, and was slightly modified for the 2008 survey. Information captured fell into two categories: (1) Community-specific information: distances to nearest markets, schools, health facilities, HIV and AIDS services, and all-weather roads; changes in local services and infrastructure; main employment opportunities locally; unpaid labour contributions by community members. (2) Project-specific data: PSNP public works activities in the last three years; market wage rates for different farming and non-agricultural activities; PSNP payment procedures; targeting, appeals and graduation processes; transparency in PSNP implementation at community level. Key informants were interviewed in each village or community where fieldwork was undertaken. For the community-specific information, knowledgeable local individuals were asked to provide these facts (more than one person was consulted, to triangulate or verify the responses), but for the PSNP project-specific data, the people interviewed were familiar with implementation of the PSNP in the local area. On average, five persons were interviewed as key informants in each village. They included Kebele Chairmen, members of the Kebele Council, elders, development agents, traders, and informed community members including youth and women. The community questionnaire is appended as Annex 3 to this report Market survey While the enumerators were conducting household interviews, the supervisor of each team asked key informants about local markets (nearest markets for basic food items and other commodities, frequency of market days, seasonality of prices and supplies). Where possible, the supervisor also visited the most relevant local market to observe trader activity, volumes of commodities traded, and so on. Key informants, consumers and traders were also asked 8

15 retrospectively about prices and volumes of basic commodities during the past year, in an attempt to establish whether the operation of the PSNP in local communities had any impact on market response. Finally, supervisors attempted to collect any available price data from the Woreda Administration, specifically on the major crops traded in the local area, for the previous two to three years. 2.2 Sample size and sampling frame The sampling methodology was established during the 2006 survey, when the total sample size for the household survey was determined by the following formula: n z k 2 p(1 d 2 p) = 3 x 2 ( 1.645) x 0.50( ) = 894 n k z p desired sample size number of stages of sampling (i.e. 3-stage cluster sampling, in this case) standard normal deviation (1.96 for 95% confidence level, for 90% confidence) proportion of target population estimated to have characteristic (since no estimate is available, we use 0.5) q 1-p d degree of accuracy required (usually 0.05, as here, or 0.02 for greater significance) A stratified random sampling procedure was followed, with three stages of stratification (kebele, village, household). To ensure that at least 894 households were actually surveyed, this figure was rounded up to 960, disaggregated as 120 households per woreda, and 60 households per kebele (at two kebeles per woreda), and 60 households per village (at one village per kebele), unless the selected village had a small population, in which case 30 households were interviewed per village, in two contiguous villages per kebele. Table 1 shows how the sampling frame for the household survey was constructed. Regions and woredas were pre-selected. The next section explains how kebeles, villages and households were selected for the survey. Table 1 Sampling frame for the household survey 4 Teams Region Zone Woreda Kebele Village Team 1 Amhara (1) North Wollo Bugna Team 2 Amhara (2) South Wollo Kalu Team 3 Oromiya Chiro Fedis Team 4 SNNPR (1) Sidama Boricha {1 high PSNP} {1-2 villages =60 hhs} {1 medium PSNP} {1-2 villages =60 hhs} {1 high PSNP} {1-2 villages =60 hhs} {1 medium PSNP} {1-2 villages =60 hhs} {1 high PSNP} {1-2 villages =60 hhs} {1 medium PSNP} {1-2 villages =60 hhs} {1 high PSNP} {1-2 villages =60 hhs} {1 medium PSNP} {1-2 villages =60 hhs} {1 high PSNP} {1-2 villages =60 hhs} {1 medium PSNP} {1-2 villages =60 hhs} 4 In all tables and figures in this report where data are disaggregated by region, the four regions are ordered alphabetically: Amhara, Oromiya, SNNPR, Tigray. 9

16 {1 high PSNP} {1-2 villages =60 hhs} Team 5 SNNPR (2) Derashe {1 medium PSNP} {1-2 villages =60 hhs} {1 high PSNP} {1-2 villages =60 hhs} Enderta {1 medium PSNP} {1-2 villages =60 hhs} Team 6 Tigray {1 high PSNP} {1-2 villages =60 hhs} Wukro {1 medium PSNP} {1-2 villages =60 hhs} Total 4 Regions 8 Woredas 16 kebeles 960 households 2.3 Sampling procedure Sampling for the household survey was done at several levels: first the regions, then woredas, then kebeles, then villages, and finally PSNP beneficiary and non-beneficiary households Region and woreda selection Regions and woredas were not part of the sampling frame, as they were pre-selected in 2006 based on a number of criteria that are external to the survey design. It should be noted that the woredas surveyed reflect a range of implementation modalities. In 4 of our 8 woredas one in each region the PSNP is implemented by the government (Kalu, Derashe, Fedis and Wukro). In 3 woredas the implementing agency is an NGO (Bugna, Chiro and Enderta). In the eighth woreda (Borcha in SNNPR) the PSNP is implemented by the government and an NGO Kebele selection The survey kebeles were selected in advance of the 2006 survey, the main criteria being each kebele s level of participation in the Productive Safety Net Programme and its involvement in complementary programmes, specifically Other Food Security Programme (OFSP) activities. Differential levels of PSNP participation were considered important for the analysis of targeting issues in the PSNP, while the presence of OFSP activities is important for the analysis of linkages with complementary programmes and public services. In 2006, to identify the kebeles to be surveyed, the survey supervisors visited the office of the Bureau of Agriculture and Rural Development (BoARD) at each woreda capital, where they obtained a list of all kebeles in the woreda, including the percentage of households registered for inclusion on the PSNP by kebele, the total population of each kebele and the numbers of PSNP participants (to allow the percentage of inclusion to be calculated). Next, the list of kebeles was ranked by the percentage of households registered for inclusion on the PSNP. For example, if the highest percentage of PSNP participants was 93%, this kebele was ranked first, while if the lowest percentage was 17%, this kebele was ranked last. The next step was to select two kebeles from this list: one with a high participation percentage and one with a medium level of participation. This was done by selecting the highest-ranked kebele and the middle-ranked kebele. For example, if there were 15 kebeles in the woreda, the two selected kebeles were those that had the highest PSNP participation rate and the one that ranked eighth in terms of PSNP participation rate. At least one of these two kebeles also had to have ongoing Other Food Security Programme activities to be selected in If neither kebele was an OFSP kebele, it was replaced by one 10

17 ranked just above or just below it on the PSNP participation list that did have OFSP activities (e.g. the kebele ranked eighth was replaced by the one ranked seventh or ninth). If no kebele in the woreda had OFSP activities at the time of the survey, another kebele was selected where other relevant programmes are ongoing. A final consideration was logistics. If one kebele was inaccessible by road, or if the two kebeles were located far apart, this might have been impractical given the tight time constraints for the fieldwork. This constituted another reason for replacing a kebele with another that had a similar PSNP participation rate (e.g. the kebele ranked second might be chosen, rather than the one that ranked first in the woreda). During the April 2008 round of fieldwork the same kebeles were revisited as in Village selection As with the kebeles, survey villages were also selected in Once the two kebeles were selected in each woreda, one village had to be identified for the fieldwork in each kebele. Alternatively, since 60 households had to be interviewed in each kebele, two adjacent villages were selected if the first village identified had too few households. The same principles were followed as for kebele selection. In kebeles with high levels of PSNP participation, a village was selected with one of the highest participation rates. In kebeles with medium levels of PSNP participation, a village was selected with an average level of PSNP participation. If the rate of PSNP participation was not known at the level of individual villages, a village was selected at random from each kebele Household selection In the 2008 panel survey no household selection was conducted except for replacing missed households from the 2006 surveyed households. In 2006 two options were considered for sampling households within each selected village: (1) simple random sampling; (2) stratified sampling with quotas for each household category. Although there are many advantages to random sampling, since one objective of this study is to compare trends in households receiving food transfers with trends in those receiving cash transfers, the decision was taken to survey approximately equal numbers of cash and food beneficiaries overall. Also, on the assumption that approximately 80% of PSNP beneficiaries are employed on public works projects while no more than 20% are receiving direct support, these proportions were also reflected in the total sample. Finally, a number of non-beneficiaries (20% of the total) were also interviewed for their views on the PSNP programme. Non-beneficiaries were randomly selected from the pool of households that were not participating in the PSNP within each village visited. After replacing missing households four categories of household had been adequately covered in the 2008 panel survey: (1) Direct Support beneficiaries; (2) Public Works participants; (3) Past beneficiaries; (4) Non-beneficiaries. A new category of past beneficiaries is identified, comprising households that were PSNP beneficiaries in 2006 but have since left the programme, for a variety of reasons. The total 11

18 numbers of households in each category is summarised in Table 2, which also provides guidelines on the numbers and percentages of households interviewed per category. Table 2 Sample stratification by PSNP beneficiary status Household characteristic Number Percent PSNP transfers * Cash only Food only Cash + food PSNP category * Public Works Direct Support PSNP status Current beneficiaries Past beneficiaries (2006) Non-beneficiaries All households % * PSNP transfers and PSNP category are for current beneficiaries only (n=701) Sampling for the panel survey In this round the team revisited the same households sample in In addition, on the basis of their PSNP status, missed households were replaced by randomly selecting from household roster obtained from kebele administrations. 2.4 Qualitative methods The qualitative methodology was based on the methods used in the linkage and targeting studies in Fieldwork was carried in two woredas in each region, covering a kebele and one or more villages in each woreda. At the federal and regional levels, key informants were interviewed and secondary data and documents collected. Further fieldwork was carried out in the 8 rural woredas covered by the 2006 study. In each woreda the targeting team worked in one kebele and village. This sampling of these kebeles was not designed to be statistically representative, rather it provided an opportunity to revisit the woredas covered by the 2006 study and follow the changes in the PSNP that have taken place since 2006 and the impacts that these have had on poor households. One regional team comprising two researchers was deployed in each region to cover fieldwork for two woredas. Each team included one researcher who had worked on the 2006 targeting study. Staff from A-Z Consult, ODI and IDL also participated in fieldwork. As with the 2006 targeting study, a combination of different qualitative approaches was developed (Table 3). Information was collected primarily through qualitative field methods at woreda, kebele, village and household levels. A field procedures and methods guide was developed and finalised during a short period of training / familiarisation. This data was supplemented with a review of secondary sources and key informant interviews at regional and national level. The design of the qualitative fieldwork was coordinated with the re-survey to cross-reference the different data sources and ensure coverage of all areas of the TORs. 12

19 Fieldwork was undertaken between May and August The full fieldwork methodology is available as a supporting document on request. Table 3 Qualitative data collection methods Methods Level of data collection National Regional Woreda Kebele Village Household Review of documents Key informant interviews FSTF group interviews Focus group discussions Household case studies 2.5 Data collection Data collection training was provided at two levels. First level training was given to survey coordinators from seven branch offices at CSA head office in Addis Ababa. Following this, the survey coordinators provided training to supervisors and enumerators in their respective branch offices. The trainings were focussed on the familiarisation of survey instruments and some hints on effective management of survey. The first training was facilitated by CSA s head office staff and Dadimos, while the second round of training was managed by the survey coordinators. Each team comprised 1 Supervisor + 4 Enumerators, travelling in 1 vehicle with a driver. The six teams were as follows: Team 1. Amhara North: Team 2. Amhara South: Team 3. Oromiya: Team 4. Oromiya: Team 5. SNNPR Sidama: Team 6. SNNPR Gedeo: Team 7. Tigray: 1 team covering 1 woreda in North Wollo (Bugna) 1 team covering 1 woreda in South Wollo (Kalu) 1 team covering 1 woredas in West Hararghe (Chiro) 1 team covering 1 woredas in East Hararghe (Fedis) 1 team covering 1 Sidama-speaking woreda (Boricha) 1 team covering 1 Gedeo-speaking woreda (Derashe) 1 team covering 2 woredas (Enderta, Wukro) The reasons for deploying 7 teams rather than four (i.e. one team per region) included: 1. Distances between North and South Wollo would have required more travelling days to be added if a single Amhara Region team was deployed. The two woredas in this region are also under different branch offices of CSA. 2. Chiro and Fedis are under different branch offices and this has required to setup two different teams. 3. The two SNNPR woredas speak different languages and under two different CSA s branch offices so two teams were required. 13

20 2.6 Data entry and analysis The household survey questionnaires were entered using a user-friendly and self-automated data checking household survey data entry programme that was developed using CSPro 3.3. This programme has facilities to clean and export data to SPSS and STATA, which were both used for survey data analysis. In order to check the quality of data entry about 40% of questionnaires were double-entered and different entries were compared to identify critical areas for re-checking. Open-ended questions and market price data were entered and analysed in SPSS. The main software package for analysing the data both descriptive statistics and regressions was STATA 10. The data analysis was led by IDS in collaboration with Dadimos in Addis Ababa. Unlike in 2006 where we only had one cross-sectional data set, we now have two datasets with a high percentage of the same households. This enables us to perform some panel data analysis using fixed effects estimators Limitations of fieldwork Although the total sample size is generated by statistical theory applied during 2006, the sample of eight woredas is not representative of all of Ethiopia, nor of all communities where PSNP is being implemented, so the findings will be presented as indicative rather than statistically significant. This caveat applies even more to data presented at the regional, woreda and kebele levels, and to data presented for each of the five household categories. A statistically significant survey of the impacts of PSNP transfers within beneficiary households would require a randomised sampling frame and/or a larger stratified random sample. 2.9 Characteristics of surveyed households The average household size across the entire sample is PSNP households are smaller than the average, at 5.05, and significantly smaller than non-psnp households, at Direct Support households, with 3.16 members, are smaller than Public Works households, with 5.59 members. Almost two-thirds of Direct Support households, but only one in four Public Works households, are female-headed (Table 4). This is consistent with expectations: Direct Support households have little adult labour compared to Public Works households. 5 Note that this is not reported upon in this draft report and will be presented at the mid-term review workshop only if the authors have sufficient time to conduct the analysis. However this analysis is extremely important and will be completed and presented within the following six weeks. 14

21 Table 4 Household demographic characteristics, by PSNP status and region Demographic characteristics PSNP status Household size Dependency ratio Labour capacity index Femaleheaded households Public Works % Direct Support % Past PSNP beneficiaries % Non-PSNP beneficiaries % Region Amhara % Oromiya % SNNPR % Tigray % All households % Note: Dependency ratio is defined as the number of household members aged 0-15 years, plus the number of household members aged 60 and older, divided by the number of household members aged between 16 and 60 years old. This is calculated over 888 households as 72 households had zero adults making the ratio impossible to calculate (numerator =0). 15

22 CHAPTER 3 UNDERSTANDINGS OF THE PSNP This chapter explores perceptions and understandings of the PSNP by the key stakeholders involved in the programme. Aspects discussed include the objectives and principles of the PSNP, targeting procedures, payment schedules, and graduation. 3.1 Programme objectives and principles Overall, understanding of the main objectives and principles of the PSNP has improved since 2006, by programme implementers and programme participants, and even (to a limited extent) by non-beneficiaries. At woreda level, there is an improved understanding of the PSNP objectives and principles. Key officials of Woreda Food Security Task Forces (WFSTF) are able to outline the programme s objectives and explain its fundamental principles and processes. Officials draw more on the Programme Implementation Manual (PIM) than they did in Not all members of every WFSTF have this in-depth understanding some members, particularly those who were newly posted, were less clear on the purpose and operations of the programme. On the other hand, although staff turnover continues to be high (which affects the understanding of the programme and its implementation), there is some evidence that increasing numbers of WFSTF members have had exposure to the PSNP in other woredas where they have worked. At kebele level, the understanding of programme purpose and operations was generally clear and had improved since the previous fieldwork in Across all eight woredas visited, the Development Agents (Das) were a critical source on knowledge and understanding while the knowledge of Peasant Associations (PAs) and other Kebele Food Security Task Force (KFSTF) members was more variable. At community or village level, understanding of PSNP principles and objectives was less clear than at the woreda or kebele level, and there was significant variation. In some cases committees did not exist (Chiro, Fedis and Derashe) while in others (Enderta) there was a reasonable understanding. In one case (Kilte Awlelo), the recent reconstitution of the task force with community members who required orientation meant that knowledge of programme rules and processes had actually fallen since Beneficiaries also demonstrated an improved understanding of the objectives and operation of the PSNP in all woredas since the 2006 survey. In almost all focus group discussions, Public Works participants were able to explain the broad purpose of the PSNP. Men tended to be clearer than women, although on some operational issues women were very clear about their rights. In Kalu, for instance, pregnant women reported that they now know that they should apply for a transfer from Public Works to Direct Support (see below). Understanding among Direct Support beneficiaries was less strong than among Public Works participants. In general, understanding of the PSNP s principles and objectives is weakest among non-beneficiaries. Lack of understanding can contribute to intra-community tensions. The 2006 PSNP targeting study found that lack of transparency around targeting processes and priorities was a significant source of intra-community conflict. In Lasta, lack of transparency around the PSNP targeting process in 2006 led to significant tensions between beneficiaries and non-beneficiaries, and between non-beneficiaries and the FSTF. Overall, it appears that improved understanding of the programme has reduced tensions between beneficiaries and non-beneficiaries in some focus 16

23 woredas, including Lasta. However, in some woredas (e.g. Boricha), lack of understanding appears to generate ongoing tension between beneficiaries and non-beneficiaries. In parts of Chiro, the 2006 linkages study found that beneficiaries and non-beneficiaries were attending different mosques. Broadening the reach of familiarisation meetings to non-beneficiaries could be one mechanism for widening understanding of the PSNP, but may be difficult where non-beneficiaries are already refusing to participate in community activities following their exclusion from the programme. Although the level of understanding of the programme generally declines from woreda level to the grassroots, there are only small differences in how different stakeholders interpret the PSNP. At all levels, the PSNP is understood very much in terms of promoting development objectives rather than welfare objectives. In SNNPR, woreda level implementers use language like livelihoods, cooperatives and savings to explain the PSNP. The programme is an enabling effort to change the livelihoods of the poorest of the poor to be self-sufficient households. This programme differs from the earlier emergency programmes in that it enables people to organise cooperatives, it encourages savings, and it has strategies to fill in the food gaps that exist currently (Derashe WFSTF). A similar focus was found in Amhara. Among PSNP implementers in both Lasta and Kalu, three of the four cited objectives of the PSNP are associated with productive activities and assets, while only one refers directly to meeting the food gap. filling the annual food gap of the target groups (but not for those who migrate); ensuring food security through the protection and creation of assets; facilitating public works which contribute to environmental rehabilitation and community development; encouraging public work beneficiaries to engage in various programme packages that promote asset creation and generate income. This should not be interpreted as implying that implementers are not concerned about tackling food insecurity. Rather, it shows that the food security objective of the PSNP is expected to be achieved mainly indirectly through asset protection and asset creation, through employment and through income generation rather than as a direct ( welfarist ) result of the cash and food transfers. When discussing the main PSNP objectives and principles, while Direct Support beneficiaries liken the programme to emergency relief, Public Works participants stress environmental rehabilitation, infrastructure development and asset creation. They also discussed the role of the PSNP in preventing the loss of assets due to distress sales but, like programme implementers, few focused on the direct impact of the transfers on food security. Again like the implementers, beneficiaries also suggested that food security was achieved indirectly, through income from work and the creation of assets rather than the food or cash that they directly received. This view has important implications for the strategic priorities of programme implementers, and it influences many subsequent policy and implementation decisions. One example is how the concern with maximising Public Works outputs creating and protecting as many assets as possible might influence the size of the quota for Direct Support. In Enderta, the focus on expanding Public Works to maximise developmental effects from the PSNP translated into high 17

