Report No. 48230-NE Republic of Niger Impacts of Sustainable Land Management Programs on Land Management and Poverty in Niger February 2009 Environmental and Natural Resources Management Africa Region Document of the World Bank TABLEOFCONTENTS I WhyanImpactAnalysisofSustainableLandManagementProgramsinNiger? . ......1 I1 Poverty and LandDegradationinNiger . .................................................................... 17 I11. SustainableLandManagement Programsand Practices inNiger. ........................... 21 IV Methods ..................................................................................................................... 26 4.1 Data Sources and Sample Selection .................................................................. 26 4.2 DataAnalysis ..................................................................................................... 29 4.3 Strengths and Limitations .................................................................................. 30 V Findings ....................................................................................................................... . 32 5.1 Coverage of SLM Programs .............................................................................. 32 5.2 Community LandManagement Investments ..................................................... 34 5.2.1 Tree Plantations ..................................................................................... 34 5.2.2 Areas Protectedfor Natural Regeneration ............................................. 42 5.2.3 Other Community SWC Investments .................................................... 44 5.3 Household LandManagement Practices ........................................................... 44 5.4 Impacts of Poverty on Adoption of Household LandManagement Practices ..49 5.5 Production Impacts and Profitability o f Household LandManagement Practices ............................................................................................................. 49 5.6 Impacts of SLMPrograms on Crop Production andIncomes ........................... 52 5.7 Other Impacts Reported by Participants inSLM Programs .............................. 55 5.8 Summary of Main Findings and Implications ................................................... 61 VI.Recommendations ....................................................................................................... 68 Annex 1 Projectswith SLMComponentsinNiger . ........................................................ 74 Annex 2 . Description of Common LandManagementPractices .................................... 77 Annex 3 . Cumulative DistributionPlots of Yields With andWithout Practices............81 References ......................................................................................................................... 67 ... 111 TABLES Table 111.1:NRMOutputs of SelectedDevelopment Projects inNiger ........................................ 23 Table V.l: Characteristics of Major SLMProgram andNon-program Villages inNiger ............32 Table V.2: Access to Markets. Infrastructure. Services and Inputs in Program and Non-program Villages inthe Community Survey Sample ........................................................................... 33 Table V.3: Mean Characteristics of Program andNon-program Households ............................... 34 Table V.4: Mean Values of Community LandManagement Investments .................................... 35 Table V.5: Estimated Aggregate Benefits and Costs of Plantations andProtectedAreas ............39 Table V.6: Plot-LevelLandManagementDecisions by Program Type (meanvalues) ................45 Table V.7: Estimated Impacts of Programs on Plot-Level LandManagement ............................. 48 Table V.8: Crop yields on plots with andwithout zai and organic and inorganic fertilizer ..........50 Table V.9: Estimated Impacts of Land Management Practices on Value of Crop Production (percentage change) ............................................................................................................... 50 Table V.10: Estimated Impacts of Programs onValue of Crop Production and HouseholdIncome per Capita ............................................................................................................................... 53 percentage change ......................................................................................................................... 53 Table V.11:EstimatedImpacts of Programs on Value of Crop Production and HouseholdIncome per capita by Wealth of Household ....................................................................................... 54 Percentage change ........................................................................................................................ 54 Table V.12: Summary of Findings andImplications .................................................................... 63 Table VI : Matrix of Recommendationsfrom Stakeholder Working Groups .1 ............................. 70 iv FIGURES Figure I11: Yields of Major Crops inNiger. 1980-2006 . ............................................. 17 Figure 11.2: Per Capita Food Production inNiger. 1990-2006 .................................... 18 Figure 11.3: Overall Trends inGreenness of Vegetation inthe Sahel. 1982-2003 .......19 Figure 11.4: Overall Trends inthe ResidualNDVI. 1982-2003 ................................... 19 Figure IV : Locations of Villages Participating inSLMPrograms inNiger .1 ...............27 Figure IV.2: Locations of SelectedProgram andNonprogram Villages ....................... 28 Figure V.l: EstimatedMeanInternal Rate of Returnto Community Tree Plantations and ProtectedAreas ......................................................................................... -38 Figure V.2: Reported Qualitative Impacts of Community Tree Plantations................-41 Figure V.3: Reported Qualitative Impacts of Community Protected Areas .................41 Figure V.4: Perceptionsof PAC Impacts from PAC Participants ................................ 55 Figure V.5: Perceptionsof PSPRimpacts from PSPR Participants ............................. 57 FigureV.6: Perceptionsof PGRN Impacts from PGRN Participants ........................... 58 Figure V.7: Perceptionsof Program Impacts by Other Government Program Participants ................................................................................................................... 59 Figure V.8: Perceptionsof Program Impacts by Nongovernment Organization Program Participants ................................................................................................ 60 Figure A3.1: Matched Plots With and Without zai ........................................................ -81 Figure A3.3: Matched Plots With andWithout Inorganic Fertilizer ............................. -81 Figure A3.2: Matched Plots With andWithout Organic Fertilizer ................................. 81 Figure A3.4: Matched Plots With andWithout Stone Bunds ........................................ -81 Box Box 111.1: Project-led LandRehabilitation inthe IlldaDistrict ofNiger ..............12 PHOTOS Photo A2.1: BanquettesUnder Construction ................................................................. 79 Photo A2.2: Demi-LunesBeingConstructedby Community Workers ........................ 79 Photo A2.3: Stone Bunds ............................................................................................... 79 Photo A2.4: Zai .............................................................................................................. 