24 numbers of Public Works participants and very low numbers of Direct Support beneficiaries (less than 2% of total PSNP households). In Fedis, respondents focused on the importance of achieving food security through labour rather than transfers. The PSNP came to make people see that, whatever someone receives, it is from their own efforts. There are also relatively low levels of Direct Support beneficiaries in the Oromia woredas visited for this study. In subsequent sections of this report other examples of the implications of this bias towards the developmental aspects of the PSNP will be discussed, including how the focus on public works outputs rather than on labour inputs makes it difficult to assess whether caps on the number of days worked by individuals are upheld in practice; the growing pressure to graduate households from the programme; and the use of unified Direct Support and Public Works beneficiary lists. 3.2 Targeting principles and procedures At all levels, the understanding of targeting principles and procedures has improved since (This includes Amhara Region, where the 2006 survey found that many wealthier households were targeted initially, to ensure that graduation targets would be achieved.) All stakeholders now recognise the importance of targeting the poorest households, though they do not always agree on the best way to achieve this. There are also differences in targeting comprehension between levels. In Chiro, for example, there is a strong understanding at woreda level. At kebele level, training has been undertaken by officials, and the Task Force is aware of the need for a unified list of beneficiaries, for family targeting, and for ensuring that pregnant and lactating women are not working. Beneficiaries understandings are weaker. Most beneficiaries recognise the main principles for selection in a general sense (poverty, lack of land and labour), but very few are aware of the specific eligibility criteria (such as receiving food aid for three years before the PSNP started) (Table 5). Also, they often do not understand why some people are in the PSNP but others are not. In these cases they depend on a well-functioning CFSTF, though in a number of places, including the village where the qualitative team worked in Chiro, this was absent. It is also difficult to identify whether targeting has been done correctly when PSNP resources are limited and the caseload of eligible households is high. This leads to households being excluded from the PSNP despite having a large food gap, simply because there are not enough resources. The differences between the poorest households and those slightly further up the list are very small and this makes it difficult for people to know whether the targeting is correct and fair. This problem has been partially addressed through discussions and consultations at the community level. Where these discussions have taken place, they have resulted in improved beneficiary understanding of the targeting process and enabled community members to respond to unfair practices or mistakes when they occurred. In Lasta, Public Works participants argued that the community discussions led to an open discussion which helped the targeting process remain transparent. As a result, it was felt that the very poorest households in the community were now benefiting from the PSNP, while the wealthier households were now rightly excluded. 18

25 Table 5 Rank Beneficiaries understanding of why they were selected for the PSNP Explanation Current [n=701] Past [n=102] 1 Our household is poor 569 (81%) 86 (84%) 2 We have a small landholding 112 (16%) 33 (32%) 3 We don t produce enough food 104 (15%) 17 (17%) 4 We have no family support or remittances from relatives 85 (12%) 6 (6%) 5 We are landless 43 (6%) 8 (8%) 6 We have no labour 41 (6%) 4 (4%) 7 We can t get enough food to eat 39 (6%) 3 (3%) 8 The household head is female 38 (5%) 6 (6%) 9 We have poor quality land 34 (5%) 4 (4%) 10 We own no livestock, or only a few livestock 29 (4%) 8 (8%) 11 We have no off-farm income, or very little off-farm income 24 (3%) 2 (2%) 12 The household head is old 23 (3%) 0 (0%) 13 Our household is participating in other food security programmes 22 (3%) 0 (0%) 14 Household members are disabled or mentally challenged 20 (3%) 0 (0%) 15 Household members are sick 14 (2%) 0 (0%) 16 We have not received other government assistance (e.g. food aid) 12 (2%) 4 (4%) 17 We received food aid or emergency cash transfer in previous years 4 (1%) 1 (1%) 18 We were included only after we complained about being excluded 3 (0%) 1 (1%) 19 We don t know 2 (0%) 0 (0%) 20 There was no selection everyone in the village received something 1 (0%) 0 (0%) 21 Other reason 30 (4%) 8 (8%) Note: Multiple responses were permitted One additional concern raised during fieldwork was the way in which poor households that took an OFSP household package 6 on credit were excluded from subsequent rounds of the PSNP (see Box 2). In Kilte Awlelo, households believed they were coerced into taking packages as a precondition for getting onto the PSNP list. In Enderta, former PSNP beneficiaries complained that they were removed from the programme because the assets they received through the OFSP packages made them ineligible in terms of the PSNP targeting criteria, even though the loans were not yet repaid and the assets were not yet productive. 6 The Other Food Security Programme (OFSP) household packages complement the Productive Safety Net Programme, and are discussed in more detail later in this report. 19

26 Box 2. Experiences of targeting in Tigray During the first assessment [in 2005] we were included in the list of eligible beneficiaries. Some of us were benefiting in previous years, because initially it was not full family targeting, which gave us a chance to be included. But due to the limited quota allocated to the kushet, and the change of instructions to full family targeting, we are now excluded from the list of beneficiaries, with the understanding that we are better off. The criteria were based on asset ownership such as cattle and goats. We all took credit from the package (including those in the safety net), which we used to buy animals, which are now considered assets and became a reason for us being excluded. We have managed to reproduce them, but those who did not use the credit for the intended activity are now considered as poor and are included in the safety net. This shows that our capacity to use the credit properly which we are still repaying became a reason for us to be excluded from the safety net. Source: Focus group discussion, non-beneficiaries, Enderta Woreda, Tigray The mechanisms by which information filters from the federal down to the household level are crucial for an improved understanding of PSNP principles and procedures. Manuals and guidelines are the most important source of information on targeting. They do not reach everywhere, but kebele-level officials appeared to have much better access to information to assist them in implementation of the PSNP than previously. In Tigray, all manuals have been translated into Tigregna and are sent to woredas to be transferred to all administrative levels. For example, 8,000 copies of the Programme Implementation Manual (PIM) were distributed to the kushet level in At the regional level in Tigray, the main challenge to ensuring that people have the information they need was thought to be the rapid staff turnover at woreda level, which limits the sharing and distribution of manuals and guidelines. 3.3 Payment schedules In all woredas, most beneficiaries knew the PSNP payment schedule, though it was not always reliable. But there were geographical and gender differences in payment schedule preferences. Men and women expressed different preferences about payment schedules. In Enderta, men noted the importance of receiving cash at harvest time. This, they argued, would enable them to buy more food at lower prices, and enable non-food expenditures at an important time of year. Women questioned this approach, arguing instead that payments of cash or food should be made during the pre-harvest food deficit months. They argued that paying a regular one-month payment or a larger sum either in August (immediately before the Ethiopian new year) or at harvest would tempt beneficiaries to squander their transfers on non-essential expenditures. In Lasta, where PSNP households received food for four months and cash for two months, it was agreed that the cash should be delivered as early as possible in the hungry season in other words, in December-January, before grain prices begin their seasonal rise through the hungry months until the harvest. Food payments should be made later in the hungry season, starting around February, when prices are high and food is becoming increasingly difficult to procure using cash. Similarly, in Kalu, households receiving four months of cash and two months of food noted that their cash payments have become almost worthless due to inflation, and no longer allow a half-month s food ration to be purchased, as was originally intended. In Kalu, PSNP beneficiaries could see both sides of this difficult policy decision. They recognised that earlier payment would enable households to buy grain at lower prices, but that this might 20

27 lead to misuse in particular, consumption of all the grain before the months of greatest need but they also noted that the purchasing power of the transfer would be lower if payments are made later during the hungry period. In Fedis in 2007, public works participants were required to work during the agricultural slack season so that PSNP work requirements did not undermine productive labour on their own land or create greater dependency on the PSNP. However, because the slack season did not correspond with the hungry months, woreda officials delayed payment to prevent beneficiaries from spending this money until the hungry months before the harvest. However, this decision resulted in the cash transfers being disbursed only after food prices had already begun their seasonal rise. The experiences in Fedis and Kalu show the difficulty of getting the timing of payments right, because the seasonal peaks in hunger, labour requirements and food prices all coincide at the same time. Providing transfers when prices are low would provide beneficiaries with greatest purchasing power but not at the time of greatest need. Completing work earlier in the season and deferring payment to the hungriest months is often not acceptable to participants because their level of trust in the programme is not yet great enough. Continued efforts to improve predictability and reliability, and thereby generate greater trust in the programme, offer one pathway for overcoming this resistance. However, the quandary about the optimal timing of PSNP transfers remains unresolved. 3.4 Graduation The PSNP aims to reduce the annual food gap in food insecure households and to protect household assets against distress sales for food. This is the livelihood protection objective of the PSNP, but there is also a livelihood promotion objective. After receiving regular transfers for several months in 3 or more successive years, many households are expected to be ready to graduate from dependence on the PSNP, especially those households that also received livelihood packages in addition to PSNP transfers. The logic is that PSNP transfers should stabilise food consumption and protect household assets, while the livelihood packages would facilitate income generation and asset accumulation. One problem has been lack of clarity over the concept of graduation. Graduation from what? from food insecurity? from chronic poverty? from dependence on food aid? from participation in the PSNP? The Food Security Programme conceptualises graduation in terms of chronic food insecurity: a household has graduated when it is no longer chronically food insecure even after it no longer receives PSNP transfers. The Graduation Guidance Note issued by the FSCB in December 2007 describes graduation from the PSNP as a movement from being chronically food insecure to food sufficient, which is more precisely defined as follows: A household has graduated when, in the absence of receiving PSNP transfers, it can meet its food needs for all 12 months and is able to withstand modest shocks (FSCB, 2007: 1). A second source of confusion has surrounded the instruments that are intended to produce graduation. Is it the PSNP alone, or the livelihood packages alone, or the combination of PSNP plus livelihood packages? (Recalling that the Food Security Programme (FSP) includes both the PSNP and other food security programmes (OFSP).) Are households receiving Direct Support 21

28 also intended to graduate, or only those with labour capacity who are employed on the Public Works Programme? The Graduation Guidance Note clarifies this: It is not expected, however, that the PSNP will itself move households towards food sufficiency and graduation. For this to occur, additional interventions are required to build household assets and address vulnerabilities that make households food insecure (FSCB, 2007: 3). A third source of confusion surrounds the benchmarking of graduation. How can graduation be measured in an operational sense? How can programme implementers assess whether and when a particular household has graduated? What are the most appropriate and measurable indicators of graduation, and what thresholds should be set for monitoring these indicators? While graduation is not an official objective of the PSNP, it is an objective of the wider Food Security Programme of which the PSNP is a central component. Furthermore, government officials at all levels continue to be preoccupied with the achievement of graduation, as part of their PSNP implementation activities. For example, the government has commissioned work by IFPRI on establishing benchmarks for graduation, while at the local level Food Security Task Force committees are tasked with measuring graduation from the PSNP. Thus, for as long as graduation continues to be a priority for implementers and might even become a yardstick, officially or unofficially, for evaluating whether the PSNP has succeeded or failed resolving the contradictions identified above is crucial. In May 2007, the IFPRI commissioned report ( Benchmarks for Ethiopia s Food Security Program ) proposed a set of benchmarks for graduation, based on household assets. The authors noted that the benchmarks make no allowance for vulnerability to shocks such as droughts A threshold that incorporated resilience to drought or other shocks would be higher than those reported here (Gilligan et al., 2007: 2, 9). On the other hand, the PSNP was designed to address chronic food insecurity; it had no ambition to address transitory food insecurity. For the management of drought and other livelihood shocks, humanitarian relief would continue to be required. The PSNP hopes to strengthen the resilience of rural households against shocks, and thereby to reduce the need for emergency relief when shocks strike, but the PSNP cannot be expected to displace the need for humanitarian interventions when major shocks strike, even if the level of need is reduced. The IFPRI document argues that the benchmark for household graduation should be defined in terms of the assets needed to ensure that the household does not have an annual food gap the number of months that a household finds it difficult to meet its food needs being a common indicator of food insecurity in Ethiopia. This asset-based approach builds on the dual function of assets as generating income while also being used to secure access to food through sale or exchange when needed. Assets are considered more stable than income over time and more visible, so easier to monitor and count. Based on data collected for the Ethiopia Food Security Programme Baseline Survey in June to August 2006, the authors find a positive relationship between household assets (farm tools, livestock, education level, household size, land holding) and household food security status. 7 7 For example: in Tigray, 1,000 Birr of assets per person in the form of livestock and tools increases the probability that a household is food secure by about 11 per cent. a male-headed household in Tigray of average size, where the head has no formal schooling, the household has more than one hectare of land and has livestock and tools worth 1, 900 Birr per capita is predicted to have a 60 per cent probability that it is food secure (Gilligan et al., 2007: 4). 22

29 The IFPRI document notes that a trade-off exists between the level of the benchmark and the confidence that the household has achieved food security. A low benchmark risks graduating households from the PSNP prematurely, but a higher benchmark becomes unrealistic at the operational level. 8 It follows that the decision about setting thresholds for graduation is a policy choice about how much exclusion error the implementers of the PSNP consider acceptable. A worrying conclusion from the IFPRI analysis is that the baseline asset holdings of PSNP households was so low that most would have to accumulate several times the value of their assets in 2006 before they could be said with confidence to have graduated from the PSNP. There are regional variations in this conclusion: PSNP households with relatively high baseline asset holdings (especially in Tigray) or which had a year of good rainfall before the baseline survey (Amhara) were closer to graduation benchmarks than households with low baseline assets (SNNPR) or which had a year of bad rainfall before the baseline survey (Oromiya). A further complication is that the benchmarks relate asset prices to food prices, since the purpose is to ensure household food security, so the benchmarks are sensitive to changes in food prices. If food prices are rising faster than asset prices, then the benchmarks need to be increased both to account for the rise in food prices and, in addition, the fact that assets have lost some of their purchasing power (Gilligan et al., 2007: 9). Since food prices have risen very sharply in Ethiopia in the past 18 months, this suggests that the benchmarks for graduation as calculated by IFPRI in May 2007 are already out of date and would to be adjusted upwards. A final problem with the IFPRI proposal was that it has limited operational usefulness, because instead of providing a single quantitative threshold for graduation it presented a range of scenarios, disaggregated by combinations of household characteristics and levels of confidence that the household had achieved graduation. In late 2007 the Food Security Coordination Bureau (FSCB) issued a Graduation Guidance Note, which set out the indicators and criteria for graduation from the PSNP. The Guidance Note states that each Community Food Security Task Force (CFSTF) is supposed to make an assessment every year to determine whether each PSNP household in their jurisdiction should remain in the PSNP or not, on the basis of its food insecurity status. Importantly, since the PSNP alone is not expected to graduate households, only households that are participating in the PSNP and have taken an OFSP household package and credit are assessed. The Note set out the level of assets a food sufficient household should have in order to graduate from the PSNP. This is benchmarked by region and by year, and can also be adjusted by livelihood zone, to capture different asset bundles required for different livelihood activities. The Note states explicitly that, until the household reaches this level of assets, it remains eligible and cannot be taken off the PSNP (FSCB, 2007: 4). According to the Regional Guidance Note on Graduation for Oromiya, criteria for graduation are: 1) the household is able to feed itself for all 12 months of the year 2) the household includes able-bodied adult members 8 This trade-off arises because the benchmarking exercise only indicates the probability that a given household will be food secure. For instance, the authors find that a household in Tigray with assets worth 5,000 Birr per person has a 97% probability of being food secure. This is an extremely high threshold, given that most PSNP participating households in the FSP Baseline Survey had asset holdings worth less than 1,000 Birr per person (Gilligan et al, 2007: 7). On the other hand, even setting the benchmark at 5,000 Birr still risks excluding 3% of households who are food insecure. 23

30 3) the household s total assets (fully owned) plus income are worth 19,187 birr 4) at least 75% of the loan taken for the OFSP package has been repaid. Since 2006, PSNP graduation guidelines have been developed at the national level and at the regional level in Amhara, Oromiya, SNNPR and Tigray. Awareness of these graduation guidelines at other administrative levels varies. At the woreda level awareness is good most members of WFSTFs broadly understand the principles and recognise the regional threshold or benchmark from graduation. In Tigray, graduation benchmarks were sent to all administration levels in Tigregna and Amharic. In other woredas, officials were similarly able to quote the regional benchmarks for their region. However, there is less understanding about how they might apply the benchmark in practice. In Fedis, officials articulated their confusion about whether the benchmark applied to household income and expenditure, or assets. They argued in favour of income, because assets are not always liquid or profitable. a beneficiary must leave this program after he has created assets and sustainable income, and that is after he ensures food security. The benchmark is to have more than 18,000 birr income. This does not include animals that he has but production and other incomes that will add to 18,000 birr. In Lasta, the woreda officials found the regional benchmark useful but endorsed an alternative approach where benchmarks are agreed at woreda level, with technical support from the region. In Chiro, woreda officials have the graduation guidelines and know about the regional threshold for graduation. However, they raised questions about the timeframe over which households are meant to maintain their assets and income at or above the threshold level, in order to have truly graduated from the PSNP. These officials are concerned about how resilient households should be to shocks and stresses before they can be assessed as ready to graduate. At kebele level, most DAs and kebele administrations have received some training or orientation on graduation, but mistakes are common. In Kilte awlelo, there is much confusion regarding the measurement of graduation. The woreda cited a figure of 18,000 birr per year while the kushet cited 18,000 birr per month. Where the regional benchmarks are known at kebele level, there are many concerns that they are over-ambitious or unrealistic. In most places the benchmarks are significantly higher than the annual salaries of DAs and other lower-level government officials. A number of woredas, including Chiro, have identified their own graduation criteria and argued that households above a certain threshold of livestock ownership and landholding will not be considered for the PSNP. The KFSTF in Chiro also argued that graduation has three different meanings. However, only two of these imply an exit from the programme rather than in improvement in well-being. At community level there is a much more rudimentary level of understanding about the principle and operationalisation of graduation. Our survey found that only half of respondents (49.8%) have even heard of the term, ranging from 74% in Amhara to just 15% in Tigray, and that knowledge about graduation is higher among past beneficiaries (62.7%) than among current beneficiaries (47.8%) (Table 6). By far the most common source of information about graduation is the local Kebele Food Security Task Force (KFSTF), followed by neighbours, the Community Food Security Task Force (CFSTF) and woreda officials (Table 6). In many places this lack of 24

31 knowledge or limited understanding of graduation is creating tension and anxiety among PSNP participants, particularly where graduation has already been implemented (e.g. SNNPR). Table 6 Knowledge of the term graduation from the PSNP Have you ever heard of the word graduation from the PSNP? Region Tigray Amhara Oromia SNNPR Total Yes 34 (15%) 146 (74%) 107 (56%) 112 (58%) 399 (49.8%) No 189 (85%) 50 (26%) 84 (44%) 80 (42%) 403 (50.3%) If yes, from whom did you hear? Source KFSTF CFSTF Woreda office Neighbour Total Current PSNP 213 (64%) 34 (10%) 27 (8%) 48 (14%) 335 (47.8%) Past PSNP 43 (69%) 3 (5%) 8 (13%) 6 (10%) 64 (62.7%) Progress towards graduation is low in all woredas surveyed in this study for a range of reasons, not least the recent failure of rains in many parts of the country. In Chiro, five households had reportedly been graduated, but this was before the graduation benchmark had been set. According to the woreda head, these five households had met the ETB 18,000 benchmark. The woreda was preparing to graduate 170 households in 2008, based on an assessment of household assets. In Fedis, a database has been developed to track the assets of each individual registered PSNP beneficiary, and to identify households ready to graduate in the future. However, the food crisis has left the woreda highly sceptical that any graduation would be possible any role that the PSNP can play in avoiding deaths would be a significant achievement in itself. In Kalu and Lasta the woreda received how to instructions on graduation. Kalu planned to graduate 436 beneficiaries in 2008, having developed a format for assessing graduation by measuring assets and establishing woreda benchmarks, but the subsequent failure of the belg rains and food price inflation forced a reassessment of the benchmark. In Lasta, the woreda attempted to assess potential graduates based on the benchmark, but could not find anyone who qualified. Instead, they sought volunteers for graduation ( self-graduation ). In Derashe, the regional benchmark is perceived as unrealistic and no households have been identified that have managed to reach the asset status required for graduation. In Boricha, 164 beneficiaries were graduated, based on a combined income and asset assessment, but 18 of these households appealed to the woreda and the region, and 11 were reinstated on the PSNP in Other graduates had resisted graduation, complaining that they were compelled to leave the PSNP and denying that they had met the benchmark. In Tigray, less progress has been made and by June 2008 work had not yet begun at the woreda level to revise the regional benchmark, drawing on household asset and income data for each woreda. In Enderta, tabia Task Force members said that they did not have access to the graduation guidelines. The regional graduation indicator of 18,000 birr per year was not followed; instead, PSNP beneficiaries and Task Force members explained that graduation happens if your income is greater than someone on the waiting list. This was in stark contrast to woreda officials and some tabia officials who expect graduating households 25

32 to follow a linear increase in income 6,000 birr in the first year, 12,000 birr in the second year and 18,000 birr in the third. Public Works beneficiaries in Tigray said that graduation is unthinkable at this time. According to our survey, two-thirds of current beneficiaries have no confidence that they will graduate from the PSNP. This figure is highest in SNNPR (over 80%), followed by Oromia and Amhara (over 60%) and lowest in Tigray (one-third). Conversely, less than one beneficiary in ten (9%) is confident or highly confident that they will graduate from the PSNP. This figure is lowest in SNNPR (less than one in 20) and highest in Tigray (but still only one-third) (Table 7). Table 7 Beneficiaries confidence of graduating from the PSNP, by region Confidence level Tigray Amhara Oromia SNNPR Total No confidence 11 (32.4%) 114 (61.3%) 98 (63.0%) 116 (82.3%) 339 (65.7%) Limited confidence 8 (23.5%) 37 (19.9%) 19 (12.3%) 9 (6.4%) 73 (14.2%) Confident 6 (17.7%) 11 (5.9%) 12 (7.7%) 6 (4.3%) 35 (6.8%) Highly confident 6 (17.7%) 5 (2.7%) 2 (1.3%) 0 (0.0%) 13 (2.5%) Cannot tell 3 (8.8%) 19 (10.2%) 24 (15.5%) 10 (7.1%) 56 (10.9%) Total 34 (100%) 186 (100%) 155 (100%) 141 (100%) 516 (100%) Note: Only current beneficiaries (n=701) were asked this question, of whom 516 (73.6%) answered. According to Table 6 above, most households in Tigray do not seem to know about graduation, which is indicative of the patchy progress across Ethiopia with disseminating and implementing PSNP graduation guidelines and benchmarks. But the qualitative research found an alternative possible explanation. Focus group discussions and key informant interviews discovered that a diverse range of ways in which households left or exited the programme were all described as graduation in some places, while in other places there was a much narrower understanding. So different levels of understanding of the term graduation across the four regions may reflect differences in language (translation) or terminology, as well as differences in how graduation is operationalised in each region. The diverse views of graduation which are neither exhaustive nor mutually exclusive included: households leaving the PSNP voluntarily ( voluntary graduation, or self-graduation ) households removed from the list because a household member either migrated, or returned from resettlement households removed from the list for absenteeism from Public Works projects households removed from the list because their income exceeds the income of a non-beneficiary on the PSNP waiting list households removed from the list because they have acquired assets through OFSP packages and are therefore considered to be better off households removed from the list because their income is above the graduation guideline threshold (either regional, woreda or local community guidelines). While all these examples describe alternative ways in which households leave the PSNP, most of these exits do not constitute graduation according to the PIM definition. Only the final exit in this list should be labelled graduation, because it definitely shows an increase in the well-being 26

33 of households and is consistent with graduation guidelines. In other cases, the conditions of households that are removed from the PSNP could have worsened rather than improved. For example, households might self-graduate because they get a job elsewhere or because they become too ill to perform Public Works but are not transferred onto the Direct Support list. Households that migrate often do so out of desperation (including pressure to repay their OFSP package debts), and not necessarily because they have already found work. As was seen in Box 2 above, households that take OFSP packages on credit are not necessarily better off until their assets generate income streams and the loan has been repaid, so graduating these PSNP participants immediately could be premature. It is clear that different understandings persist of graduation and how it is achieved. In particular, perspectives at the kebele level show that graduation guidelines have not yet been distributed widely enough or understood well below the woreda level. In Lasta, the Woreda Food Security Task Force admitted that it had not been able to create enough awareness about graduation through orientation and training at the kebele and village levels. But improvements have been made. Targeting is now reaching the poorest households, rather than those who are most likely to graduate. The danger is that all exits continue to be described as graduation. Continuing efforts to establish, clarify and operationalise graduation benchmarks could play a critical role in ensuring that graduation is not invoked to explain away other kinds of exits from the PSNP. Table 8 summarises the different perspectives noted on three aspects of graduation (concepts, instruments, and benchmarking) at five levels (federal, regional, woreda, kebele, and village). Table 8 Level Federal Regional Woreda Kebele Village Perspectives on graduation at different levels Aspects of graduation Concepts Instruments Benchmarking Graduation occurs when a household is no longer chronically food insecure, even when PSNP support is removed Focus is on principles of secure food access, manageable credit and resilience Similar to Regional but with greater emphasis on secure access to food Graduation occurs when you: are food secure through the year and/ or are not the poorest and most needy and/ or when you don t attend Public Works PSNP plus OFSP PSNP plus OFSP PSNP plus OFSP From PSNP particularly because of Public Works benefit; occasional mention of OFSP Process for establishing asset-based benchmarks established Establishment of benchmarking in progress Stark variation in benchmarking criteria and processes; Differing levels of compliance with benchmarks imposed from above some local adaptation 27

34 CHAPTER 4 PSNP IMPLEMENTATION This chapter assesses several aspects of PSNP implementation: targeting procedures, work norms on Public Works projects, the type of transfers paid to beneficiaries and their preferences for cash, food, or cash plus food ; the reliability and predictability of PSNP transfer payments; collection of transfers (transport, risks), and livelihood packages. The chapter concludes with a review of developments in implementation in each of the eight study woredas. 4.1 Targeting General targeting process There has been significant progress in terms of targeting procedures since As shown in chapter 3.2, even at household level, more people understand the targeting system now than in the earlier years of the PSNP. In Amhara, where wealthier households were targeted in the first year of the programme in order to maximise graduation, respondents reported significant improvements in targeting and argued that the system is now more transparent, accountable and open. Access to information (often through manuals and guidelines), orientation and training have played a critical role in fostering these improvements. The improvement is particularly notable in Kalu, where all respondents at all levels confirmed that the targeting process starts from the community itself the list was agreed using community ranking and then submitted to the KFSTF. At regional level in Tigray, key informants differentiate between targeting problems associated with the early implementation of the PSNP, and other targeting shortcomings in the programme that have been more pervasive. Key informants from the Regional Food Security Task Force argue that amendments to manuals and retargeting have enabled them to ensure that the poorest households were included in the programme, but there are guidance gaps, including half versus full family targeting and the targeting of pregnant and lactating women. Where targeting guidelines have been distributed widely (for example in Tigray where 8,000 new PIMs were distributed at kushet level), there appear to have been improvements in reaching the poorest households through PSNP targeting. In Tigray, a targeting checking exercise takes place every six months, though solely to update the list so that people who have migrated out of the tabia or woreda and to replace those that have died with new beneficiaries on the waiting list. Beneficiaries are not moved out of the programme during the checking exercise based on their increased income or asset status. Elsewhere, there are still challenges: In Chiro, the key targeting change was a shift of targeting from kebele level to ganda level. However, concerns were voiced by beneficiaries that the ganda level officials were not targeting the PSNP fairly. Similarly, in Fedis, where targeting had shifted down to ganda level, there were concerns that Task Force members were targeting themselves. In SNNPR, the pressures associated with targeting are more acute, due to lower levels of PSNP coverage but very high levels of need. 28

35 An unresolved issue is the targeting and registration of polygamous households specifically, whether these are treated as a single (large) household or as a number of separate households. Three approaches can be identified. In general, all individuals who eat from the same pot are considered to be members of one family, and are registered for the PSNP as a single household even if this is a male-headed household that includes two or more wives. However, if a man has more than one wife and each wife cooks separately, this family is treated in one of two ways by the PSNP. Most commonly, especially when there are only two wives in the polygamous family, both wives and their children will be considered along with the household head as one single household even if the wives live in different houses. This is because the community considers the family to be managed by the husband, and since only he is registered as a PA member all his wives and children are registered as household members under this name. Finally, in a few instances, especially where there are more than two wives, each wife and her children is registered as a separate household, with the husband allocated to the senior wife s household. However, this is unusual. There have been cases of appeals to the woreda when a third or fourth wife has not been targeted as a separate household, nor even considered as part of the polygamous household. In those cases, the woreda was instructed to include these excluded individuals in the PSNP, though not necessarily as a separate household. It is clear that there is no consistency in the treatment of polygamous households by the PSNP. According to the Head of the Chiro Food Security Office, targeting is supposed to take account of polygamy, but practice varies not only between regions and woredas, but even from kebele to kebele. This inconsistency should be addressed standard guidelines should be developed that insist on registering each wife as a separate household, and these rules should be applied by PSNP implementers throughout Ethiopia Full family targeting In 2006, the targeting and linkages studies found that PSNP resources were being diluted by sharing them among a larger number of beneficiaries or households than intended. High levels of dilution mean that the impact of the assistance received by each household is very weak. This is of concern for PSNP implementers, because of its adverse impacts both on welfare (food consumption) and on graduation. In 2006, two types of dilution were identified. The first was the rotation of beneficiary lists (such that one household receives transfers for three months but is then replaced by a different household for the remaining three months). In 2008, there was no evidence of households or beneficiaries being rotated, even in SNNPR where the practice had been most common. The second, and more common, type of dilution was partial family targeting, where eligible households are registered but not all of their members are counted. The shift to full family targeting has largely been implemented, with the exception of SNNPR where this type of dilution still takes place. The shift to full family targeting has largely been implemented, with the exception of SNNPR where, despite the regional level allocation of the contingency fund to ensure full family targeting, this type of dilution still takes place. There are occasional errors where administrative mistakes left individuals out of a household list but on the whole woredas have successfully shifted to full family targeting. In Oromiya, Tigray, and Amhara there have been improvements toward full family targeting. There is now full family targeting in all woredas studied in these woredas. In Kalu, children born since 2006 are not included in the PSNP, the argument being that young children do not require full rations. With the exception of SNNPR, beneficiaries and non-beneficiaries appeared to accept the principle of concentrating resources on a smaller number of households. In SNNPR, low levels 29

36 of PSNP resources were juxtaposed against large food gaps in a high number of households. Very few households accepted the concentration of resources through full family targeting and wanted the resources to be spread among a wider number of households. Those who argued for full family targeting said that the transfer is so small that its impacts on hunger would be too small if it is widely diluted, and that graduation is unthinkable without full family targeting. However, the pressure on woreda, kebele and village-level officials was very great, and partial family targeting remained in some areas. As a result of this pressure, in SNNPR (Dereshe and Borecha) full family targeting has not been achieved at local level. Larger households are still only partially covered. It is not clear whether the allocation of the SNNPR regional contingency has not been applied at local level or the contingency is simply not enough to ensure full family targeting. The pressures felt in SNNPR are not unique. Elsewhere, where full family targeting is being implemented, some woredas and Kebele FSTFs (e.g. in Lasta) have requested that they be allowed to spread PSNP resources more broadly if there are no plans to increase the number of beneficiaries. A number of difficulties are associated with the operation of full family targeting. In some places there is confusion regarding elderly people and full family targeting. In a number of tabia, including in Enderta, elderly people are now counted with their adult children as part of full family targeting. It is unclear whether this is a positive change because elderly people who were previously excluded are now included or whether it has a negative impact on households which now have additional non-working members for whom public works must be completed. In Tigray, the official mechanism for allocating benefits to children under eight is based on their calorific requirements, so two children under eight years receive the equivalent of one adult transfer. The qualitative fieldwork did not find evidence of this at tabia or kushet level and there was confusion about how to provide support for small children and infants. In Boricha, some public work beneficiaries suggested that the PSNP will never achieve its objectives in particular graduation in the absence of full family targeting of households. Some non-beneficiaries accept this explanation for their exclusion, recognising that partial family targeting is unfair on the poorest households. In Enderta, by contrast, full family targeting is unpopular among non-beneficiaries particularly those who were taken off the list as a result. They repeatedly argued that the differences between targeted households and those removed as a result of full family targeting are negligible: We are all poor Benefit incidence Table 9 classifies households in our sample by income quintile and examines participation in the PSNP by income ranking. In 2006 PSNP participation was effectively random in terms of income: PSNP households were equally likely to come from the fifth (wealthiest) quintile as from the first (poorest). (On the other hand, this was a sub-sample only of PSNP participants, so it could imply that all PSNP households were more or less equally poor.) In 2008, there has been a shift towards the wealthier quintiles: 23% of PSNP participants are located in the fifth quintile against 15% in the first. One interpretation is that this reflects regressive targeting (selection of wealthier households ahead of poorer households). A more positive interpretation is that this is an indication of PSNP success in generating income gains for participants. Since this is a panel survey, many households are becoming wealthier because of their participation in the PSNP, and are gravitating towards the upper end of the income distribution. 30

37 Table 9 PSNP participation by income quintile, 2006 and 2008 Region Quintile: Poorest Richest PSNP households (20.2%) 146 (20.0%) 163 (22.3%) 147 (20.1%) 128 (17.5%) 2008 (adjusted) 106 (15.1%) 134 (19.1%) 146 (20.8%) 153 (21.8%) 162 (23.1%) Non-PSNP (20.5%) 37 (19.0%) 25 (12.8%) 36 (18.5%) 57 (29.2%) 2008 (adjusted) 53 (33.7%) 34 (21.6%) 25 (15.9%) 25 (15.9%) 20 (12.7%) Past PSNP (32.3%) 24 (23.5%) 20 (19.6%) 15 (14.7%) 10 (9.8%) Total households (20.3%) 183 (19.7%) 188 (20.3%) 183 (19.7%) 185 (20.0%) 2008 (adjusted) 192 (20.0%) 192 (20.0%) 191 (19.9%) 193 (20.1%) 192 (20.0%) Exclusion All the non-beneficiary households in our survey were asked if they feel that their exclusion from the PSNP was fair. Two-thirds (65%) believe that the decision to exclude them was not fair. The majority of these households complained, but reported that none of their complaints were successful. This suggests either that the criteria for PSNP inclusion and exclusion are not being clearly communicated to all community members and understood by everyone, or that there is substantial under-coverage and involuntary exclusion of people who meet the criteria but are left out, presumably because of budget constraints. 4.2 Work norms Two major concerns about working practices in the PSNP were assessed during qualitative fieldwork: the timing of public works; and whether beneficiaries are working an inappropriate amount of time Timing of public works There has been broad improvement since the 2006 study in the timing and flexibility of public works, though there is still room for improvement. There continues to be direct overlap with the agricultural work season in Chiro, Fedis Kalu, Lasta and Kilte Awlalo. There is more flexibility in working arrangements for example, public works stop when the rains come and the heaviest labour is required, and is rescheduled for less busy times. However, it remains the case that the timing of public works creates the risk of increasing dependency on the PSNP, by keeping working adults away from their own land and livelihoods. Figure 1 shows the average number of days worked per household on PSNP Public Works projects from January 2007 to March This figure illustrates the seasonality of employment provided by the PSNP, with most work being provided between January and June each year, and almost no public works activities occurring between September and December. 31