80 V FOREWORD Out of deep concern for the degradation of its natural resources, and fully aware of the linkages between that disaster and poverty, particularly in rural areas, Niger has undertaken significant measures to promote Sustainable Land Management (SLM) practices with the support of its partners. Several programs and projects have enabled the country to implement tangible efforts in the field which to date, have led to few, if any, socio-economic impact evaluations. I n order to remedy this situation, the government has undertaken this study with the support of the World Bank and the TerrAfrica Partnership. The results of the study clearly demonstrate the positive impact of efforts carried out during the past two decades on the lives of rural communities in the Niger, thus becoming a genuine tool for advocacy and the mobilization of actors for SLM. These results also highlight the need to pursue actions in order to extend said benefits to the entire country. In this regard, the Rural Development Strategy (RDS) has become a most convenient and suitable framework for the development and implementation of an innovative programmatic approach to SLM promotion. This programmatic approach to SLM, grounded on the development of a Strategic Investment Framework (SIF-SLM), aims to promote the alignment and harmonization of support provided by the various stakeholders (State, financial and technical partners, local governments and civil society), and to further mobilize resources in order to broaden SLM practices in Niger. Through the RDS Joint Ministerial Steering Committee (JMSC/RDS), as well as the existing consultation frameworks, we will further encourage national institutions and technical and financial partners to pursue and strengthen SLM efforts as a way to improve living conditions for the people of the Niger. Minister of Agricultural Minister for t Development, Chairman of the and Desertifi JMSC/RDS Iss Mahaman Y .... 4 5 --, \~ :,e '\ /'.._-__..-- v i ACKNOWLEDGMENTS The core team who worked on this report included Christophe Crepin (World Bank Task Team Leader), John Pender (IFPRI), Jupiter Ndjeunga (ICRISAT), Patrick Verissimo, Taoufiq Benounna, Amadou Alassane and Florence Richard. The team received valuable inputs and guidance from Stephen Mink, Francois Le Gall, Ousmane Diagana, Daniel Sellen, Joelle Dehasse, IdahZ. Pswarayi-Riddihough, Ray Rist and Abdoulaye Toure. The peer reviewers for the study included Erik Fernandes, Leonard Berry (University of Florida) and Bekele Shiferaw (ICRISAT). Beula Selvadurai and Hadidia Djimba provided administrative support to the team. Francois L e Gall and Marjory Anne Bromhead provided management oversight. John Pender andJupiter Ndjeunga are the leadauthors ofthe report. This work has been developed and completed in the context of the TerrAfrica platform for sustainable landmanagement insub-SaharanAfrica. Tahirou Ide and Mohammed Assadeck of the Community Action Program (CAP) provided advice and support throughout the conduct of this work as well as specific comments and suggestions on the report. Chris Reij provided comments on a previous draft of the document; Amadou Abdoulaye and EdwardKat0 provided excellent research assistance; Saidou Koala and other representatives o f ICRISAT provided essential logistical support. Many representatives o f the Government of Niger, of development partners and of research organizations operating in Niger participated in several consultations and workshops conducted during the course of the work, provided valuable comments and suggestions, and provided secondary data and other informationusedinthe report. Any errors and omissions are solely the responsibility o fthe team. The team is most grateful to the Ministers of Environment and Fight against Desertification, Water Resources, Agriculture and, Livestock and Spatial Planning o f the Government of Niger for co-hosting a final national consultative workshop inNiamey inJuly 2008 with more than one hundredparticipants. The team is also gratehl to the more than 2,000 community representatives and rural household heads who participated inthe community and household surveys conducted for this work. Without them this task would not have beenpossible. This report is dedicated to them, to their fellow rural people in Niger, and to their efforts to combat poverty and land degradation. We hope that the information in this report will contribute to broader and more effective efforts to achievethese goals inNiger and elsewhere. vii Abbreviationsand Acronyms ASAP1 Appuipour la SBcurit6Alimentaire par laPetite Irrigation [Support for Food Security by Small Scale Irrigation] AVHRR Advanced Very HighResolutionRadiometer (data) CARE Cooperative for Assistance and ReliefEverywhere, Inc. cdf Cumulative distribution function C D M Clean development mechanism CILSS Comite Permanent Inter-Etats de Lutte contre la Secheressedans le Sahel [Permanent Inter-State Committee for Drought Control inthe Sahel] CIRAD Centre de Coopkration Internationale en Recherche Agronomique pour le Dkveloppement [French Agricultural ResearchCentre for International Development] CRESA Centre Rkgional d'Enseignement Sp6cialis6 en Agriculture [Regional Centre for Specialized Studies inAgriculture] CSIF Country Strategic InvestmentFramework FCFA West African Francs GIS Geographic information system GTZ Deutsche Gesellschafi fiir Technische Zusammenarbeit [German Agency for Technical Cooperation] HIPC HighlyIndebted Poor Countries initiative ICRISAT International Crops ResearchInstitute for the Semi-Arid Tropics IFPRI InternationalFoodPolicy ResearchInstitute IFAD InternationalFundfor Agricultural Development INRAN IstitutoNazionale di Ricercaper gli Alimenti e laNutrizione [National ResearchInstitute for Food andNutrition] IRR Iv Internal rate of return Instrumentalvariables (econometric regression estimator) LTCs Landtenure committees LUCOP Lutte Contre la Pauvret6 [Fight against Poverty] NDVI Normalized difference vegetation index NGO Nongovernmentalorganization NN Nearest-neighbor (econometric matching estimator) NPV Net presentvalue NRM Natural resource management OLS Ordinary least squares (econometric regression estimator) PAC Programme d'Actions Communautaires [Community Action Program] PADER Projet d'Appui au Dkveloppement Rural de Dosso [Rural Development Support Project inDosso] PADL Projet d'Appui pour le D6veloppement Local [Local Development Support Project] PAFN Projet d'Am6nagement des ForQtsNaturelles [Natural Forest Management Project] PASP Projet Ago-Sylvo-Pastoral [Agro-Forest-Pastoral Project] PDRM Projet de D6veloppement Rural de Maradi [Maradi Rural DevelopmentProject] PDRT Projet de Developpement Rural de Tahoua [Tahoua Rural DevelopmentProject] PGRN Projet de Gestion RessourcesNaturelles [Natural ResourcesManagement Project] PIK Projet Int6gr6Keita [Integrated Development Project inKeita] PIP2 Projet de Promotion de 1'Irrigation Priv6e [Private Irrigation Promotion Project] Phase 2 PLCE Programme de Lutte Contre 1'Ensablement du fleuve Niger [Program to Fight Sedimentation o fthe Niger River] PSPR Special Program of the President viii PSM Propensity score matching SLM swc Sustainable land management Soil and water conservation VCR Value-cost ratio All currencies are in US.dollar amounts unless otherwise speciped. ix EXECUTIVESUMMARY A. WHY AN IMPACT ANALYSIS OF SUSTAINABLELAND MANAGEMENT PROGRAMSINNIGER? 1. Since the early 1980s, the Government of Niger and its development partners have invested more than 200 billion West African Francs (FCFA) in programs to promote sustainable land management (SLM) and other activities to reduce poverty andvulnerability. Overall, more than50 programs have promoted SLMinNiger. 2. Despite large investments in SLM programs, their impacts on land management, agricultural production, poverty, and other outcomes are not well known. A few studieshave documented impacts of particular projects andlandmanagementpractices in selected locations, finding many favorable impacts. However, although these studies provide valuable insights, they are limited in scope and by the methods used. A common problem is the absence o f suitable counterfactual observations to compare to outcomes for communities and households participating in programs or using particular land managementpractices, or inadequatedefinition of the counterfactual used. 3. This analysis is intended to contribute to knowledge about the impacts of SLM programs in Niger by addressing some of the methodological limitations of prior studies. The evaluation is based on a secondary database o f major SLM program activities and village characteristics assembledfor all villages of Niger, and a community and household survey conducted with more than 1,200 households in 139 villages selectedto represent the rural regions of Niger where most SLMprograms have operated. Threats to external and internal validity ofthe findings were addressedby usingstatistical sampling methods to assure representativeness of the findings, and quasi-experimental matching and econometric methods to assure that the program and counterfactual non- program villages and households were as comparable as possible in terms of observable characteristics that affect program placement, participationand outcomes. B. POVERTYANDLAND DEGRADATIONNIGER IN 4. The interlinked problems of poverty, vulnerability, land degradation and low agricultural productivity are severe in rural Niger. Niger is one of the poorest countries in the world, ranking at the bottom of countries in terms of the Human Development Index, and with two thirds of the rural population living in absolute poverty. Malnutrition increased in Niger during the 1990s and the rates of infant and child mortality are the highest in West Africa. Average crop yields have increasedonly modestly since the famines o f the 1980s and food production per capita has remained stagnant, with hightemporal and spatial variability. 5. Land degradation is an important factor contributing to low agricultural productivity,poverty, and other social and environmentalproblems. ManyNigeriens view land degradation as one o fthe main causes of poverty and vulnerability, along with population growth and drought. Soil fertility depletion and soil erosion are major problems in both croplands and rangelands, resulting fiom low and declining use of 1 fertilizers, declining fallow periods, expansion o f cultivation into marginal lands, overgrazing o f rangelands, deforestation, droughts, landtenure insecurity, erosive rainfall andwinds, andother factors. 6. Not all trends in land managementin recent decades are negative, however. Since the early 1980s, there has been a well documented "re-greening" o f much o f southern Niger and many other parts o f the West African Sahel. Although improved rainfall since the early 1980s i s responsible for much o f this recovery, there are large areas o f Niger in which the re-greening exceeds what can be explained by changes inrainfall alone. Other factors, including S L M programs as well as indigenous responses to the famines o f the 1970s and 1980s and changes in government policies are cited by some observers as important causes o f improved land management. However, rigorous evidence o f the impacts o f S L M programs on landmanagement, poverty and other outcomes i s lacking. c.SUSTAINABLE LAND MANAGEMENT PROGRAMSAND PRACTICESINNIGER 7. At least 207 billionFCFA are estimatedto have been spent by programswith SLM components in Niger since the early 1980s, considering31 of the largest programs. However, this figure excludes two o f the largest programs: the Projet IntCgrC Keita (Integrated Development Project in Keita; PIK), estimated to have spent $75 million between 1984 and 1999; and the Special Program o f the President (PSPR) (estimated costs unavailable). Other large programs include the Projet de Gestion Ressources Naturelles (Natural Resources Management Project; PGRN), which ended in 2001 and the Programme d'Actions Communautaires (Community Action Program; PAC), which began in 2003. Both PGRN and PAC were/are supported by the World BdInternational Development Association. 8. Sustainable Land Managementis one importantpriority in Niger, however,not the only one. Therefore, importance given to land degradation through these large SLM programs i s to be considered in light o f competing demands for public and private financing. In this context, review o f and guidance to improve effectiveness o f SLM investments is critical. 9. Many of these programsare multipurposedevelopmentprogramsfor which natural resourcemanagement(NRM) activitiesare only one component. A rough estimate is that about one third o fthe total expenditures o fthese programs are for NRMactivities. 10. Most of these programs have promoted water harvesting and soil and water conservation (SWC) measures, tree planting, and other measures to rehabilitate lands. Water harvesting and SWC measures that are commonly promoted include improved planting pits (zai), half-moons, stone bunds, banquettes (embankments with trenches), small dikes, and water spreading dams. Vegetative measures promoted include tree nurseries and plantations, vegetative bands, windbreaks, areas protected for natural regeneration, and sand dune fixation. At least 250,000 hectares o f land have been rehabilitated with such measures - mainly with tree plantations and SWC measures. In addition, regeneration o f naturally occurring trees is reported to have occurred on more than3 millionhectares ofprivate land(Adamet al. 2006). 2 11.The reported outcomes of these investmentsinclude increasedvegetation, reduced erosion, rehabilitation and increased use of degraded land, increased agricultural yields, more fodder for livestock, improved water availability, improved food security, improvedwelfare of vulnerable groups, and reducepoverty, among others. For example, Reij and Steeds (2003) reported that an International Fundfor Agricultural Development (IFAD) project in IllCla District of Tahoua Region rehabilitated 9,000 hectares of degraded land, mostly using zai on private land, leading to incremental annual economic benefits of $65 per hectareresultingfrom increasedcrop production, compared to annual investment costs o f $250 per hectare. According to Hassane, Martin and Reij (2000), this project earned anestimated 20 percent economic rate o f return, contributed to development ofthe local landand labor markets, and ledto diffusiono fzai to other areas. Ambouta, Moussaand Daouda (2000) estimated a 30 percent reduction inwater andwind erosion, as well as reduced runoff and increased water infiltration resulting from interventions of various projects in the Tahoua region. Adam, et al. (2006) report improvements in a wide variety of outcomes in the project villages studied in Maradi, Tahoua and Tillaberi regions, compared with little improvement in the non-intervention villages studied. As noted earlier, however, these studies are beset by various methodological limitations, some of which are addressedinthis assessment. D. METHODS 12. This analysis is based on secondary data collectedfor every village in Niger, and on a community- and household-level survey conducted in 139 villages with 1,218 households in the DOSSO, Maradi, Tahoua, Tillaberi and Zinder regions. The secondary database includes information on the presence and activities of 12 of the largest SLM programs as well as on village characteristics, such as geographic coordinates, agroclimatic zone, dominant soil type, distance to Niamey, distance to the nearest town, and access to infrastructure and services. This database was used as the sample frame for selecting the sample villages. 