38 Figure 1 Mean days worked on PSNP per household, 2007/08 Evidence from other countries shows that the solution is not as simple as switching public works to outside the agricultural labour season. The agricultural labour season also corresponds the hungry months. In one case, public works took place earlier in the season and implementers faced a difficult decision about the timing of payments: They could pay at the time the work took place but risk households not having the capacity to save the food or cash until the hungry months, or created confusion among beneficiaries by not paying until the hungry months a number of months later. In Kilte Awlalo Public Works participants have to work particularly hard. They work 8 months on the PSNP and are still required to work another 2 months on other community labour. In Tigray, PSNP households complained that the demands and timing of PSNP public works are undermining their capacity to work on their own land. Proportional piling methods confirmed that Public Works participants in Enderta are doing public works projects during the two busiest agricultural months of the year. In Kilte awlelo too, PSNP public works labour is required well into the peak agricultural season Caps on working days per month In 2006, both the targeting and linkages studies raised grave concerns that some of the poorest households that had limited labour capacity were having to work full-time for six months or more in order to meet the public works obligations. Even when working full-time, they did not always meet the required input to receive transfers for all their household members and received only partial family benefits. On paper, an improvement in work requirements has been brought about by revisions in the PIM, which establish a cap on the number of days that beneficiaries or households can work. In Amhara and SNNPR, the cap is set at 20 days per month for both men and women. If there are more than two workers in a household there is a maximum of 25 days. In Boricha, the workload has been temporarily reduced because of the drought. In Kalu, there were reports that the caps had been increased to 25 days for men and 20 for women. In Tigray, men work a maximum of 20 days and women work a maximum of 15 days. In Fedis, a cap of 30 days was reported at kebele level but in Chiro there was not sufficient knowledge regarding caps. 32

39 Finding evidence that the cap is being implemented in practice is very difficult. The ways in which public works are organised in most woredas make it difficult, if not impossible, to establish how many days are worked by each person in each household. In Chiro, the team carried out key informant interviews to assess how households with limited labour capacity were treated within the public works system. It was clear in Chiro (and in other woredas with the same public works organising principles) that the 20 days per month guideline was barely compatible with the system by which public works were organised. An explanation of the organisation of public works in Chiro is found in Box 3. Box 3. Public works implementation in Chiro The operation of public works in Chiro is drawn from the previous EGS system. Public works tasks are organised and teams are formed to complete them, in order to receive their safety net transfer. The amount of time required for each task is expressed in person-days and is based on agreed norms. Indicative norms are as follows: Road maintenance: 500 person days per km Road construction: 3,000 person days per km Pitting (pits are 0.5m deep): 15 pits per person-day Planting seedlings: 50 seedlings per person-day Soil and water conservation terracing: 200 person days per km. The payment for one person-day under the safety net is 3 kg. So in principle, each person would have to work for 5 days for each dependent in their household in order to achieve the entitlement for all the household. However, PSNP transfers are paid when public works tasks are completed, not when people have worked 5 days for each beneficiary in their household. Case study: Maintenance of 4.5 km of road in Chiro Under public works norms, the maintenance of 4.5 km of road requires 2,250 person-days and will provide a total of 6,750 kg of food; equivalent to the monthly entitlement of 450 people. Therefore, 90 households, each with 5 members, were organised into a public works team to complete this task. Teams were always constructed in this way. Workers all came from households of the same size but household composition could vary. Some households had 2 adults and 3 dependent children (we call this household type A), while other households had 1 adult and 4 dependents. Each household designated one person to be a member of the public works team. In theory, each member would then contribute 25 person days to the team effort. But the actual number of days varied depending on an agreement within the team to work faster. Working faster required more workers and teams could agree that every member would bring an additional worker. Household type A brought the second adult household member. For household type B, either children were deployed to work teams or a system of guza was used, whereby a friend or neighbour completed the work in return for food or chat. The amount of food that was exchanged is small (perhaps a plate of injera rather than ½ the PSNP transfer of 3kg per day) but these labour transactions are usually based on reciprocity (i.e. in another month, the relationship will be reversed). Source: Key informant interviews, Chiro Practices to limit the number of days included: allocating only 3 or 4 set days each week for public works. On a 3-day-week basis, public works can only be carried out for 15 days maximum, less in months where more 33

40 days fall on weekends. In Chiro, households worked only on Mondays, Wednesdays and Fridays. arranging public works teams composed of workers from the same size of household so each worker has the same work commitment. The team can then be allocated a specific size of task (1 km of road or 100 m of terracing) that is achievable within the number of days that each team member must work. However, on the ground, there were many cases where the cap was not upheld. The most common reason was that established work norms did not match up with beneficiaries experiences of completing difficult work. Terracing took much longer than the work norms allowed for. In Derashe, public works participants reported that they often worked more than the 20 day cap in order to complete the task. In Enderta, for example, respondents complained that they sometimes worked longer than 5 days for each participant to finish the work norm and their public works had over-run into a seventh month because they had not completed the required output. Work teams often sought to finish tasks faster than the allocated time by bringing in additional labour. For households with more labour, this was much easier that for those with limited labour capacity. The latter were forced to buy in labour from friends or neighbours. While the wages or guza (reciprocal support) underpinning this labour were small, they reduced the impact of the PSNP transfer in beneficiary households. Together, these processes meant that, while there has been some success in implementing a cap, the PNSP public works system is still very tough on households with limited labour capacity. The focus on maximising public works outputs, rather than ensuring a fair, reasonable and equitable contribution by participants, stems in part from the leaning at woreda and kebele levels towards maximising public works and the developmental aspects of the programme over and above core welfare objectives. The key finding regarding public works norms is that, as long as they continue to follow an output-based approach, that is, one based on length of road constructed, number of seedling plants etc, they are likely to continue to discriminate against needy but labour-poor households. In Tigray, key informants at the regional level shared these concerns. They recognised that, while there were improvements in the implementation of the PSNP programme, and a good understanding of the outputs of the programme, much less is known about the actual outcomes of the programme. The focus on measuring and reporting public works exacerbated the focus on outputs rather than outcomes. With regard to implementation, there is improvement in both quality and quantity. But what we see is the output and we have to be able to see the outcome as well. And we do not keep data to evaluate the outcome. It is important, therefore, to ensure that the output approach does not create wrong incentives in programme. Predictability of transfers and the welfare or direct food security objectives of the PSNP should not be compromised by Public Works expectations and arrangements, especially when Public Works affect the predictability of transfers paid to Direct Support beneficiaries. 34

41 4.2.3 Unified lists and Direct Support quotas The use of unified lists, alongside the abolition of the quota for Direct Support, is important for two reasons. First, they ensure that people are targeted for inclusion in the PSNP based on their food gap and not on their capacity to work. Second, households that are temporarily unable to work, for example pregnant and lactating women, can be moved between Direct Support and Public Works. Since 2006, unified beneficiary lists have become part of the revised PIM. Overall this change has been implemented. Most woredas have broadly followed the targeting process as laid out in the PIM and the Technical Guidance Note and have developed unified lists of beneficiaries. This has made it easier to transfer Public Works beneficiaries to Direct Support. In Amhara and Oromia there is a unified list of beneficiaries. In Fedis, the targeting process began with a wealth/asset ranking, before beneficiaries were split into Direct Support and Public Works. In practice, the kebele still ends up with two lists (one Direct Support and one Public Works), but they are able to change an individual s status from Direct Support to Public Works, or vice versa. In Tigray and in SNNPR, the targeting process starts with a wealth ranking exercise which leads to the creation of a unified list and then household labour is assessed. The formal and explicit quota or cap on the number of Direct Support beneficiaries (maximum 20% of total PSNP caseload) that was evident in 2006 is less prevalent in In Tigray there is no official quota and in some cases, the proportion of Direct Support beneficiaries has increased (e.g. for example REST woredas now average 13%, up from 11% in 2006). There are, however, ongoing pressures from the regional government to ensure that Direct Support beneficiaries are the most deserving, amidst concern about inclusion errors and dependency. In Oromiya, there is no official quota for the Direct Support : Public Works ratio. In Fedis the woreda instructs kebeles not to select more than 20% for Direct Support, but the actual number is 29%. In Chiro, there is no quota. In Amhara, officials claim that there is no quota and in some kebeles in Lasta Direct Support has increased to 25% of beneficiaries (Bugna woreda averaged 16% in 2006). In SNNPR there are reports of a quota of 16% at regional level, but not at woreda or kebele level. The abolition of quotas appears to have had two different effects across the woredas. In some places, for example Lasta and Kalu, there have been increases in Direct Support. In Kalu, this suggests a stronger focus on the welfare function of the PSNP, including kebele-level autonomy in decisions over the proportion of Direct Support beneficiaries. These general increases in Direct Support in some woredas contrast with a range of pressures on woreda and kebele officials to minimise Direct Support beneficiaries. In some places, the proportion of beneficiaries receiving Direct Support remains very low or has become lower. Maximising the outputs from public works and concerns about long-term dependency on the programme are at the heart of pressures to keep Direct Support beneficiaries down. 35

42 In other places, including Fedis, Chiro, Enderta, Kilte Awlelo, there is frequently a much lower level of Direct Support measured at woreda level than when the quota was used as a quasiguide. In some kebeles, Direct Support beneficiaries constitute less than 1% of the kebele population. Box 4. Improvements in unified lists and Public Works : Direct Support ratios in Lasta In Lasta (Amhara Region) it was reported that, rather than having a Direct Support quota imposed by the woreda, the kebeles had greater flexibility in deciding the quota for Direct Support. This has enabled the kebeles to have a single unified list of PSNP beneficiaries incorporating both Public Works and Direct Support components, so enabling flexibility in moving people between Public Works and Direct Support as appropriate. In kebeles with a high number of chronically ill, elderly, orphans, disabled and pregnant or lactating women, the Direct Support component had to be increased. If the situation in the kebele is the opposite, the Direct Support component is smaller. This is a positive way in which the welfare or direct food security objectives of the PSNP can be maximised, but this will only work if kebele-level officials are focused on welfare rather than pursuing maximum Public Works outputs. In SNNPR the situation is different and kebeles have been given autonomy to identify Direct Support beneficiaries based on need and capacity. There remain divergent views about whether the way forward is a unified list or two separate but linked programmes for Public Works and Direct Support. A number of regional key informants note that long-term social protection will be required for Direct Support beneficiaries who will not be able to graduate from the PSNP. It may not be in the best interests of those beneficiaries to continue to be linked to a programme that is tied to public works. For example, when Public Works transfers are late, Direct Support beneficiaries receive their transfers late too. In Fedis, Direct Support beneficiaries reported that We do not get the transfer on time. We have to wait until the Safety Net workers finish their work. At the same time, however, separate programmes and the existing quotas for Direct Support will make it more difficult to maintain the flexibility required to meet the needs of those who are unable to work in the short term for example, pregnant and lactating women Treatment of pregnant and lactating women under the PSNP Evidence from most woredas suggests that good progress has been made in ensuring that pregnant and lactating women are no longer forced to undertake public works in order to receive PSNP transfers. In almost all cases, women now stop working at the sixth month of pregnancy and do not return to public works until 10 months after the birth. The least progress has been made in Derashe where the kebele officials had just been trained regarding the new policy but were only in the early stages of implementation. Beneficiaries recognise the public health benefits of the new approach and also argue that it means fairer working conditions for others doing public works, who are no longer expected to take up the slack for pregnant and lactating women. Concerns were also voiced in some woredas that full family targeting and the switch of pregnant and lactating women onto the Direct Support component of the programme are leading to increasing population pressure. There is no evidence of this effect, and it is not compatible with the view of beneficiaries in Lasta that the size of PSNP transfers only barely meets a six-month food gap, while most households have a nine-month gap. However, in Lasta, one KFSTF was 36

43 so concerned about the length of time that women would be on Direct Support that they did not implement the full 10-month lactation period. In Fedis, while the official position is that pregnant and lactating women should not work, there is contradictory evidence in this regard. A women s focus group discussion claimed that all pregnant and lactating women were still working. There appear to be two reasons for this. First, in households with two adults, the total work requirement is not reduced if women are taken out of public works when pregnant or lactating. Rather, the remaining adult is expected to fulfil the work requirement for all other household members. Second, the system for implementing public works is forcing women to work while pregnant or lactating. Works are issued to individuals rather than teams at kebele level. As a result, family members, including children and pregnant or lactating women, have to support the assigned household member to complete their task. One respondent explained that a pregnant woman s husband will take her to the field, because she has to help him finish the measured work given to him in order for them to get the transfer. In this case, the establishment of a unified beneficiaries list did not help to ensure that women were not counted under Public Works. 4.3 Transfers Types of transfers The PSNP introduced an innovative approach to social assistance in Ethiopia, namely the disbursement of cash transfers rather than food aid. This was not done universally throughout the programme but was restricted initially to woredas with the administrative capacity to manage cash transfers, and where markets were assessed as resilient enough to respond to the incremental demand pressure that cash injections into poor rural communities would generate. The expectation was that cash transfers would gradually displace food transfers in other woredas as administrative capacity and markets strengthened, and as the expected positive impacts from PSNP cash woredas generated lessons for the wider programme. In our 2006 survey, 81% of PSNP beneficiaries received payment in the form of cash, either in full or in combination with food. Only 19% of households received food only (Figure 2a). In the 2008 survey, the proportion of beneficiaries receiving cash plus food has fallen from 66% to 53%, while the proportion receiving cash only has risen from 15% to 21% (Figure 2b). Over the same period, PSNP beneficiaries receiving food only have not declined as planned, but have increased from 19% to 26%. 37

44 Figure 2 Type of transfers actually received by PSNP households, 2006 and 2008 (a) 2006 (b) 2008 As a rule, Public Works participants and Direct Support beneficiaries in the same community receive the same form of payment at the same time. (In the 2006 survey this synchronicity raised a major concern, as delays in implementing Public Works projects resulted in delayed payments for Direct Support households as well.) This practice can be seen in the close correlation between the proportion of Public Works and Direct Support households receiving cash only, food only and cash plus food transfers, both in 2006 and in 2008 (Table 10). Table 10 Type of transfers received by PSNP households, 2006 and 2008 Transfers Public Works Direct Support Public Works Direct Support Cash only 82 (14.6%) 31 (16.1%) 111 (20.4%) 33 (21.9%) Food only 118 (21.0%) 29 (15.1%) 150 (27.5%) 34 (22.5%) Mixed (cash + food) 362 (64.4%) 132 (68.8%) 284 (52.1%) 84 (55.6%) Total households 562 (100%) 192 (100%) 545 (100%) 151 (100%) PSNP households participated in the programme between January and August in 2007, after which the PSNP was dormant for four months between September and December, before it resumed in January 2008 (Figure 3). 38

45 Figure 3 Households receiving cash or food from the PSNP, by month Beneficiary preferences There has been a significant shift in preferences by PSNP beneficiaries for different forms of payment since the 2006 survey. Two years ago, almost half of all beneficiaries surveyed (45%) expressed a preference for receiving all or part of their transfers from the PSNP in cash either as a combination of half cash and half food (36%) or cash only (9%) (Figure 4a). By 2008, this preference for cash has shrunk to just less than one in six (16%), and the proportion of beneficiaries choosing food only has risen from just over half (55%) to more than four in five (84%) (Figure 4b). These findings are disappointing from the perspective of one of the secondary objectives of the PSNP to shift the delivery of social assistance in rural Ethiopia away from food aid towards cash transfers. Figure 4 Preferences for types of assistance from PSNP (a) 2006 (b) 2008 The qualitative fieldwork also uncovered a widespread desire among beneficiaries to receive food rather than cash in current cash plus food and cash only woredas, due to increasing food 39

46 prices (Box 5). In Lasta, households receive a combination of 4 food payments and 2 cash payments. Beneficiaries claimed that in 2008, the cash transfer only enabled them to buy half a wheat grain ration. In Fedis, there has been a shift from cash plus food to cash only despite concerns at woreda level that this is not in the best interests of beneficiaries because of rising food prices. In 2008 in Fedis, the price of grain escalated from birr per kadiye up to 5 birr. According to respondents, this was far above the normal seasonal variation. Due to the increase in the price of food items, beneficiaries are increasingly requesting to change from cash to food transfers. As a result of these requests, 11 woredas have changed from cash to food transfers. In these woredas there have been difficulties with the re-establishment of transport and import logistics Value of PSNP transfers Average transfers of cash and food to PSNP households were substantially higher in 2007 than in Most of this increase was driven by the massively increased value of food transfers, due to higher food prices an average of 4.1 Birr/kg, as opposed to 1.7 Birr/kg in The market value of food transfers was more than three times the value of direct cash transfers in For example, cash transfers to male-headed households averaged 482 Birr, whereas food transfers averaged 1,682 Birr 3.5 times higher (Table 11). If 2006 food prices were used to value food transfers in 2007, the difference would be much less drastic food transfers would be worth 697 Birr, just 1.4 times higher than cash transfers. This largely explains the sharp shift in beneficiary preferences towards food transfers reported above. It also demonstrates conclusively that the value of direct cash transfers is inadequate and has fallen dramatically in real terms in terms of food purchasing power. PSNP planners should take this finding very seriously, because it compromises the programme objective of shifting towards cash transfers and breaking Ethiopia s chronic dependency on food aid. Table 11 Total value of cash and food transfers to PSNP households, 2007 Category Household type Total cereal for household (kg) Cash value of cereal (Birr) Total cash received (Birr) Total value of food and cash received Male-headed , ,809 Female-headed , ,157 PSNP status Cash only n/a n/a Food only ,364 n/a 2,364 Cash + food , ,450 Note: Calculations of cash value of food are based on an average cereal price of 4.1 Birr/kg The changing value of food and cash transfers also affects the calculation of what proportion of their entitlements under the PSNP participants actually received. If each individual is entitled to a maximum of 30 days at 8 Birr, this amounts to 240 Birr per capita each year (it was 180 Birr in 2006, when the daily transfer was set at 6 Birr). Cash only recipients received two-thirds of 40

47 their entitlement in 2007 (166.1 Birr = 69% of 240 Birr). Conversely, recipients of cash plus food received one-third more than they were entitled to (316.8 Birr = 132%), while food only recipients received almost twice their entitlement, in cash equivalent terms (455.9 Birr = 190%) (Table 12). This calculation shows vividly the effects of food price inflation and the failure of the programme to increase cash transfers sufficiently to match inflation on the transfer value. Table 12 Total value of cash and food transfers to PSNP households, per capita, 2007 Category Household type Total cereal for household (kg) Cash value of cereal (Birr) Total cash received (Birr) Total value of food and cash received Male-headed Female-headed PSNP status Cash only n/a n/a Food only n/a Cash + food Note: Calculations of cash value of food are based on an average cereal price of 4.1 Birr/kg 4.4 Reliability and predictability of PSNP transfer payments Changes in the delivery of the PSNP are common in many regions and woredas as the programme is revised and refined. In Tigray, PSNP beneficiary numbers remained more or less the same in 2006, 2007 and 2008 but there were significant changes within the beneficiary caseload. Households switched between Direct Support and Public Works, and households moved on or off the programme to be replaced by others. In other woredas, there have been shifts in payments between different years of the programme. In Tigray, there is a difference in the type of transfer. Some beneficiaries are receiving cash only, some receive cash plus food, while some receive food only. The initial allocation was based on woreda preferences. During the first year of PSNP implementation five woredas decided to transfer benefits in cash, including Enderta. Nine woredas received cash and food, and 17 woredas received cash. By 2007 there was a significant shift towards transferring cash: 25 out of 31 woredas included in the PSNP received cash transfers. In some woredas, there has been a shift back to food transfers in response to food price inflation in 2008, but in the majority of cases the shift towards a cash only approach continues. A critical issue with the general shift from food to cash transfers is predictability. Unfortunately, the qualitative fieldwork found that there have been only limited improvements in the timing of transfer delivery since In all regions, public works beneficiaries know the amount and type of transfer they will receive but there is great uncertainty about when the transfer will come. Knowledge is patchier among Direct Support beneficiaries as these include elderly and disabled beneficiaries. In the Tigray, Amhara and Oromiya woredas, the transfers are intended to be on a monthly basis. In practice, 41