13. The survey sample excludedAgadez and Diffa regions, because few villages in these regions participated in SLM programs. A sample of 70 program villages was selected using a random sample, stratified by region and major program type, and 69 matching non-program villages were selected from the same region and agroclimatic zone, and at a similar distance to an urban center, as the selected program villages. The purpose of this sampling procedure was to assure adequate representation of major regions and SLM programs, and to include counterfactual nonprogram villages that are similar inagroclimatic and market access characteristics. 14. Within the sample program villages, six program participants and six non- participants were randomly selected for the household survey; in non-program villages six households were randomly selected. The purpose of selecting these three groups was to be able to investigate direct and indirect impacts of SLMprograms. Since non-participants in program villages are more likely to be affected by indirect effects of the programs than households in non-program villages, comparisons of outcomes betweenthese groups helps to assess such indirect effects; while comparisons between 3 participants and non-participants inprogram villages enables assessment of direct effects of program participation. 15. Matchingmethods and econometricestimatorswere usedto assure comparabilityof households in these different groups and to estimate the impacts of programs on household-level land management practices, crop production and income. Both propensity score matching (PSM) and nearest neighbor matching with bias correction (NN) were used, because each method has advantages and disadvantages. In addition, probit, ordinary least squares (OLS) and instrumental variables (IV) regression models were used, andthe robustnessofthe findingsto analytical methods investigated. 16. The economic costs and benefitsof household levellandmanagementpracticeswere estimated based on the matchingand econometric analysis of the householdsurvey data, and at the community level using the community survey data. In addition, community members' and households' perceptions of impacts o f SLM programs and community land management investments on a wide range of social and environmental indicators were investigated. 17. The analytical methods used have several strengths, including using a large and representative survey to assure broad representativeness of the findings, using quasi- experimental methods to assure appropriate counterfactuals in assessing impacts, enabling investigation of program spillover impacts within program villages, controlling and testing for biases due to "selection on unobservables" using IV methods, estimating economic costs and benefits of SLM programs and land management practices, and combining qualitative with quantitative datato deepenthe insights gained. 18. There are also limitations of the methods used, including the absence of suitable baseline data, lack of use of remote sensingor other historical data on changes inlanduse and management, lack of objective measurable indicators of social and environmental impacts, focus mainly on private costs and benefits, and inability to explore in depththe reasons for success or failure of SLM programs, due to budgetary, time and other constraints. However, this assessment has provided a foundation upon which further actions as well as complementary work canbe built to addressthese limitations. E. FINDINGS COVERAGE SLMPROGRAMS 1: OF 19. Despite the large investmentsin SLM programs, their coverage has been limited and focused in areas of more favorable agro-climaticconditions and marketaccess. Less than 8 percent of villages in Niger participated in any o f the 12 major SLM programs included inthe national village database. Program villages are more common in the southern, higher rainfall areas of the country, closer to Niamey and to the nearest town, and inareas with degradedsoils. These are areas with lower poverty indicators than other areas, but are where the program impacts are likely to be more favorable. Even withinour community sample (for which program andnon-program villages have similar agro-climatic conditions and access to towns), program villages have better access to infrastructure and services andpeople inthese villages are more likely to use agricultural inputs. Thesedifferences may reflect impacts ofthe programs, butthis is not certain. 4 20. Within program villages, participants in several types of programs tend to be wealthier in some respects than non-program households. For example, households participating in PAC own more irrigated land, livestock, farm equipment and household assets than either nonparticipants inPAC villages or householdsinnon-program villages. However, PAC households had larger endowments of these assets even in 2000/1, prior to implementation ofthis program. Hence, the greater current wealth o fPAC participants reflects at least inpart a tendency for wealthier households to participate inthe program. Similar findings (greater wealth in.2000/1 of program participants) are evident for participants in PSPR, other government programs and NGO programs. Further analysis i s necessary to determine whether these programs also contributed to increased wealth, by promoting increased income. By contrast, participants inPGRN had less household assets than non-participants in 2000/1 (while that program was in operation); though without baseline data from before that program it i s not clear whether this reflects pro- poor targeting of PGRN. F. FINDINGS2: COMMUNITYLAND MANAGEMENT INVESTMENTS 21. Tree plantations are the most common community land management investment promoted by SLM programs. Tree plantations were found in 90 percent of program villages but inonly three non-program villages inthe sample. Program involvement thus appears essential for widespreadadoption oftree plantations to occur. 22. The average tree survival rate in plantations is only 54 percent, although there is substantialvariationacross plantations. Survival rates are significantly higher inPAC promoted plantations and for certain tree types (Azadirachta indica (neem) and Acacia seyal). SWC measures, especially half-moons, are commonly used in tree plantations. However, no significant impact of such measures on tree survival was found using regression analysis. Further investigation of this i s needed, but this result suggests that targeting inthe use of SWC measures inplantations needto be improved. 23. Acacia Senegal is the most commonly planted tree in plantations, yet little gum Arabic is beingharvested. Inpart, this reflects the young age of most plantations, but even inolder acacia Senegal plantations little gum Arabic is harvested. This representsa missed opportunity to obtain economic benefits from such plantations, and reflects numerous constraints affecting the harvesting, handling and marketing of gum Arabic. Investigation of the key constraints and appropriate remedies are needed to achieve the economic potential ofthese plantations. 24. The main benefitsof community tree plantationsare the value of wood and fodder produced. The average estimated value of the wood in tree plantations is about 17.2 million FCFA per plantation (368,000 FCFAhectare, considering the average size of 47 hectares), representing about 7 percent o f the average wealth of all householdsinvillages with plantations. The annual value of fodder produced is estimated to average in the range of 1.4 million to 2.8 million FCFA per plantation (31,000 to 61,000 FCFAhectare) using two different estimation methods. This is much larger than the value of other materials collected from plantations, which averagedonly 6,000 FCFA per plantation. 5 25. On average, community tree plantations are highly profitable to the communities investing in them, although there is large variation in returns across plantations. The average annual benefit cited above is larger than the average cost (to the communities) of establishing plantations - about 1.3 million FCFA per plantation. However, there is very large variation in returns, with only about one-fifth of the sample plantations earning positive benefits in 2006. Much of the reason for this is the recent nature of most of the plantations earning zero income (two thirds o f these planted after 2002); but many older plantations are also earning zero benefits. Thus, despite high potential for profitability, most plantations are not yet realizing this potential. The main sources o f variation in the profitability o f tree plantations are variations in tree survival rates, labor investment costs, and the value of fodder browsed by livestock. This suggests that the profitability o f plantations could be substantially improved in many cases through improved targeting of tree species and use of improved management practices (such as watering seedlings) to increase tree survival rates; better targeting of labor-intensive SWC investments to ensure that they are effective in improving tree survival and that such costs are avoided or reduced where unnecessary; investments in livestock management to improve utilization of available fodder in plantations; and increased harvesting and marketing of valuable products such as gum Arabic. Still, the estimated value of wood in the plantations exceeds the investment costs for almost all plantations, so even though current benefits are often low, these investments are paying off by increasing the wealth of communities. 26. Considering the full costs of establishing plantations (including costs supported by SLM programs), tree plantations are an attractive investment option for SLM programs. Based on the average benefits estimated from the survey and the costs of establishing plantations reported inthe literature, the internal rate of return(IRR) for tree plantations is estimatedto range between28 percent (low fodder value estimate) and 45 percent (high fodder value), while the estimated net present value (NPV) (at 10 percent discount rate) ranges between 125,000 and 307,000 FCFNha. If gum Arabic were harvested and marketed (considering plantations o f Acacia Senegal), the IRR would increaseto a range of 33 percent (low fodder value) to 48 percent (high fodder value). 27. The estimated aggregate economic benefits of tree plantations are greater than the amount estimated to have been spent by programs on NRM activities. Considering the average estimated NPV of fodder and wood produced per tree multiplied by the number of trees plantedwith support of programs andtheir average survival rate, the 65 million trees planted with support of the programs reviewed are estimated to yield an aggregate NPV of 71 billion to 111 billion FCFA, larger than the roughly 69 billion FCFA (one third of total program costs) estimated to have been spent by these programs onNRMactivities. Hence, the value o ftree plantations alone appears sufficient tojustify the investmentsmadebythese programs. 28. The net benefits of tree plantationswould be even larger if the value of social and environmental benefits were included. For example, valuing the amount of carbon sequestered at $4.20 per ton of C02 equivalent (tCO2e) - the amount to be paid by a BioCarbon Fund project in Niger, but less than the average amount being paid internationally for carbon credits ($10.90 per tCO2e on average in2006) -the IRR of tree plantations (including the value of gum Arabic) increases from 33 to 35 percent (low 6 fodder value) or 48 to 49 percent (high fodder value). Hence, includingthe carbon sequestrationvaluewouldmodestlyincreasethe profitabilityof tree plantations,but i s notamajor determinant of social profitability. 29. Many other social and environmentalbenefits of tree plantationsare also reported by community members - including improvements in the availability o f fodder, fuel wood and useful wild plants; reduced soil erosion; and improved food security, revenues and welfare. However, most o f these benefits were not included in the estimated value (except income benefits). In some cases, tradeoffs are apparent among some social and environmental indicators. Such potential tradeoffs should be considered when tree plantations or other investments are promoted. 30. Some communities have protected communal areas for natural regeneration of vegetation, in most cases promoted by a SLM program but also in several non- program villages. Program involvement appears less essential to establish protected areas thanplantations, probably because o f lower costs andtechnical expertise required. 31. The main economic benefit of protected areas - the value of fodder material produced - is less in protected areas than in plantations, but much greater in protected areas promoted by a program than in non-promoted protected areas. Thus although program involvement is not essential to establish protected areas, it appears to increase their benefits. 32. Although the benefits of protected areas are less than those of tree plantations, the IRRfor protectedareasis higherbecausethe costs of establishmentare muchlower. The estimated mean IRR for promoted protected areas i s greater than 170 percent (low fodder value estimate) and 46 percent for non-promoted ones. Still the N P V h a o f protected areas i s less than that o f plantations (at 10percent discount rate), indicating that plantations are a more profitable land use if the financial and technical constraints are overcome. 33. Overall, protected areas promoted by programs have generated a total estimated NPV of about 2 to 3 billion FCFA. This is much less than the aggregate benefit estimated for plantations, both because o f much smaller areas o f promoted protected areas and because o f the lower N P V per hectare o f protected areas. Nevertheless, this still contributes significantly to the aggregate benefits o fthe programs. Inaddition, as for tree plantations, protected areas are perceived to result inmanypositive impacts on social and environmental indicators (increased availability o f fodder and beneficial wild plants andanimals; improved food security, revenues andwelfare). 34. SWC investmentsare also made in other communal areas (without tree plantation or protection) in 14 percent of villages, almost always promoted by programs. These investments were more costly than protected areas and yielded very low economic returns. Only 5 percent o f these areas yielded any positive benefits in2006, andthese were small even when positive. 