48 however, transfers are often rolled together when there is a delay. In SNNPR, capacity constraints mean that transfers were planned to be paid every two months. In Kalu, there are still problems with the timing of payments. While a uniform recording and reporting system has been established for DAs, this has not yet enabled the processing of payments to happen more efficiently. In Fedis there were also delays usually of days each month. The shift from food to cash was difficult because the woreda did not have the logistics systems in place to make the change. In Kilte Awlelo, the evidence shows that the timing of transfers is improving but still poor. Households prefer receiving transfers every month, suggesting that their priority is meeting basic household needs rather than creating assets or investing in livelihoods. In Enderta, the experience is similar, with some limited improvements in the regularity of transfer delivery. In both woredas in Tigray, it is clear that a major challenge to the predictable payment of transfers every month is the complex arrangements for measuring, signing off and reporting progress on public works. One key informant explained that delays in payments to beneficiaries result from delays in the completion of field activities and attendance records, and the overlapping of different public works tasks or activities. Officials are concerned about the amount of red tape in the PSNP system and in Public Works in particular (Box 5). At woreda level many different officials are required to approve Public Works records to trigger subsequent transfer payments to beneficiaries. At the regional level, officials report that efforts to make the transfer available at the right time had not been successful. While the federal Bureau transferred the capital budget and the first two months by December, the transfer still did not reach the woreda until March. In SNNPR, there continue to be problems with the timing of transfers. In both Boricha and Derashe, the planned payments schedule corresponds with the hungry months in poor households, but respondents report that transfers were never on time, arguing that payments reach us after we are all on the verge of death (Public Works participants, Boricha). In Lasta, there have been significant improvements in the delivery of transfers. In 2007, transfers were made in five months February, April, June, August and October with the final transfer being for two months of payment. In 2008 each payment was made on time on a monthly basis between February and July. Progress in Lasta appears to have been made through good diagnosis of the bottlenecks in the payments system and appropriate steps to address the bottlenecks. Officials argue that previously the transfers were not predictable because of the late release of the budget and lengthy processes of attendance confirmation and payroll approval. The woreda appears to have solved these problems and was able to make payments on time. They achieved this by cutting some of the chains in the Rural Development and Agriculture office to shorten the procedure for approving the (consolidated) attendance sheets and payroll. It was argued that officials at woreda level have gained experience that enables them to better manage the attendance sheet in particular. The release of the regional budget on time also helped the process. 42

49 Box 5. PSNP payment and public works arrangements in TIgray The role of the Office of Food Security at regional level is to write to the Bureau of Finance and Economic Development (BOFED) in the case of cash transfers, and to the Disaster Prevention and Preparedness Agency (DPPA) in the case of food transfers, to make the transfer to the woreda. The Office of Food Security then writes to BOFED with a copy to each woreda indicating the amount to be transferred to each woreda for capital budget and cash transfer. Then BOFED send cash to its woreda line offices. The Public Works attendance lists are submitted by the kebele to BOFED, then BOFED prepares the payroll and completes the transfer. In the case of food transfers the system is different and is handled by the DPPA. With regards to REST project sites the transfer is made by REST and a report is submitted to the Office of Food Security. Key informants claim that, while the presence of monitoring and evaluation, focal point officials and the rapid response team (RRT) to monitor Safety Net activities all contribute to the success of the programme, there are also serious challenges related to delays in reporting. In particular, accessing transport to enable the region to monitor woreda activities and accessing motorbikes for monitoring field activities in a timely way is a key constraint. The findings about the timing of transfers in the PSNP are not all negative. In Fedis, despite problems with the predictability of payments, beneficiaries prefer to work on the PSNP than either other NGO or government projects or seeking casual work. In providing six months of support, there is greater continuity on the PSNP, and the availability of other casual wage labour can be problematic. In SNNPR, the fact that beneficiaries are able to use their PSNP membership as collateral against loans suggests that, for moneylenders at least, the PSNP is seen as reliable. This finding is supported by similar experiences in Fedis and Chiro. In Lasta, when improvements were made, they had a significant impact on beneficiaries. In 2008, when payments were made on time, PSNP beneficiaries were able to get credit from their neighbours on the understanding that they would repay when the payments came. The timeliness of PSNP transfers prevented late repayments of credit. Previously in Lasta, it was reported that credit systems were undermined as potential creditors had concerns about government willingness to pay the PSNP. A number of other effects were also attributed to the reliability of payments in Lasta, including reductions in food shortages, asset depletion, and the procurement of credit at high interest rates. At the same time, the availability of credit from shops increased. In Chiro, beneficiaries perceive the PSNP as reliable but not predictable. They know what transfers they will receive, in what form, but they do not know when they will receive them. However, in common with other woredas, beneficiaries use the PSNP in order to take small consumption loans from neighbours. Their neighbours know that they will be repaid when the PSNP transfer is made. Table 13 presents subjective findings from the household survey on beneficiary perceptions about the delivery of PSNP cash and food transfers. Most beneficiaries do receive payment in full (66% agree), but not necessarily on time (56% disagree). Respondents are divided as to whether payment reliability has improved since the PSNP started (38% agree, 47% disagree), presumably reflecting improved performance in some woredas but not others. There is a strong preference for payments to be made every month (91% in favour) rather than a double payment 43

50 every second month (47% in favour). Probably because of the unreliability of payments, most respondents do not believe they can plan ahead knowing that PSNP payments will be made. There remains much room for improvement in this aspect of PSNP implementation. Table 13 Beneficiary perceptions on reliability of PSNP payments I always know when and where to go, to collect my payment from the Safety Net Programme The Safety Net Programme always pays me in full The Safety Net Programme always pays me on time, every month Payment used to be unreliable when the Safety Net Programme started, but it has improved I can make plans for some months ahead, knowing that I will get paid by the Safety Net Programme It is best to get paid every month, when the Safety Net Programme is running It would be better to get paid a double payment from the Safety Net Programme, every second month Agree strongly Agree Neither agree nor disagree Disagree Disagree strongly 107 (15%) 234 (33%) 85 (12%) 204 (29%) 71 (10%) 207 (30%) 250 (36%) 55 (8%) 161 (23%) 21 (3%) 59 (9%) 163 (23%) 85 (12%) 318 (46%) 71 (10%) 63 (9%) 205 (29%) 97 (14%) 254 (36%) 78 (11%) 50 (7%) 141 (20%) 99 (14%) 285 (41%) 121 (17%) 419 (61%) 206 (30%) 26 (4%) 27 (4%) 11 (2%) 101 (15%) 221 (32%) 45 (7%) 168 (25%) 147 (21%) The overarching finding regarding the timing of transfers is that transfers are dependable but not predictable: transfers are often not made on time but they do eventually arrive. This means that they are reliable enough to be used as collateral in credit arrangements, but not predictable enough that poor households can avoid taking credit in the first place. According to our survey, almost three beneficiaries in four (71.3%) reported that they had encountered delays in PSNP payments in These delays had negative implications for most of the affected households; only one in six (16.5%) said that there were no negative consequences. Several responses to delayed payments by affected households were recorded. These are very similar to coping strategies or responses to poverty and hunger by food insecure households. Half of all affected households rationed their food consumption (51.5%), one in ten borrowed cash to buy food (10%), slightly less than this sold livestock to buy food (8%) or consumed household food stocks (7%) (Table 14). The linkages study in 2006 highlighted the ways in which predictability was important for PSNP beneficiaries. In 2008, predictability was found to be similarly important. When payments were made in a timely manner, households were able to appropriately plan their expenditure, including investments. 44

51 Table 14 Implications of delayed payments for PSNP households Implication Number % Reduce amount of food consumption per day % No implication at all % Borrowed cash from neighbours or relatives to buy food % Unwillingly sold livestock to procure food % Consumed household food stocks % Obtained food from neighbours or relatives for free % Family members went for paid labour 7 1.4% Other % Total % Although transfers are often delayed, beneficiaries know that they will be delivered. Over the past few years, this has built confidence and trust that the PSNP will deliver, allowing households to plan and act with a greater degree of confidence. It has been critical in giving households the confidence to hold onto, rather than sell, productive assets during difficult times. Late payments in various woredas have limited the impact of the PSNP as an income smoothing device. To cope with late payment in Derashe and Boricha, people took credit in most months in the place of timely transfers in order to meet their household s needs. High interest loans are not ideal. While they may tide over a household until the PSNP payment arrives, when the PSNP is paid against interest, less money is left for expenditure on food. 4.5 Payment collection Transport and travel time Almost all PSNP beneficiaries (current and past) travelled to their payment collection points on foot (776/803 =97%). One in five used pack animals, such as donkeys, on some occasions (21%), either their own animals or hired. One in four paid for public transport (13%) or private transport (12%). Walking to PSNP collection points took the longest time. For half of these beneficiaries (51%), the journey took longer than an hour each way, while for a quarter (26%) the walk took over two hours each way. One respondent claimed that the return journey took them eight hours. Pack animals were used to reduce the travel time, and to carry food home (either food rations or food purchased with cash transfers). Only one in five respondents who used pack animals took longer than an hour to reach their collection point (61/302 =11%). Motorised transport took the shortest time less than an hour in almost all cases (1/202 =99.5%). A relatively high proportion of current beneficiaries spent a night away from home during the last 12 months, while collecting their PSNP payments (18.4%). This is a concern because it has implications for HIV transmission, as discussed elsewhere in this report. Transport costs are inversely related to travel times. Walking takes the longest time but costs nothing. Using pack animals is slower than motorised transport, but cheaper. For beneficiaries who paid to reach their collection points, the average cost was 5.9 birr. One person paid 80 birr 45

52 for a return journey to collect and transport PSNP food rations back home. This is extremely expensive, recalling that the PSNP cash payment in 2008 amounted to just 8 birr per day Risks associated with collection of PSNP payments PSNP beneficiaries were asked to report on various risks they might have faced while collecting cash or food payments. Reported incidents of personal abuse or loss of property were very low, both for personal experiences and for knowledge of these negative events occurring to other PSNP beneficiaries. Only two risks were reportedly experienced by more than one in a hundred beneficiaries surveyed: verbal abuse at collection points, and theft of PSNP food or cash (2.6% each). Theft of other personal property was reported by a few beneficiaries (0.5%), drunkenness and physical violence by even fewer (0.29% each), and one woman reported an incident of sexual abuse (0.14%). Although any abuse or loss of property is regrettable and should be minimised as far as possible, these percentages are encouragingly low, suggesting that little action is required. Any programme staff who are found verbally abuse claimants should be dismissed or at least trained to improve their attitude and behaviour, and security at collection points should be assessed and tightened up in places where theft has been reported. 4.6 Livelihood Packages One in five of all current beneficiaries in our survey (21%) claim to have received a livelihood package loan under the Food Security Programme since the PSNP started. One in eight past beneficiaries in our survey (12%) said that they had received a package. In our 2006 survey, only one PSNP beneficiary in sixteen surveyed had received a livelihood package (49/764 =6.4%). This suggests that adoption of livelihood packages is increasing over time: from 6% to 12% to 21% between 2006 and Disaggregating current beneficiaries that have taken livelihood packages by income quintile reveals a bias towards wealthier households. More than half the beneficiaries in the two wealthiest quintiles have taken packages (61% of the top quintile), compared to only one in five of the poorest quintile (20%) (Table 15). Table 15 Current beneficiaries receiving livelihood packages, by income quintile Quintile Households receiving packages % of quintile % of all households receiving Total households in quintile % of current beneficiaries % 7.2% % % 13.3% % % 18.1% % % 27.7% % % 33.8% % Total % 100% % Adding current and past beneficiaries together, a total of 318 households in our sample have taken a total of 380 packages. The four most popular packages are all related to livestock dairy, sheep rearing, sheep or goat fattening, and oxen fattening. The fifth most popular package is beehives for honey production. These five packages account for 90% of all packages taken over the four years from 2005 to 2008 (Table 16). 46

53 Table 16 Livelihood packages taken by PSNP beneficiaries, by year Package Total packages Dairy 26.8% 35.0% 20.5% 18.9% 31.5% Sheep rearing 26.3% 31.7% 34.9% 25.4% 12.3% Sheep/ goat fattening 13.5% 13.3% 15.8% 23.0% 20.6% Oxen fattening 11.5% 13.3% 15.1% 18.0% 1.4% Beehives 10.8% 1.7% 2.1% 5.7% 31.5% Poultry 4.3% 3.3% 1.4% 3.3% 2.7% Seed 1.5% 1.7% 3.4% 1.6% 0.0% Marketing 1.5% 0.0% 0.7% 1.6% 0.0% Water harvesting 0.5% 0.0% 0.7% 0.0% 0.0% Fruits 0.3% 0.0% 0.7% 0.0% 0.0% Vegetable 0.3% 0.0% 0.7% 0.0% 0.0% Water harvesting 0.3% 0.0% 0.0% 0.0% 0.0% Irrigation 0.0% 0.0% 0.0% 0.0% 0.0% Other 2.6% 0.0% 4.1% 2.4% 0.0% Total 100% 15.0% 36.6% 30.6% 18.3% Several benefits were reportedly derived from the Livelihood Packages. The most common is increased livestock ownership (reported by 63% of respondents who took packages), a function of the livestock (cattle, sheep and goats, poultry) that are transferred directly by the packages. The second most common benefit is increased household income (41% of respondents), a secondary effect that suggests the packages are generating profitable income streams for a significant proportion of beneficiaries. No other benefit was reported by more than 10% of households. Nonetheless, several respondents reported that the packages had enabled them to send their children to elementary or high school, or to construct a house (Table 17). Table 17 Self-reported benefits from livelihood packages Benefits Households % Increased livestock holding % Increased annual income % Sent my children to elementary school % Constructed a house % Gained business capital in the form of cash 8 2.7% Increase the size of land cultivated in recent years 7 2.4% Sent my children to high school 4 1.4% Gained business capital in physical assets (e.g. farm tools) 4 1.4% Other % Total % 47

54 4.7 Implementation issues at woreda level There are many overlaps in findings between different woredas, but in each there is a unique story about progress and remaining challenges in the PSNP. The key elements in each woreda are discussed below Fedis, Oromia Fedis is one of a number PSNP woredas that has gradually shifted from food to cash. In 2007 it distributed a combination of food and cash that in Only cash was distributed despite implementers concerns about whether the switch to cash only was in the interests of beneficiaries, because of rising food prices. Acute price rises were anticipated in Fedis as the woreda is not a surplus producer and must procure grain from elsewhere even in good years. PSNP operated in all kebeles in the woreda and covered about 27% of the total woreda population. Direct support constitute close to 11% of the PSNP caseload with higher proportions in some kebeles. Beneficiaries in Fedis stress the critical importance of the PSNP in protecting their existing assets. They suggest that, without the PSNP, they would have entered into distress sales of their few livestock in order to meet household needs when the belg rains failed in The focus of policy and implementation changes in Fedis is maximising the development potential of the PSNP. Objectives include establishing a stronger work ethos among beneficiaries, tracking the accumulation of assets by beneficiary households and concentrating resources in a smaller number of households through for family targeting. Progress towards these objectives is undermined by three main issues. First, there are capacity and logistical constraints and a lack of lesson learning in the implementation process at woreda level. Second, graduation is undermined at the implementation level by the failure of transfers to reach beneficiaries, in a timely way. Much time is spent verifying incomplete attendance lists from public works, which delays payments. Third, there are grave concerns that households still have very limited capacity to respond to certain shocks and that without greater resilience they will not truly graduate. The main change in targeting in Fedis has been a shift from kebele to ganda level targeting. At this stage it does not appear that the change has improved accountability and transparency. Community consultation meetings do not appear to have successfully improved transparency as is the case in other woredas. In the appeals process, beneficiaries and non-beneficiaries continue to be shunted between different levels of the appeals structure Chiro, Oromia At 11.2%, Chiro has much lower PSNP coverage than other woredas in the study. Direct support, coverage is also low, at 6% of the PSNP and less than 1% of the total woreda population. The contingency fund had been used in 2008 to support one additional kebele when the belg rains failed and a small additional caseload of households whose land had been flooded. Beneficiaries focused on the consumption and protection aspects of the PSNP. Women argue that the PSNP means that You don't sell the boat that is in front of you. Instead, you are eating 48

55 with help from the government. the capacity of the PSNP to support beneficiaries in this way is undermined by a lack of predictability. This has been caused by the late receipt of food and funds by the woreda. There is limited evidence that beneficiaries perceive the PSNP as reliable as they use it takes more consumption loans from neighbours. While the targeting process had shifted from kebele to ganda level, the community food security task force did not have a clear understanding of targeting and beneficiaries were complaining about the targeting process. There were significant tensions between the community and the ganda level chairperson. The grain store of a task force member had been burned down by nonbeneficiaries who were unhappy with the targeting process. Some households had been graduated from the programme but closer inspection suggested that they had simply been taken off the programme rather than experiencing any significant or sustainable improvement in welfare or livelihoods Kilte awlelo, Tigray At 67%, PSNP coverage is high in Kilte Awlelo compared to woredas in other regions. Beneficiary numbers have remained stable over the last few years but there have been significant changes in the timing and size of benefits. Households now receive 13 kg of grain each month per beneficiary for a total of 8 months (compared to 15 kg for 6 months). On an annual basis this constitutes 104 kg per beneficiary as opposed to the norm of 90 kg. This is an effective concentration of PSNP resources on the basis of 90 kg, an additional 12,000 beneficiaries could be included in the programme. In 2006, the linkages and targeting studies expressed concerns that poor households were being excluded from the PSNP unless they took an agricultural package. In 2008, evidence was found showing that the process continues, despite woreda warnings to kebeles that it should cease. Households with a household member seeking work in town were also often removed from the list, even if the search for work was unsuccessful or no remittances were sent. There have been improvements in the accountability and transparency of targeting though households appealing at community consultations are forced to name another household that they should replace on the programme. This process is having a negative impact on community relations. There continue to be some problems with the timely delivery of transfers. Officials put these delays down to the complex public works reporting systems and high levels of staff turnover and red tape at woreda level Enderta, Tigray High coverage levels and a very high proportion of public work beneficiaries (97%) in Enderta point to an understanding of the PSNP that focuses very much on the developmental aspects of the programme the creation and protection of productive assets, linkages to other food security programmes such as agricultural credit packages rather than the direct welfare and food security impacts of the transfers. In common with other woredas, payments are not predictable. Delays are attributed to the complexity of recording public works outputs and attendance and the large numbers of officials at woreda level required to authorise the payroll. The public works system is an output-based system it focuses on the assets created rather than the number of days that public works participants input. Completing the tasks often takes more days that are required under the PSNP. Beneficiaries are reluctant to raise grievances about this and other aspects of the PSNP 49