35. Environmental and social benefits were reported for these SWC investments in many cases, though less commonly than for plantationsor protected areas. Given the relatively high costs and low returns o f these investments and less favorable perceived impacts, SLMprograms should reconsider whether and when to promote these. 7 G. FINDINGS HOUSEHOLDMANAGEMENT 3: LAND PRACTICES 36. Land managementpractices used by farmers in Niger include (in decreasingorder of use) organic fertilizer, inorganic fertilizer, zai, half-moons, stone bunds, tree planting and mulching. Organic fertilizer is used on more than 40 percent o f plots in the sample and inorganic fertilizer on 18 percent; all other practices are used on less than 10 percent o f plots. When inorganic fertilizer i s used, it i s used in "micro-doses" averaging only about 20 kg per ha. Tree plantingi s only found on 4 percent o f plots, and only 12trees per ha were plantedon average when plantingoccurred. 37. Although tree plantingis rare, trees are commonly found on households' cultivated plots. More than three-fourths of the sample plots have trees on them, and about one- thirdhaveFaidherbiaalbidatrees, a leguminous tree that loses its leaves duringthe rainy season and hence can be beneficial for crops. These findings are consistent with the findings of other studies, suggesting that farmer managed natural regeneration of trees is occurringon a largescale inNiger. 38. SLM programspromote adoption of several land managementpractices,especially water harvestingand SWC practicessuch as zai, half-moonsand stone bunds.These impacts are not only upon program participants, but also affect nonparticipants in program villages. Program impacts on tree planting and protection by households are more limited, and no significant impacts o f programs on use o f organic or inorganic fertilizer were found. 39. SLM programs can have durable impacts by promotingdurable investments, such as SWC measures and tree planting. They also may affect adoption o f land management practices even after the programs end, by increasing farmers' awareness o f such practices and their benefits. Available evidence for PGRN, which ended in 2001, does not show this latter impact, however. No significant impacts were found o f PGRN on adoption o f landmanagement practices since 2001. H. FINDINGS IMPACTSOFPOVERTYONHOUSEHOLDMANAGEMENT 4: LAND PRACTICES 40. Adoption of many land management practices is inhibited by aspects of poverty. Farmers with less access to irrigation are less likely to adopt zai, stone bunds, and inorganic fertilizer. Households with less male labor are less likely to invest in stone bunds andhalf-moons. Households with less livestock are less likely to invest inorganic fertilizer. Land tenants are less likely than owners to invest in zai, stone bunds, or organic fertilizer. These findings reflect the fact that poverty may undermine sustainable landmanagement by limitingfarmers' access to complementary assets (e.g., irrigation or livestock) or by causing tenure insecurity. 41. However, some aspects of poverty contributeto adoptionof some land management practices. For example, households with more land are less likely to invest in stone bunds on a given plot. Households with more livestock are less likely to use mulch. Households with more farm equipment are less likely to invest in stone bunds. 8 Households with more durable household assets are less likely to invest in stone bunds, planttrees, or use organic fertilizer. More educatedhouseholds are less likely to investin half -moons. Many of these findings may be because wealthier households have higher labor opportunity costs and are thus less prone to use labor intensive land management practices. Such complex relationships between household endowments or other aspects o f poverty and land management decisions are consistent with theories o f agricultural householdbehavior inthe presence of imperfect markets, and are found ina largenumber of studies in Africa. Thus, investments to reduce poverty are not likely to be sufficient to reverseland degradation. I.FINDINGS5: PRODUCTION IMPACTS AND PROFITABILITY OF HOUSEHOLD LAND MANAGEMENT PRACTICES 42. Zai, organic fertilizer and inorganic fertilizer significantly increase mean crop yields and reduce the risk of low yields. Usingmatching and econometric methods we find that crop yields on plots with zai are as much as 24 percent higher on average than yields on comparable plots without zai. Organic fertilizer increasesyields by as much as 33 percent, while inorganic fertilizer increases yields by as much as 19 percent. All of these practices, as well as stone bunds, reduce the probability o f low yields and have a more favorable impact on low productivity plots. None of the other practices considered (half-moons, mulch, trees) has significant and robust impacts on yields. 43. Based on the estimated production impacts and costs of these practices, zai appear unprofitable, organic fertilizer breaks even and inorganic fertilizer is marginally profitable on average for the sample households. The mean estimated impact o f zai increases income by about 10,000 FCFA per ha, compared to an estimated average initial cost of 50,000 FCFA per ha and annual costs of 24,000 FCFA per ha. These results indicate that zai are not profitable (at mean estimated returns and costs) irrespective o f the discount rate used to evaluate profitability, since the value of the mean increase in production is less that the estimated annual cost of maintaining zai. The mean impact of organic manure increases income by 11,000 FCFA per ha, but also costs about 11,000 FCFA per ha in labor costs to apply. Inorganic fertilizer use has an estimated marginal value cost ratio of 1.6, indicating marginal profitability. 44. These estimates underestimatethe profitability of zai to the extent that plotswith zai were unusable prior to rehabilitationwith zai. It is often claimed that zai plots were not usable prior to rehabilitation with zai; if this is the case, the proper counterfactual to production with zai is zero production. Unfortunately, the IFPRVICRISAT survey did not investigate the use of zai plots prior to establishment of zai, so this couldn't be investigated. The matching methods suggest that many cultivated plots without zai are similar insoil characteristicsand other characteristics to zai plots, however. 45.Even if zai and organic fertilizer were not profitable on average in the survey sample and year, they may be profitable in other years or other locations, or for households with low labor opportunity costs. The rainfall in the study year of 2006 was relatively favorable in Niger, and this may have affected the impact of water harvesting measures such as zai. Zai may also be more profitable inparticular locations 9 than in the broad sample in this survey. Furthermore, even if these measures are unprofitable if labor is valued at the market wage level, they may be profitable for householdswho have low labor opportunity costs at the time whenzai are constructed, or may be attractive for their riskreducing advantages. J. FINDINGS IMPACTSOFS L MPROGRAMS ONCROP PRODUCTIONAND INCOMES 6: 46. Most S L M programshave limited impact on the value of crop production. PAC is an exception, which is found to have a robust negative impact on crop production. Regardless of the estimation method used, PAC households are found to have significantly lower value of crop production per hectare (30 to 35 percent lower) than comparable nonparticipants in PAC villages. This may be due to promotion of other income generating activities by PAC. Supporting this i s the finding that the income of PAC households is not significantly different from that of nonparticipating householdsin PAC villages, which suggests that PAC households reduce crop production as they pursueother activities, but compensateby having higher income from other sources. 