56 system, arguing that that should eat what they give you, hear what they tell you, rather than questioning the system Boricha, SNNPR Boricha is a highly food insecure woreda in which beneficiaries receive three months of food and three months of cash under the PSNP. There are 40,911 beneficiaries of whom 2,300 are covered by the direct support component of the programme. There is a good understanding of the principles and objectives of the PSNP with a leaning towards livelihoods rather than welfare and the encouragement of a work culture among beneficiaries. Despite the emphasis on public works, there is some flexibility in the arrangements for works and tasks are organised based on each household sending only one person. Pregnant and lactating women receive direct support this change is a direct result of successful PIM training. Key challenges in Boricha include pressure on officials to dilute PSNP resources among a larger number of households with food gaps and how to balance this against providing households with a level of support that could enable graduation. Partial family targeting remains common and household size rarely matches up with the benefits received. There are a very small number of examples of people transforming their lives through the PSNP but generally the graduation benchmarks are difficult to apply and unrealistic at community level. Prospects for graduation are also undermined by unpredictability in the payment system which forces households to take credit at high interest rates. In 2008, 11 appeals were heard from households that had been graduated. Eight of the appeals were upheld and the households were returned to the programme Derashe, SNNPR There are almost 20,000 PSNP beneficiaries in Derashe and 20% of these receive direct support. The failure of rains in SNNPR meant that Derashe and other woredas were facing a serious emergency humanitarian caseload. Derashe, along with other PSNP woredas in SNNPR, faces relatively low PSNP coverage compared to woredas in other regions. It is estimated at woreda level that there are at least 8,000 people in households who meet the PSNP eligibility criteria but there are not resources to increase coverage. At community level, the number of households facing food shortages but not included in the PSNP was more than double the number of PSNP beneficiary households. As a result, some of the recommended changes in particular the shift to full family targeting are much more difficult to implement and to justify. It is difficult for woreda, kebele and village level officials to justify the concentration of resources to a smaller number of households in the community when so many households are in need of support and there are strong claims that a very large number of households are equally poor. There are a large number of appeals and significant tensions between both KFSTF and non-beneficiaries, and between beneficiaries and non-beneficiaries. In the face of pressure to dilute PSNP resources among a larger number of households, expectations are graduation are low compared to other regions. At woreda and kebele level, the regional benchmark of ETB16,000 is well-known but officials argue that the benchmark is not 50

57 realistic. It appears that households have not been able to improve their asset status to meet the benchmark. While the predictability of payments continues to be a problem, some progress has been made. Beneficiaries know what and how much they will receive but not when the transfers will be made. This knowledge is not enough to enable precise planning of expenditure but it is enough to provide collateral against loans from moneylenders Kalu, Amhara PSNP beneficiaries in Kalu receive four months of cash payments and two months of food transfer. In 2007 they received cash payments only but have opted to switch partially to food in the context of increased food prices. Just over a quarter of the woreda population of 205,000 people participate in the PSNP, just under 10% of these as direct support beneficiaries. The total number of beneficiaries has remained stable over the last two years. In 2005, the first year of the PSNP, the targeting of richer households for the PSNP in order to achieve graduation led to the exclusion of the poorest households and division within communities. Since 2006, the process and accountability of targeting has improved. These improvements have reduced the load on task force and appeal committee members. PSNP beneficiaries are also supported through other food security programmes. Targeting improvements are one outcome of the sharing of information, including the PIM, with a wider range of implementers. Other major achievements since 2006 include the capping of individual public work requirements, consideration of pregnant / lactating women s capacity to work. Payment errors have been ironed out through the issuing of a permanent identification card. In part, these improvements are an outcome of the development of uniform PSNP reporting systems for kebele development agents (DAs). DAs report the type and scope of public works activities, numbers and types of beneficiaries. The uniform reporting system has enabled the timeliness of payments to improve Lasta, Amhara In the 2006 targeting, linkages and trends in transfers studies, fieldwork was carried out in Bugna woreda in Amhara. Since 2006, there have been cadastral revisions and the area in which the targeting study worked has now become part of Lasta woreda. The PSNP in Lasta woreda provides a combination of cash and food transfers to 43,254 beneficiaries almost two-fifths of the population. Almost 16% of beneficiaries are under direct support. In 2006 the studies identified significant problems with targeting in the first two years of the PSNP programme in Amhara. In Bugna, that is middle-level households with higher wealth status, were targeted in order to maximise graduation. It was assumed that the poorest households should be targeted for resettlement and the richest would be support through agricultural extension programmes. In 2006, retargeting to explicitly identify the poorest households led to wholesale changes in the list. This led to a significant reduction in appeals which are now mainly focused on changing household circumstances resulting from marriage and divorce. 51

58 Other key changes in the woreda are caps on the maximum number of days of public works (20 days for a single person, 25 days per 2-person household); full family targeting; increased transparency in the targeting process; and improved timing of payments; and the shift of pregnant and lactating women onto the direct support component of the programme. Reliability of payments had improved in 2008 with positive impacts on asset depletion rates, food stocks, the availability of credit at reasonable rates, participation in community activities, distress migration rates, and school enrolments. There remain concerns about the appropriateness of some of the public works arrangements in Lasta where many public works participants have to walk for up to five hours to reach the location of public works tasks. While beneficiaries reflect positively on the assets that have been created and improved awareness regarding food security, the woreda has been unable to find any PSNP beneficiaries who can meet the regional bench mark for graduation. 52

59 CHAPTER 5 IMPACTS OF PSNP TRANSFERS This chapter reports on impacts of the Productive Safety Net Programme at the level of households, communities and markets. 5.1 Consumption uses of PSNP transfers Table 18 summarises the ways in which recipients of food transfers from the PSNP used that food. Direct Support beneficiaries, who are poorer on average than Public Works participants, were more likely to consume all this food (86% versus 71%). So for three in four food recipients (74%), PSNP food transfers were allocated entirely to meeting the immediate subsistence needs of beneficiary households, and contributed directly to enhancing household food security. A significant minority of Public Works participants ate some food but sold the rest (15%), sometimes to buy other types of food and sometimes to meet non-food needs. Altogether, one in five Public Works participants (21%) and one in thirteen Direct Support beneficiaries (7.8%) sold some or all of their food transfers. These percentages are higher than in 2006, reflecting the swing towards food transfers and away from cash more beneficiaries received food only in 2008, so meeting their non-food needs required converting some food aid into cash. A smaller proportion of PSNP beneficiaries (7%) gave away some of their food to other people. Table 18 Use of PSNP food transfers (last 12 months) Use of food Public Works Direct Support All current PSNP We ate all the food 332 (70.6%) 99 (86.1%) 431 (73.7%) Sold some food, ate the rest 68 (14.5%) 6 (5.2%) 74 (12.6%) Gave away some food, ate the rest 20 (4.3%) 6 (5.2%) 26 (4.4%) Sold the food to buy other food 21 (4.5%) 0 (0.0%) 21 (3.6%) Gave some food as payment, ate the rest 13 (2.8%) 0 (0.0%) 13 (2.2%) Sold all the food for cash 10 (2.1%) 3 (2.6%) 13 (2.2%) Gave all the food to livestock for feed 1 (0.2%) 0 (0.0%) 1 (0.2%) Gave all the food to others as a payment 0 (0.0%) 1 (0.9%) 1 (0.2%) Other 5 (1.1%) 0 (0.0%) 5 (0.9%) Total 538 (78.3%) 149 (21.7%) 687 (100%) Table 19 summarises the consumption uses of PSNP cash transfers over the 12 months preceding the survey. The overwhelming majority of cash recipients spent some or all of this cash on buying staple food (194/230 =84% of households), and this was also the item that took the largest proportion of the cash by any spending category (248 birr per household). The second most common spending category was groceries (61% of households), although this consumed a smaller amount of cash (49 birr). Third in the list is clothing (43% of households, 124 birr). Smaller numbers of households used some of their cash transfers for various social obligations and paying taxes. Relatively few cash recipients spent some of this money on non-staple foods such as meat (11%). 53

60 An insignificant number gave or lent some of this money to others much fewer than the number of food recipients who donated some food to help others. This is an interesting finding which has been confirmed elsewhere: people who receive social transfers are generally more inclined to share if the assistance is provided in the form of food rather than as cash. This could be interpreted as suggesting that food transfers reinforce traditional food sharing practices in rural communities, whereas cash transfers reinforce trends associated with modernisation such as commercialisation and individuation. However, a recent study of cash and food transfers in Swaziland found that this conclusion is misleading. Many cash transfer recipients bought food with the cash and then shared this purchased food, so they were no less inclined to share than were food recipients on the same programme (Devereux and Jere, 2008). Table 19 Consumption uses of PSNP cash transfers (last 12 months) Consumption items Public Works Direct Support All Current PSNP HHs Birr HHs Birr HHs Birr Bought staple food (e.g. grain) Bought groceries (e.g. salt) Bought clothes or cloth Social obligations Paid taxes Bought other food (e.g. meat) Gave some cash to help others Lent some money to others 1 n/a 0 n/a 1 n/a Total households Investment uses of PSNP transfers Table 20 summarises investment uses of PSNP cash transfers over the 12 months preceding the survey. The most popular item here is investment in education of children (28% of cash recipients), which reinforces a pattern across Africa cash transfers subsidise education costs in every country where they have been introduced. Even if education is fee free, there are many associated expenses (school books and stationery, uniforms, transport, etc) and cash transfers provide much needed cash to meet some of these expenses, and to retain children in school rather than withdraw them due to poverty or following a livelihood shock. Health bills were also settled using PSNP cash in several households (12%), so a significant impact on human capital can be inferred in terms of household investments in both education and health. In terms of investment in productive activities, many cash recipients bought seeds or fertiliser for farming (20% of households), while a small minority invested in their business (2%). The cash transfers did facilitate asset accumulation by many, especially livestock purchases (10% of cash recipients), which was also the most expensive investment item (averaging 226 birr). PSNP cash transfers also allowed some households to pay off debts (16% of households), some of which were quite large (averaging 161 birr), so there was a significant positive impact in terms of reducing the indebtedness of poor families. 54

61 Table 20 Investment uses of PSNP cash transfers (last 12 months) Investment items Public Works Direct Support All Current PSNP HHs Birr HHs Birr HHs Birr Paid for education costs Debt repayment Bought seeds for farming Paid for health costs Bought livestock Bought fertiliser for farming Used for business (e.g. trading) Other Total households The qualitative data suggest that the PSNP is making only a very limited contribution to asset building. In Lasta, beneficiaries argued that the amount of the PSNP transfer per month was very low and could only cover their food gap with the most prudent use. In practice, therefore, the transfer was not sufficient to enable households create or maintain assets at the benchmark required for graduation. In Chiro, beneficiaries reported that they were not selling off their livestock during the hungry months because the PSNP enabled them to eat but the research team found no examples of the PSNP enabling investments in productive asset that subsequently led to graduation. Similarly, in Derashe, respondents claimed that the PSNP had successfully protected beneficiary households from distress sales of assets, but there was no evidence that the PSNP had enabled investments. In both Derashe and Boricha this was attributed to the unpredictability of transfer delivery. Beneficiaries also argued that the arrangements for other food security programmes made it difficult to make investments profitable. For example, in Lasta, they argued that the period before which credit repayment was required was too short. Those taking packages were forced to begin repayment four months after they received credit. In most cases this was before the investment had become productive. For example, one public works participant in Lasta took credit to purchase goats but his repayments began before the goats had produced offspring. As a result, he began making repayments by selling goats that he had purchased for rearing. When he finished making the repayments he had sold all of the goats and made no profits. In Fedis, there are indications that limited savings and investments occurred indirectly as a result of the PSNP: Beneficiaries received enough to eat from the PSNP and, as a result, did not sell their assets. At the same time, the consumption support of the PSNP enables households to draw on package support from the government, the EU or the World Bank. However, it was argued that the switch to cash only transfers in Fedis and the simultaneous increase in grain prices had undermined the purchasing power of cash transfers so savings and investments were undermined. As a result, it was not possible for public works participants to invest in assets due to high cost of grain in

62 5.3 Household incomes Between 2006 and 2008 the nominal incomes of Public Works and Direct Support households more than doubled (by +124% and +115% respectively). Even after adjusting for inflation, real incomes of PSNP households increased by more than 50% in just two years (+59% and +52%) (see Table 21 and Figure 6). Of course it is not clear how much of this improvement in living standards should be attributed to the PSNP, but one indication is a comparison with a control group of households that did not receive support from the PSNP. Over the same period the nominal incomes of non-psnp households surveyed increased only marginally (+13%), and in real terms actually declined (-20%). Non-PSNP households were significantly wealthier than PSNP households in 2006 (+23%), but significantly poorer in 2008 (-38% in both nominal and real terms). These findings suggest two conclusions. First, that the increased incomes of PSNP households must be largely attributable to the PSNP, and second, that the poverty of households excluded from the PSNP has deepened their real incomes in 2008 are lower than the incomes of Public Works participants were in Even the gap with Direct Support beneficiaries has narrowed: in 2006 non-psnp households had incomes three times higher than Direct Support households; but by 2008 the difference was only 48%. Another indicator comes from comparing households that have taken OFSP packages with those that have not. Across all households in our sample, the average income of households that have taken a package is 15% higher, at 2,267 birr compared to 1,972 birr (Table 21). Interestingly, there is no significant difference among current PSNP beneficiaries, presumably because the income-generating effects of the packages have not yet been realised, but among past PSNP households the income of households that took packages (n=10) is double the income of those that did not (n=92), suggesting that the packages have made a substantial positive difference to their livelihoods. The same is true for non-psnp households that have taken packages though once again the number of households involved is small (n=11). Table 21 Annual household incomes, by PSNP status, 2006 and 2008 (birr) Economic characteristics Public Works Current PSNP Direct Support Past PSNP household Non-PSNP households All households , n/a 1,949 1, ,559 1,482 1,954 2,200 2, (deflated) 2,525 1,052 1,386 1,561 2, (deflated) No OFSP package 2,222 (n=554) 1,284 1,454 1,972 OFSP package 2,210 (n=147) 2,326 2,983 2,267 Notes: All numbers are average (mean) values for the category Figure 5 graphs the incomes of current, past, and non-beneficiaries in 2006 and Figure 6 shows the changes in real household income for each category of households. 56

63 Figure 5 Household income, by PSNP status, 2006 to 2008 Note: 2008 (deflated) is 2008 income at 2006 prices Figure 6 Changes in real household income, by PSNP status, 2006 to 2008 Note: Real income is 2008 income at 2006 prices Table 22 and Figure 7 disaggregate incomes by region. Households in Tigray have substantially higher incomes than any other region, while households in SNNPR have the lowest. Per capita incomes in Tigray have more than doubled in just two years since Households in SNNPR have also experienced significant increases, but remain considerably poorer than the other three regions. Nominal incomes in our Amhara sample have fallen, while incomes in Oromia have stagnated. (Note that these results refer only to our small, non-representative samples from two woredas in each region, and should not be extrapolated to infer income trends at the regional level.) 57

64 Table 22 Changes in mean per capita income, , by region (birr/year) Households Amhara Oromia SNNPR Tigray Total Male-headed % change -34% +6% +82% +145% +35% Female-headed % change +4% -34% +40% +211% +33% All households % change -19% -3% +31% +160% +34% Across the sample, per capita incomes in female-headed households are comparable to incomes in male-headed households (this was also the case in our 2006 survey). However, in Oromia female-headed households have lower incomes, while in Amhara the situation is reversed, and male-headed households are poorer. Figure 7 Mean per capita real income, by region and household type (birr/year) Disaggregating PSNP impacts by implementing agency also reveals some interesting findings. (Recall that the PSNP is implemented by the government in 4 of 8 woredas, by NGOs in 3 woredas, and by the government plus an NGO in 1 woreda.) The descriptive statistics indicate that respondents in NGO woredas have higher average incomes, and that NGO woredas have 58

65 seen a bigger proportional change in incomes over the period 2006 to 2008 than non-ngo woredas. In both government and NGO woredas, real incomes have risen by more than 50% for beneficiaries but have fallen for non-beneficiaries (Table 23). Note that the strong income result for beneficiaries may derive from the inclusion of PSNP Public Works payments in the aggregate income variable. In terms of asset ownership, respondents in government woredas have seen their asset values more than double over this period, while NGO woredas have also seen significant increases, though not on the same scale. Table 23 Changes in incomes and assets, by PSNP implementing agency Implementing agency Income 2006 Income 2008 (adjusted) Change (%) Assets 2006 Assets 2008 (adjusted) Change (%) Government % % Beneficiaries % % Non-beneficiaries % % NGO % % Beneficiaries % % Non-beneficiaries % % 5.4 Household assets All categories of household in our survey enjoyed dramatic increases in their asset holdings from 2006 to 2008, as measured by the value of assets at current market prices (Table 24). Even after deflating current asset holdings by 2006 prices (collected during our previous PSNP survey), the real increase in asset values more than doubled for every category of household Public Works participants (274%), Direct Support beneficiaries (201%), and non-beneficiaries (247%), with an average real increase across all households of 245% (see also Figure 8). Table 24 Household asset values, by PSNP status, 2006 and 2008 (birr) Economic characteristics Public Works Direct Support Past PSNP household Non-PSNP households All households n/a 1, ,549 1,039 2,486 4,731 3, (2006 prices) 2, ,239 3,639 2, (deflated) No OFSP package 1,974 (n=554) 1,170 3,572 2,175 OFSP package 1,883 (n=147) 1, ,056 59

66 Figure 8 Change in real asset values, by PSNP status, 2006 to 2008 Note: Asset values are calculated at constant 2006 prices Table 25 reveals the reason why asset values have increased so dramatically between 2006 and The main factor is not a large increase in the mean number of assets held per household, but a very significant increase in the number of households owning each type of asset. This effect is particularly pronounced among current PSNP beneficiaries, where the following increases in households owning assets are among the most significant: cows up from 291 to 328 households; calves up from 152 to 240 households; poultry up from 250 to 408 households; stoves up from 39 to 183 households; radios up from 50 to 141 households; gold or silver jewellery up from 6 to 155 households. The same cannot be said of non-beneficiaries, many of whom have lost assets since 2006: oxen are down from 109 to 78 households; cows are down from 114 to 99 households; pick-axes are down from 117 to 74 households. It is true that the past two years have been good harvest years, and this is likely to explain much of the increases in asset holdings across the board. In particular, it appears that many PSNP beneficiaries who held no assets before have been able to acquire assets as a result of good weather conditions and receiving PSNP transfers. On the other hand, beneficiaries that held assets before have not managed to accumulate assets at a faster rate than non-beneficiaries. This is consistent with our findings on household incomes. It confirms that the PSNP is having real positive impacts on beneficiaries, especially on the poorest beneficiary households (see also Figure 22 below). Note that these results on asset ownership include assets transferred through OFSP livelihood packages (see for instance the increase in households owning modern beehives up from 7 to 38). Either way, Table 25 confirms that the PSNP (plus the PFSP) has had a major impact on asset accumulation among beneficiaries. 60

67 Table 25 Assets owned, by PSNP status, 2006 and 2008 Asset Current Non PSNP Current Past PSNP Non PSNP Livestock Oxen (238) 1.21 (109) 1.55 (265) 1.40 (30) 1.60 (78) 1.62 Bulls (40) 1.15 (23) 1.39 (50) 1.12 (6) 1.16 (25) 1.36 Cows (291) 1.12 (114)1.25 (328) 1.19 (48) 1.18 (99) 1.37 Heifers (87) 1.05 (26) 1.23 (119) 1.07 (13) 1.07 (36) 1.41 Calves (152) 1.17 (63) 1.19 (240) 1.16 (36) 1.16 (77) 1.28 Sheep (122) 2.87 (48) 3.48 (160) 3.76 (28) 2.14 (41) 3.34 Goats (216) 3.08 (66) 3.89 (223) 3.81 (46) 3.10 (73) 4.17 Donkeys (129) 1.21 (55) 1.23 (215) 1.17 (33) 1.03 (55) 1.14 Camels (5) 2.20 (10) 2.40 (10) (5) 3.20 Poultry (250) 3.66 (75) 3.66 (408) 4.43 (52) 5.93 (100) 4.67 Productive assets Plough (342) 1.09 (121) 1.20 (379) 1.17 (53) 1.20 (109) 1.14 Sickle (machid) (458) 1.73 (146) 1.98 (495) 2.00 (79) 1.45 (128) 2.10 Pick-axe (doma) (383) 1.15 (117) 1.27 (233) 1.20 (20) 1.15 (74) 1.37 Axe (metrebia) (439) 1.28 (129) 1.50 (475) 1.36 (71) 1.19 (119) 1.68 Hoe (mekotkocha) (298) 1.45 (93) 1.64 (307) 1.60 (62) 1.96 (79) 2.05 Spade (akefa) (337) 1.06 (107) 1.07 (346) 1.10 (43) 1.04 (90) 1.11 Traditional beehive (50) 1.74 (15) 2.53 (36) 1.58 (2) 1.00 (10) 1.30 Modern beehive (7) (38) 1.89 (1) 2.00 (3) 1.03 Water pump (hand/foot) (6) 1.00 (2) 1.00 (9) (1) 1.00 Water pump (diesel (8) 1.00 (4) 1.00 (14) 1.14 (1) 1.00 (3) 1.00 Grain mill (stone) (186) 1.08 (53) 1.11 (319) 1.06 (55) 1.07 (82) 1.10 Grain mill (diesel) (1) 1.00 (2) 1.00 (3) 1.00 (1) 1.00 (1) 1.00 Household goods Charcoal/ wood stove (39) 1.02 (10) 1.00 (183) 1.09 (11) 1.00 (26) 1.26 Metal bed (14) 1.00 (18) 1.00 (51) 1.01 (8) 1.00 (19) 1.10 Consumer durables Radio (50) 1.00 (34) 1.05 (141) 1.01 (16) 1.00 (46) 1.06 Jewellery (gold, silver) (6) 2.00 (6) 1.00 (155) 4.99 (9) 3.70 (30) Wristwatch (34) 1.00 (17) 1.00 (161) 1.15 (27) 1.14 (57) 1.24 Finally, there was a small increase in households owning land across all household categories, but a more significant increase in land cultivated than land owned. Non-PSNP households cultivate more land than PSNP households, but the increase in hectares cultivated by non-psnp households between 2006 and 2008 was only 8% (from 0.98 to 1.06 ha), while the increase among Public Works households was 23% (from 0.64 to 0.79 ha). An intriguing finding is a 43% decline in land cultivated by Direct Support households (from 0.47 to 0.27 ha) (Table 26). This could signify that many households receiving cash or food transfers from the PSNP no longer need to farm for a living, which can be interpreted as either a negative outcome (a sign of rising dependency) or a positive outcome (since these households are labour-constrained). 61

68 Table 26 Land ownership by PSNP status, 2006 and 2008 Economic characteristics Households owning land (%) Public Works Direct Support Past PSNP household Non-PSNP households All households % 75.4% n/a 88.2% 86.5% % 80.2% 94.1% 92.9% 90.9% Land cultivated (hectares) n/a Notes: All numbers (not percentages) are average (mean) values for the category 957 households reported asset values; 3 households did not report. 5.5 Livelihood activities Interestingly, PSNP households are more likely to be involved in a diverse range of livelihood activities than non-psnp households. More PSNP households are engaged in livestock fattening and poultry-rearing, and more have salaried jobs. Other activities where PSNP households dominate include public works (which is obviously related to the PSNP), carrying water, and sharecropping, all of which are conventionally thought of as poor people s work. In most other livelihood activities there is little difference between PSNP and non-psnp households (see Table 27). Table 27 Household livelihood activities Sector Agriculture Employment Trading Livelihood Activity PSNP households Past PSNP households Non PSNP households Crop production 592 (84.4%) 143 (91.1%) 99 (97.1%) Rearing and selling animals 379 (54.3%) 107 (68.1%) 59 (57.8%) Livestock fattening 28 (47.3%) 8 (5.1%) 8 (7.8%) Poultry rearing for sale 332 (47.3%) 78 (49.6%) 42 (41.1%) Salaried job 68 (9.7%) 9 (5.7%) 4 (3.9%) Public works 575 (82.2%) 6 (3.8%) 48 (47.5%) Agricultural worker 21 (3.3%) 10 (6.3%) 5 (4.9%) Non-agricultural 23 (3.3%) 8 (5.1%) 3 (2.9%) Unpaid work for community 304 (43.4%) 87 (55.4%) 69 (67.6%) Food crops 35 (5.0%) 8 (5.1%) 9 (8.8%) Selling firewood 86 (12.3%) 26 (16.5%) 13 (12.7%) Water carrier 301 (43.0%) 49 (31.2%) 14 (13.8%) Sharecropping 93 (13.3%) 8 (5.1%) 1 (0.9%) It is interesting to consider levels and trends in wage rates for the common livelihood activities, bearing in mind that the PSNP wage rate has been raised since 2006 from 6 birr to 8 birr per day. The range of agricultural wages paid in cash during 2007 was 7 to 30 birr, with an average of 17.1 birr per day (more than double the PSNP payment). In Amhara, for instance, people 62

69 were employed as daily labourers for weeding and harvesting, at a standard rate of 10 birr per day (not much higher than the PSNP wage). Figure 9 presents the daily wage reported across several study sites, month by month, which reveals a narrower range (because these figures are averaged within each woreda surveyed) of 14 to 20 birr (see also annex Table 48). The fitted trend-line reveals that wages were rising steadily throughout the year, with a sharp dip in July-August, ending the year some 25% higher than at the beginning of the year. Non-agricultural wage rates are generally higher than agricultural wages, and during 2007 this gap increased. In Amhara, where agricultural workers received 10 birr/day in 2007, farmers who found work as daily labourers in nearby towns were paid 12 birr/day. In Oromia, the average pay for unskilled casual work was reportedly 15 birr. In Tigray Region, semi-skilled casual work such as construction activities and carpentry paid 20 birr in rural areas, but birr in towns. Overall, non-agricultural daily wages in our survey communities increased from around 14 birr in January to 25 birr in December (Figure 9), with an average of 21.3 birr. The increase during the year was 47%, almost double the increment in agricultural wages. Non-agricultural work is more difficult to find and often requires specific skills, which explains the higher remuneration. It is not clear whether rising wages were a reaction to rising food and commodity prices, but it is possible that the faster rise in non-agricultural wages reflects rapidly escalating living costs, especially in urban areas. Figure 9 Agricultural and non-agricultural wage rates, 2007 (birr/day) 5.6 Household food security Just over half of current PSNP participants surveyed (55.4%) suffered household-level food shortages during 2007/08. This amounts to a 35% decline from the figure of 85.6% in 2005/06 (Table 28). While this clearly represents a significant improvement in household food security status, it cannot be attributed to the PSNP, because a comparable decline in food insecurity of 33% was recorded in non-psnp households. Evidently 2007/08 was a better year in food security terms than 2005/06 (despite the high food prices of 2008). 63

70 Table 28 Households suffering food shortage in last year Household Category 2005/ /08 % change Region Amhara 205 (85.4%) 63 (26.5%) -69% Oromiya 224 (93.3%) 175 (73.8%) -22% SNNPR 226 (94.2%) 175 (74.5%) -23% Tigray 167 (69.6%) 113 (47.3%) -32% PSNP status Cash only 73 (60.8%) 61 (42.7%) -16% Food only 124 (86.1%) 93 (50.5%) -25% Cash + food 484 (97.2%) 237 (65.1%) -51% Current PSNP households 681 (89.7%) 394 (56.6%) -42% Past PSNP households n/a 59 (59.0%) n/a Non-PSNP households 141 (71.2%) 73 (47.7%) -48% Total 822 (85.6%) 526 (55.4%) -36% Figure 10 illustrates graphically the (self-reported) finding that current PSNP beneficiaries were more food insecure during 2007 than either past beneficiaries or non-beneficiaries in our survey. During the peak hunger months of June-August 2007, 40% of PSNP households experienced food shortages despite participating in the PSNP (this could be an indicator of late payments). In the early months of 2008, past beneficiaries (who suffered the lowest rates of food shortage during 2007) were struggling most, which could be an indicator that they were graduated off the PSNP too soon having no PSNP transfers to protect them, their food insecurity is rising faster than for other groups. Figure 10 Food shortages by PSNP status and month, 2007/08 A comparison with our 2006 survey results reveals overall levels of food shortage have fallen (Table 29), but this is misleading as the 2007/08 data only cover the first 3-4 months of 2008, before the hungry season months in mid-year when these rates will certainly rise. 64

71 Table 29 Households experiencing food shortages by PSNP status, 2006 and 2007/ PSNP status Food shortage No food shortage Total households PSNP households 681 (89.4%) 81 (10.6%) 762 (100%) Non-PSNP households 141 (71.2%) 57 (28.8%) 198 (100%) Total households 822 (85.6%) 138 (14.4%) 960 (100%) 2008 Current beneficiaries 394 (56.6%) 302 (43.4%) 696 (100%) Non-beneficiaries 73 (47.7%) 80 (52.3%) 153 (100%) Past-beneficiaries 59 (59.0%) 41 (41.0%) 100 (100%) Total households 526 (55.5%) 423 (44.5%) 949 (100%) Figure 11 disaggregates the prevalence of self-reported food insecurity by region for the period January 2007 to March During 2007 households in Oromiya experienced the highest levels of food shortage, followed by SNNPR and Tigray, with very few households in our Amhara sample (<20% even in the peak hunger months) reporting food shortages in These findings could indicate selection bias in targeting (some regions being more pro-poor than others), or regional differences in rainfall and harvest outcomes, or differences in the effectiveness of PSNP delivery (e.g. Oromiya has had serious problems of delayed payments). All regions are belg-producing except Tigray, where the fall in food insecurity occurs later than in other regions only in October indicating the advantage of harvesting a belg crop. Figure 11 Households experiencing food shortages by region, 2007/08 Figure 12 compares self-reported food shortages in 2005/06 with 2007/08. There is a very significant reduction in overall levels of food shortage across the years, both for beneficiaries (peaking at 60% in 2005 but 40% in 2007) and for non-beneficiaries (peaking at 50% in 2005 and at 35% in 2007). Because these imporvements are of similar magnitude for both groups, this result cannot be attributed to the PSNP but probably reflects exogenous factors such as the distribution of years of good and bad rainfall. 65

72 Figure 12 Households experiencing food shortages in 2005/06 and 2007/08 Table 30 confirms the close association between household food security status (using selfreported food shortage as a proxy for food security), and household incomes, and asset values. Food insecurity is closely correlated with income and asset poverty. Interestingly, the gap in assets between food secure and food insecure households is much wider than the gap in incomes (171% versus 34%). This suggests that asset ownership is a robust indicator of household vulnerability. Food insecure households also have lower asset values than average incomes, while food secure households have higher asset values than incomes. This could be an indication that food insecure households are forced into selling assets to raise cash to buy food. This is a common hunger coping strategy, as discussed in the following section. Table 30 Household food security, incomes and assets, 2008 PSNP status Income Assets Households Current Beneficiary Food insecure 1,964 1, Food secure 2,567 2, Non-Beneficiary Food insecure 1,378 1, Food secure 1,769 5, Past-Beneficiary Food insecure Food secure 2,091 2, All households Food insecure 1,765 1, Food secure 2,369 3, All households 2,023 2,

73 5.7 Coping strategies Table 31 and Figure 13 rank 15 coping strategies that households resorted to during 2007/08 to survive the hunger that they experienced. A fairly predictable pattern is evident, with the ranking familiar from the literature on famine coping strategies. Rationing of food consumption eating smaller portions or fewer meals per day is the most commonly adopted response by all categories of household. This is followed by cutting non-essential spending. All of these are easily reversible strategies with no long-run consequences, unlike the next most common strategies selling livestock or borrowing to buy food, or migrating to find work, all of which could compromise the future viability of household livelihoods. Table 31 Coping strategies adopted in 2007/08, by PSNP status Coping strategy Current PSNP beneficiaries Past PSNP beneficiaries Nonbeneficiaries Ate less food (smaller portions) 78.1% 81.8% 72.3% Reduced number of meals per day 72.3% 78.1% 65.2% Reduced spending on non-food items 31.8% 44.7% 31.2% Sold livestock to buy food 29.3% 29.0% 33.3% Borrowed food or cash to buy food 21.8% 26.7% 23.4% Household members migrated to find work 12.8% 24.4% 10.6% Sold firewood or charcoal to buy food 10.0% 17.4% 12.0% Relied on help from relatives/ friends 8.1% 20.9% 5.6% Rented out land to buy food 6.9% 1.1% 0.0% Sent children to work 2.7% 9.3% 6.3% Sold other assets to buy food 2.1% 8.1% 4.2% Sent children to stay with relatives 1.7% 4.6% 4.2% Withdrew children from school 0.9% 3.4% 2.1% Collected bush products to eat or sell 0.9% 3.4% 1.4% Sold land to buy food 0.7% 0.0% 0.7% For almost all coping strategies, the rate of adoption is highest for past beneficiaries, implying that they have been badly affected by being dropped from the PSNP, and have been forced to adopt various damaging behavioural adjustments to survive. Conversely, current beneficiaries have been partly protected. There is no systematic difference between current beneficiaries and non-beneficiaries in terms of coping strategies adopted. There is no clear pattern in terms of asset disposal behaviour, with one in five current, past and non-beneficiary households each selling livestock to buy food in 2008 (Table 32). 67

74 Figure 13 Coping strategies adopted in 2007/08, by PSNP status Table 32 Asset disposal behaviour in , by PSNP status (% of households) Asset disposal Sold livestock for food needs Sold livestock for non-food needs Sold or rented out land for food Sold or rented out land for non-food Non beneficiaries Current beneficiary Past beneficiary Household self-assessment Table 33 presents household self-assessments of well-being status for 2006, 2007 and 2008 (see also Figure 14, Figure 15 and Figure 16). Among current beneficiaries there has been a slight deterioration, with only 37% reporting they are doing well or okay now compared to 43% two years ago, and 63% struggling or unable to meet needs now compared to 57% two years ago. Non-beneficiaries have seen a similar deterioration, though from a better initial situation (not surprisingly, since they were not targeted as needing PSNP assistance): those doing well or okay fell from 62% to 52%, while those doing badly increased from 37% to 48%. The group that has fared worst of all, however, is past beneficiaries. Those doing well or okay have fallen from 49% to 34%, while the most alarming increase is in the number who are unable to meet household needs, which was only 14% in 2006, when these households were included in the PSNP, but rose to 47% by 2008, following their exclusion. One factor that explains the general negative trend across all household categories is rising food prices. But between groups, past beneficiaries are worst off in 2008, suggesting that being dropped from the PSNP prematurely is extremely detrimental for household well-being. 68

75 Table 33 Subjective self-assessment, , by PSNP status (% of households) Self-assessment Non beneficiaries Current beneficiary Past beneficiary Doing well Doing just okay Struggling Unable to meet household needs Total households Figure 14 Self-assessment, current beneficiaries, 2006 to 2008 Figure 15 Self-assessment, past beneficiaries, 2006 to

76 Figure 16 Self-assessment, non-beneficiaries, 2006 to Graduation Although graduation is not intended to be achieved by the PSNP alone, it is an explicit objective of the Food Security Programme (FSP), which has the Productive Safety Net Programme and Livelihood Packages as two central components. One indicator of graduation is the number of households in our panel that have left the PSNP since Of 803 beneficiary households in 2006, 102 were no longer beneficiaries in 2008, implying that 12.7% have graduated out of the need for PSNP assistance. On the other hand, the findings on self-assessed well-being (above) revealed that two-thirds of these households (66%) are struggling or unable to meet household needs in 2008, raising questions about whether these households were removed from the PSNP before meeting the criteria for graduation. It is possible that some were rotated off the PSNP to make way for other beneficiaries, and many were removed in the retargeting exercise that followed the discovery that many middle-income households had been registered initially in order to maximise the PSNP s graduation potential. Of 102 respondents that have left the programme ( past beneficiaries ), 98 said they had been included through regular targeting, and 3 said they had been included through the appeal process. Most past beneficiaries believe they were included in the PSNP because they were poor (84%) or food insecure (78%), while smaller numbers believed it was because they lacked productive assets either land (57%) or labour (8%). Among the 102 past beneficiaries, 63% said they had heard of the term graduation, as compared to 335/701 (48%) of current beneficiaries a significant difference. Explaining why they had left the PSNP, 12% claimed to have graduated from the programme, 51% reported (unspecified) other reasons according to feedback from fieldworkers, this was usually related to the 2006 retargeting while 35% did not respond to this question and 2% did not know. Only 12 past beneficiaries claimed to have received livelihood packages. This is significant because graduation is expected to be achieved mainly through the packages, not through the income transfers disbursed on the PSNP. In this context, the fact that the 12 past beneficiaries who took packages enjoyed substantially higher incomes (81%) and asset values (60%) than the 90 who did not raises questions about why these 90 households were removed from the PSNP beneficiary list, especially since they reported the lowest incomes and asset values of any other category of beneficiary or non-beneficiary (Table 34). 70