47. Major reasons for limitedimpactsof S L M programson crop productionincludethe limited impact of the programs on household land management practices, the limitedimpactof the practicespromotedon productivity,and the emphasis of these programs. Even though SLM programs do have significant impacts on household adoption of several land management practices, the increase inprobability of adoption is less than 10 percentage points in most cases. Furthermore, some of the practices promoted, such as half-moons and stone bunds, have limitedimpacts on crop yields. Soil fertility management practices, which do significantly boost production, are not much promoted by these programs. Finally, much of the emphasis of SLM programs is on community level activities such as tree planting, and on other activities besides crop production suchas establishing cereal banks, livestock fattening, and others. 48. Although the programs have limited impact on crop productionand on incomes of participants compared to non-participants in program villages, they appear to contributesignificantlyto higherincomesinprogramvillages ingeneral. Incomes of nonparticipants in program villages are found to average as much as 23 percent higher than comparable households in nonprogram villages, suggesting a broader community level impact of the programs on household incomes. 49. The impactsof S L M programsvary across wealthier and poorer households, across the country's different regionsand agroclimaticzones, and by type of program. For example, the negative impact of PAC participation on crop production was found only among the wealthier half of the sample; no impact was found for poorer households. K. FINDINGS OTHERPROGRAM IMPACTSREPORTED BYPROGRAMPARTICIPANTS 7: 50. Householdparticipantsin S L M programsgenerally report favorable impacts of the programson a wide variety of environmental,socialand economic indicators. More than half of participants in all SLM programs except PGRN reported increases in the 10 availability o f useful trees, while most participants in PSPR, other government programs and NGO programs reported increased availability and quality of pasture and fodder, increased availability and quality of cropland, increased agricultural productivity and food security, and improved welfare. Clearly, not all of the benefits of these programs are measuredby impacts on crop production and income. 51. Positiveimpactson these indicatorswere less commonly reportedby participantsin PAC and PGRN, though favorable impacts were still reported by a sizable fraction of participants in these programs. More than 50 percent o f PAC participants reported favorable impacts for only one indicator - availability of useful trees - while fewer than 50 percent o f PGRN participants reported favorable impacts on every indicator. The dominant response was of "no significant impact" for these programs. Further study is needed of why participants in PAC had less favorable impressions than participants in other programs. 52. Across programs, favorable impacts were less commonly reported for several indicators,includingthe presence of predatorsor pests, the availabilityand quality of water, and conflictsover naturalresources. Most participants inall programs report no significant impact o f the programs on these indicators, and for some indicators, a small percentage of participants reported unfavorable impacts. These results indicate that some potential trade-offs can result from SLM programs, although these appearto be uncommon. L. IMPLICATIONSOFTHEFINDINGS 53. Much more investment in SLM programs is neededto reach most of Niger's areas and people. The programs appear to be targetedto locations where they are more likely to be effective, but many less favorable areas are being left out. Within program communities, more effort i s neededto reachpoorer households. 54. Giventhe high average returns to community tree plantationsand protectedareas, more investment in such community SLM investments is economically justified. SLMprograms appear to be essentialto achieving these high returns; these benefits are unlikely to berealized on awider scalewithout expansion of suchprograms. 55. Giventhe highvariability of returnsto communitytree plantations,efforts to better target and increasethe effectiveness of these investmentsare needed. Of particular importance are efforts to increase the harvesting and marketing of gum Arabic from Acacia Senegalplantations. Promotion of other high value trees, investments inlivestock improvement, andbetter targeting o f SWC measurescould also improve returns. 56. Effortsto measureandvalue the non-economicsocial and environmentalbenefitsof community land management investments, as well as trade-offs among different outcomes, would be useful. The case for programs promoting these investments is stronger than the economic benefits alone, but further researchi s neededto quantify how important these other benefits are andwhat trade-offs exist. 57. Increased emphasis on promoting improved soil fertility management, and better targetingof use of SWC measures, is needed in programspromotinghouseholdlevel 11 land management practices. Despite the higher returns and risk reducing benefits of soil fertility managementpractices, SLMprograms are not significantly promoting these, while some of the practices they are promoting are not very effective in increasing crop production or income. 58. Since the factors affecting adoption of household land management practices and their impacts are very context-specific, a demand driven approachto promotionof such practices is essential. N o one-size-fits-all land management approach will be effective inthe diverse circumstances that farmers face; approacheshave to be adapted to local contexts andto farmers' constraints and capabilities. 59. Addressing poverty will not be sufficient by itself to prevent or reverse land degradation. Although some aspects of poverty constrain SLM, some aspects ofpoverty are associated with greater adoption of sustainable practices. As people become more educated and have higher value opportunities for their labor, they are less prone to adopt some labor-intensive practices. Strategies thus have to be developed to achieve SLM underchanging conditions ofpoverty as rural development proceeds. 60. PromotingSLM will also not be sufficient by itself to eliminate poverty. Although SLM programs are having significant positive impacts on household income, these impacts are not large enoughto bring most rural households out of poverty. Other public investments, such as in infrastructure and education, will continue to be critical for poverty reduction, even as SLMinvestmentsare scaledup. 61. Further researchis needed to understandwhy PAC householdsare obtaining lower value of crop production and have less favorable impressionsof this program than participants in other programs, whether this is a longer term problem, and what ameliorative actions may be needed. The robust conclusion that PAC households are obtaining lower crop production does not necessarily imply a problem, since other sources o f income are being promoted. But this i s worthy of further study and possible action ifit i s a longer term tendency. M.RECOMMENDATIONS 62. The following eleven recommendationswere made for improvingthe effectiveness of actions relating to SLM in Niger by working groups at a stakeholder workshop in July, 2008, basedon the findings ofthis study and other information: Recommendationsrelatedto developmentstrategiesand actions: 1. Develop a political consensus on intensifying SLM integrated with the national Strategy of Accelerated Development and Poverty Reduction and the Rural Development Strategy. 