77 Table 34 Livelihood packages by beneficiary status, 2008 Beneficiary status Households Income Assets Beneficiary No package 554 2,222 1,974 Livelihood Package 147 2,210 1,883 Non-Beneficiary No package 146 1,454 3,572 Livelihood Package 11 2,983 4,535 Past-Beneficiary No Package 90 1,284 1,170 Livelihood Package 12 2,326 1, Gender impacts of PSNP transfers Women tended to focus on the consumption and asset protection aspects of the programme while men focused on creating assets and investments in productive activities. For example, in Chiro, women argued that: Because of the safety net programme, you don t sell the goat that is in front of you. Instead you are eating with help from the government. In Fedis, there were many concerns about whether the PSNP created an additional burden for women. As well as the issue of pregnant and lactating women continuing to do public works to support their husbands, women argued that public works was difficult to combine with their other childcare responsibilities and domestic activities. When women were doing public works, their children remained at home, unattended and unfed. In Lasta, while the changes to rules governing the participation of pregnant and lactating women in public works were seen as highly positive, there remained concerns about the lack of childcare facilities for women doing public works. However, it was argued that the PSNP had improved the recognition of the importance of women s domestic and productive work. Following participation in public works, there are instances where women have developed confidence to share ideas and discuss household issues with their husbands and some men have begun to recognised women s rights to asset and property ownership. In Enderta, men were likely to spend cash on larger items in particularly, invest in productive assets while women s expenditure was focused on food and other basic household necessities. Similarly, in Derashe, officials argued that they explicitly targeted women and female-headed households precisely because they were more likely to use any cash distributed to meet basic household food needs rather than spend the money on other items. Control over cash transfers is often related to who collects the cash. In many programmes, men tend to take control of cash transfers while women are in charge of food aid, reflecting their traditional ascribed role as providers of meals to the family. On the PSNP, however, there are no obvious gender differences between the collection of cash and food transfers. Women are marginally more likely than men to collect food, but men are no more likely to collect cash than are women (Table 35). 71

78 Table 35 Responsibility for collecting PSNP cash and food payments Household members Cash Food Household head (male) 285 (40.2%) 263 (40.5%) Household head (female) 154 (21.7%) 178 (27.4%) Spouse of household head (female) 122 (17.2%) 106 (16.3%) Husband and wife together 11 (1.6%) 0 (0.0%) Other adult male from the household 28 (3.9%) 29 (4.5%) Other adult female from the household 15 (2.1%) 22 (3.4%) Young boys from the household 15 (2.1%) 31 (4.8%) Young girls from the household 14 (2.0%) 15 (2.3%) There is no cash / food payment 65 (9.2%) 5 (0.8%) Total respondents 709 (52.2%) 649 (47.8%) The household questionnaire asked PSNP participants what they would be doing if they were not working on Public Works projects, in order to establish the opportunity cost of the PSNP labour requirement. These findings are significantly gendered. Almost half (46%) of female respondents (but only 3% of males) stated that they would be doing domestic chores (Table 36). More than half of males (55%) but only one-third of females (35%) said they would be attending school, suggesting that large numbers of young people are engaged in PSNP Public Works, possibly at the expense of their education a worrying finding. The next most common activity forgone because of PSNP work commitments was farming (tending fields or livestock), which was mentioned by 27% of male respondents but only 6% of females. Very few respondents reported losing earned income because they were employed on Public Works just 3 male and 3 female respondents claimed that they would have been engaged in paid employment, either locally or by migrating, while 16 females and 16 males stated they would have engaged in trading or other micro-enterprises. Presumably it was the certainty of regular employment on the PSNP that attracted these people, whose earnings from trading or migrating in search of work would be much more unpredictable. Table 36 What PSNP participants would be doing if they were not doing Public Works Activity Female Male Total Attend school 214 (35.4%) 268 (54.6%) 482 (44.0%) Domestic chores 280 (46.3%) 16 (3.3%) 296 (27.0%) Livestock tending 27 (4.5%) 68 (13.9%) 95 (8.7%) Farm work 7 (1.2%) 66 (13.4%) 73 (6.7%) Child-care (own) 6 (1.0%) 2 (0.4%) 8 (0.7%) Paid employment (local) 2 (0.3%) 1 (0.2%) 3 (0.3%) Paid employment (migration) 1 (0.2%) 2 (0.4%) 3 (0.3%) Social or religious organisation 0 (0.0%) 2 (0.4%) 2 (0.2%) Trading/ going to market 1 (0.2%) 0 (0.0% 1 (0.1%) Teach community members 1 (0.2%) 0 (0.0%) 1 (0.1%) Other enterprise 7 (1.2%) 9 (1.8%) 16 (1.5%) Something else 9 (1.5%) 7 (1.4%) 16 (1.5%) Nothing 50 (8.3%) 50 (10.2) 100 (9.1%) Total 605 (55.2%) 491 (44.8%) 1,096 (100%) 72

79 5.11 Social impacts Gender issues: In the qualitative fieldwork, the focus of the gender research questions was very much on the gender implications of PSNP design and implementation for example, whether policy recommendations to shift pregnant women and nursing mothers from Public Works to Direct Support had been implemented. As a result there were only limited examples of intra-household gender conflicts. Two common issues emerged. The first was when men migrated to towns or other rural areas to seek work, and women were left in situ but removed from the PSNP because kebele task force officials assumed that migrating husbands had found work and would be sending remittances. Frequently, this was not the case. The second common conflict occurred when husbands doing Public Works forced their wives, even when pregnant or nursing, to accompany or replace them in works teams. Generational issues: Generational conflicts resulting from the PSNP tend to be between adults and young people of school-attending age. Conflicts between parents and students arose when migrating to school would mean that household members, or even households were taken out of the PSNP. Kebele officials in Enderta argued that households that could send their children to attend school in Mekele, were not among the poorest households. This created a disincentive to send children to school and caused tension within households. There are few conflicts between working age adults and the elderly, though this may be because Direct Support coverage remains low in most areas so the PSNP does not provide a quasi-non-contributory pension for the elderly. Community-level: The 2006 targeting and linkages studies identified several sources of social conflict generated by the PSNP. In some cases, the targeting and retargeting exercise generated antagonism toward FSTF members and kebele and woreda administrations. The 2008 review found continued instances of conflict with PSNP administrators over targeting (e.g. in Chiro, one kebele task force member claimed that his grain store was burnt down in response to exclusion from the PSNP). Overall, however, there appears to be less antagonism toward PSNP administrators in 2008 than in This may be a result of improved awareness about the PSNP and target processes and priorities (see section 3.1 above). In 2006, exclusion from the PSNP was also identified as generating social tension between beneficiaries and non-beneficiaries. In some cases, intra-community tension was sparked by PSNP households leapfrogging ahead of households who had been ranked as wealthier in community targeting exercises. This was most common in Oromia and SNNPR, particularly where coverage was lowest. There, many households had large food gaps and there were only small material differences between households selected for the PSNP and those that were richer. In Chiro, for example, this led to confusion about retargeting and whether households that had leapfrogged their neighbours should leave the PSNP the following year. Households that had been overtaken by PSNP beneficiary households revealed high levels of resentment. In Chiro, beneficiary and non-beneficiary households began worshipping at different mosques in the village as a result of the targeting process. In 2008, the issue of leapfrogging was not explicitly explored in the qualitative fieldwork. Compared to 2006, however, no evidence emerged of leapfrogging causing community tension. In SNNPR, where coverage was lowest and leapfrogging was most likely, tension between beneficiaries resulting from leapfrogging was reduced because PSNP transfers were diluted among a greater number of households through partial family targeting. Tensions 73

80 between beneficiaries and non-beneficiaries remain, however. Non-beneficiaries in Derashe derided beneficiaries as being children of the government. In Lasta, non-beneficiaries said we are smelling the food of beneficiaries, and in Enderta beneficiaries were alienated with statements like: you eat alone, you will die alone. In 2006, the appeals process was a source of social tension in some woreda. The 2008 review finds that there are fewer appeals in the study woredas and in some cases more robust appeals processes are in place. In some contexts (e.g. Tigray) appealing one s exclusion from the PSNP remains a challenge. As in 2006, there is a great deal of social pressure not to appeal. In both Enderta and Kilte awlelo, appellants must provide administrators with the names of existing beneficiaries who are better off than they are, if their appeal is to be considered. Given Tigray s tight-knit communities, this is a significant challenge, as the following quote from a potential appellant in Kilte awelo illustrates: I was requested to mention by name a better-off household from the beneficiaries during community meeting. Unfortunately, the one I know is my friend and has one family, but I have 4 family members. Considering the friendship and neighbourhood, I decided not to mention the name Impacts on HIV and AIDS One issue that emerged since the 2006 studies is the links between the PSNP and HIV and AIDS. Several possible linkages can be postulated. One risk is that the PSNP could accelerate HIV transmission, for instance if PSNP participants spend nights away from home for public works duties, or to collect food or cash transfers. The most recent Demographic and Health Survey (CSA, 2006) found that overnight absences from home is a predictor of HIV-prevalence in men. On the other hand, the PSNP offers an opportunity to deliver access to HIV and AIDSrelated information and services to large numbers of poor rural Ethiopians, many of whom are excluded from access to basic services by their poverty and remoteness. For these reasons, this study was requested to explore whether the PSNP is contributing to the spread of HIV, with a view to developing effective responses and mainstreaming this important issue in future rounds of PSNP design and delivery. As expected, it proved difficult for researchers to get at this issue without having a separate research module. Researchers did explore HIV-infection risks that could result from participation in the programme and asked whether proxy indicators were recognised in either the targeting of beneficiaries or the allocation of Direct Support or Public Works. In terms of targeting there are only limited examples of where the chronically ill were explicitly included in the PSNP Direct Support component. Primarily, Direct Support beneficiaries remain comprised of elderly and disabled people, plus pregnant and lactating women who have been temporarily transferred across from Public Works. In Tigray, it was suggested at various levels of government that the PSNP implicitly supported orphans from HIV/AIDS affected households. This was not borne out at woreda level though. For example, in Kilte awlelo the quota for direct support is based on relief programmes of previous years and a series of other factors none of which are based on good HIV/AIDS proxy indicators. For example, there is no evidence that changing household labour capacity, adult mortality and morbidity or chronic illness due to HIV/AIDS are part of the rationale for the quota. Similarly, in Kalu, officials argued that there is increasing orientation of the PSNP towards households affected by HIV/AIDS but the claims were not really substantiated. 74

81 In Chiro, women said that the targeting criteria and public works arrangements were not helpful to people living with AIDS. For example, the cap on the distance to public worksites of 5 km was not upheld at kebele level and chronically ill people were travelling further than was acceptable. In Chiro, there was no evidence that chronic illness was used as a criterion for Direct Support. SNNPR have been more progressive in addressing HIV/AIDS in their programme design and targeting. In Derashe, two households were shifted from public works to direct support when chronic illness due to HIV/AIDS began to have a serious and detrimental effect on their labour capacity. In Boricha, chronic illness due to HIV/AIDS is recognised as one of the criteria for direct support allocation. The contingency fund is being used explicitly to support HIV/AIDS patients and an orphans prioritisation scheme. In the woreda, three households were switched from public works to direct support this year as part of this process. Many women were concerned about the distances that they travelled to public works and to transfer distribution points. In Fedis, the risks were smaller because collecting transfers did not received an overnight stay. In Chiro, the distances travelled were far but were managed by travelling in groups and staying in safe houses. They overcame the risks by travelling with their husbands or with other family members and friends. In Kilte Awlele, REST moved the distribution to the tabia to decrease the risk. In Enderta, both the collection of transfers and the location of public works are near to beneficiaries homes so these do not create an increased risk of HIV infection. Other concerns, such as the proximity of the village to Mekele, children undertaking schooling in town, and local cultural practices (such as circumsicion) are thought to be a much greater issue affecting HIV infection. The greatest concern was in Lasta where women had to travel to Lalibela to collect benefits and stay there overnight. In Lasta even the distance to public works was so great that it required an overnight stay Impacts on vulnerability and emergencies There are a number of pockets of emergency responses in most woredas most years that tackle small-scale or geographically-confined emergencies such as localised flooding or hailstorm damage. Beyond this, the main linkage through the PSNP and emergency caseloads was through the contingency fund in the PSNP. In Chiro, the contingency fund was used in 2008 to support an emerging caseload, whose land had been flooded. When the belg rains failed, and additional kebele was added to the PSNP. Compared to other woredas, stakeholders in Fedis placed greater emphasis on the prevention of distress sales of productive assets by beneficiaries in the face of shocks and stresses. The concern with resilience in the face of shocks and stresses was prominent in Fedis. In Tigray, 5% of the budget is allocated to a woreda level contingency fund 15% of the budget remains at regional level for contingency. The regional contingency fund has been used to enable retargeting (additional caseload) and to tackle emergencies that are officially outside the remit of the PSNP but lacking alternative support. The federal bureau clears the use of the contingency fund outside the PSNP remit. In 2007, the contingency fund was used to support an emergency case of about 152,000 people who had become food insecure due to hailstorms and subsequent crop damage. Contingency funds that are used for emergency caseloads are targeted using the PSNP PIM and targeting guidelines, other cases are implemented through DPPA. In Enderta, the woreda worked in a similar way, using 5% of PSNP funds (25% of the 75

82 total contingency allocation) for treated emergency caseloads that arose during periods of floods or drought. The region controlled and allocated the other 75% of the contingency fund). There was a strong view that, in Enderta, the PSNP provided a critical role in enabling households to avoid distress sales of assets during crisis and so reduced the need for an emergency response (insert quote from p. 19). Indeed, no emergency relief programme was deemed to be required at the time of fieldwork in either Enderta or Kilte awlelo 9. In Boricha, the contingency fund was already being used for small-scale emergencies (localised flooding or case-by-case allocations). While a serious and acute humanitarian crisis was already emerging in Boricha at the time of the research, anecdotal evidence there and elsewhere in SNNPR suggested that people requiring supplementary feedings and children in therapeutic feeding centres were mostly not PSNP beneficiaries. This suggests that the PSNP was proving effective in protecting beneficiaries households from serious food shortages. Many beneficiaries responded that they would have faced acute hunger if the PSNP had not been in place. In SNNPR, the first of the PSNP regions to experience seasonal droughts and subsequent food shortages, one desk official at woreda level argued that even maintaining household welfare was too much for some PSNP households and they would require emergency assistance but argument was heard in only one instance during the research. In Derashe, the 5% contingency fund was used for meeting emergency caseload resulting from lowland flooding. However, as a large number of people slipped into food shortages in May and June, there was no mechanism to support an increasingly large caseload. However, overall, the team could find only very limited evidence that there were working mechanisms in the PSNP system to help provide early-warning of or to trigger emergency responses to address a large scale humanitarian crisis. In Lasta, for example, woreda officials reported that, except for access to the 5% contingency fund, there were not mechanisms to trigger emergency responses. They argued that, in the absence of a functioning RRT, they were unable to track increased caseload, appeals and food gaps in a reliable way. In Kalu, the functioning RRT had carried out a surveillance survey for early warning. This predicted that around 46,000 people, in addition to the PSNP caseload, would be in need on emergency support in The official figure agreed at regional / federal level is only 14,000 people PSNP impacts on markets and food prices Food price inflation and the real value of PSNP cash transfers Food price inflation in Ethiopia has been significant in recent years. After an extended period of steady but low to moderate inflation it took nine years, from 1998 to 2006, for food prices to double prices of a standard food basket have redoubled again in the past 18 months. Figure 17, derived from CSA data, shows that food prices are leading, after lagging behind non-food price inflation for many years, but that non-food prices are also rising faster than before, probably because of rising fuel prices and their knock-on effects on transport and other costs. However, it is rapidly rising food prices, closely related to global high food prices trends, that are dragging up the overall CPI fastest. 9 It is important to note that, at the time of interview, Tigray had not yet reached the season in which large numbers of people would annually experience food shortage and require emergency support. 76

83 Figure 17 Consumer Price Index (CPI) in Ethiopia, Source: Central Statistical Agency The price of food has been higher every year than the year before, but usually by a margin of 10-15%. Since mid-2007, however, the margin has widened and since early 2008 the price of food has accelerated to an unprecedented extent (Figure 18). This is important for the PSNP because rising food prices erode the purchasing power of cash transfers in terms of food, and the primary intention of PSNP cash transfers is to provide market access to food. The data presented in Figure 18 cover the programming period of the PSNP to date from January 2005 to mid-2008 and this reveals how the value of PSNP cash transfers has collapsed to less than half of its initial purchasing power within four years. This is one of two factors explaining the switch away from cash transfers towards food transfers by a net 1.8 million PSNP participants during 2006/07 (Pelham 2007), which is partly explained by beneficiaries expressing their preference for food which retains its real value, over cash which does not. The other factor is administrative capacity and preferences of administrators at the woreda level. Figure 18 Food price inflation in Ethiopia, Source: Central Statistical Agency 77

84 Markets During fieldwork for this study, 22 communities were asked about their nearest markets. 9 of 22 markets are weekly markets, and 8 of 22 markets are held twice weekly. Only 1 of 22 markets is a daily market, but 4 of 22 are held six days a week, with only one day off. Food retailers were found in every market, and most markets also had livestock traders and clothes sellers, but food wholesalers were active in only half these markets (12 of 22). Across all markets, there were on average 20 food wholesalers, 54 food retailers, 82 livestock traders and 117 clothes sellers. The average distance to the nearest market in walking time was one hour (57 minutes), the furthest market being three hours away (180 minutes) (see annex Error! Reference source not found.). Access to markets does not appear to have been a problem for these communities Food prices Staple food prices in the study areas averaged slightly over 4 birr per kilogram, ranging from 3.46 to 8.13 (Figure 19). Figure 19 Average price of food crops in Ethiopia, 2007/08 (birr/kg) Figure 20 shows the range of prices in local markets within the last year (mid-2007 to mid- 2008). The line represents the high-low range for each crop, while the square block on each line indicates the price prevailing on the day the market was monitored for this evaluation. In most cases the observed price is equal to or close to the highest price, because food prices have been rising rapidly during 2008, as seen in Figure 18 above. 78