2. Support decentralizedmanagementof natural resources. 3. Increase investments in SLMprograms inthe context of the Rural Development Strategy through the community development plans and enlarge the zones of intervention, includingby further working at watershed levels. 12 4. Continue and extend investments promoting SLM actions with high returns (e.g., plantations, protection areas for natural regeneration, organic and inorganic fertilizer, agroforestry). 5. Develop a global, integrated and efficient approach in SLM programs, including both public andprivate goods dimensions as well as increasedconsideration of climate risks. 6. Continue investing in other public goods and services that will have a large impact on poverty reduction andrural development, based on the priorities identified inthe PDCs. Recommendations related to policy and program monitoring and evaluation, and knowledgemanagement: 7. Implement a system of participatory monitoring and evaluation and knowledge managementat all levels. 8. Replacethe sectoral approachto knowledge managementwith a multi-sectoral approach; includingknowledge of forestry, pastoralism, socioeconomics, and other relevant fields. 9. Synthesize and conserve context-specific knowledge on SLM and disseminate findings in readily usable form to stakeholders. Recommendationsrelatedto teaching, capacity building,researchand technicalassistance: 10. Strengthenteaching of SLMinuniversity courses. 11. Conduct action research and provide technical assistance to address key issues for increasing the effectivenessof SLMinvestments. 13 Why an Impact Analysis of Sustainable Land Management Programs in Niger? 1. Since the early 1980s, the Governmentof Niger and its developmentpartners have invested more than 200 billion West African Francs (FCFA) in programs to promote sustainableland management(SLM) and other activities intendedto reduce the incidence of rural poverty and vulnerability in Niger. Such activities have included large, multi-regional programs, such as the Special Program of the President (PSPR), the Projet de Gestion des Ressources Naturelles (Natural Resources Management Program; PGRN), and the Programme d'Actions Communautaires (Community Action Program; PAC), as well as numerous regional and district-level projects initiatedby the national government or nongovernmental organizations (NGOs). Overall, more than 50 programs and projects have promoted SLM inNiger since the early 1980s. 2. Despite the large magnitude of these investments, their impacts on land management,agriculturalproduction,poverty,and other outcomes are not well known. A few studies, however, have documented impacts of particular projects or land management practices inselectedlocations: Hassan, Martin, and Reij (2000) assessed the impacts on farmers' crop yields of soil and water conservation (SWC) measures promoted project inthe IllCla District of the Tahoua region, which was supported by the International Fund for Agricultural Development (IFAD). 0 The Lutte Contre la PauvretC (Fight against Poverty; LUCOP) project conducted a similar study of the impacts of the SWC measures it has promoted in the Tillaberi region (LUCOP 2005). 0 Adam et al. (2006) synthesized results of a series of studies led by the Centre RCgional d'Enseignement SpCcialisC en Agriculture (CRESA) on the impacts of program interventions and natural resource management practices in 16 villages of the Maradi, Tahoua, andTillaberi regions (comprising 12project interventionvillages and 4 nonintervention villages). The synthesis also included a study by Larwanou, Abdoulaye, and Reij (2006) of farmer assisted natural regeneration of trees in three departmentso fthe Zinder region. These studies reported many favorable impacts of SLM interventions and improved land management practices, including increased crop yields, livestock production, and high- value crop production; increased availability o f useful trees, other vegetation, and fodder and fuel wood; increased availability o f water; expanded cultivable area; improved food security; and reducedpoverty. 3. The aforementionedstudies providemany valuable insights,but they are limitedby the methodologiesemployed. Each study was conducted ina fairly small number of locations, and the selection of the locations was purposeful, limiting the external validity of the findings (the ability to generalizethe findings to other populations or regions not included inthe studies).' The internal validity of the findings in these studies (the ability to attribute the program ' Theconcepts of external and internal validity, and threats to assuring them, are discussed at length in Cook and Campbell (1979). 14 intervention or land management practice considered as the cause of the outcomes measured) was also compromised by the lack o f suitable counterfactual observations.* With the exception o f the village studies synthesized in Adam et al. (2006bwhich did not explain how the nonintervention villages were selected or how they were similar to the intervention villages- non-intervention sites were not studied in most cases. Similarly, some studies of the impacts of land management practices on crop yields compare yields on intervention or "treatment" plots with those on non-intervention or "control" plots (for example, Hassane, Martin, and Reij 2000; LUCOP 2005), but they don't explain how the different plots were selected or how similar they are inaspects that could affect yields. Hence, the differences in yields on these different types of plots could be attributable to factors other than the land management practice being evaluated. Another limitation is that none o f these studies investigates the impacts o f program interventions on measured household income. Although Adam et al. (2006) reports differences between intervention and non-intervention villages in terms of many social and environmental indicators, these are reported perceptions and may be subject to reporting biases, particularly among participants seekingadditional program involvement. 4. The review undertaken for this study is intended to contributeto knowledgeabout the impacts of SLM programs in Niger and to address some of the methodological limitations of prior studies. The current review was based on a secondary database o f major SLMprogram activities andvillage characteristics assembledfor all the villages ofNiger, along with a large community and household survey of more than 1,200 households in 139 villages spread throughout five of the seven rural regions of Niger, where most SLM programs have operated: Dosso, Tillaberi, Tahoua, Maradi and Zinder. The review was thus able to investigate the impacts of several major programs in a wide range o f settings. Threats to both internal and external validity were addressed by using a stratified random sample to select the program intervention villages, and by selecting comparable ("matching") non-program villages from the same region and agro-ecological zone as each program village, and with similar access to markets. Participating and nonparticipating households were randomly selected for the survey, andthen matching households were selectedfor the analysis basedontheir asset and educational endowments and other factors expected to affect program outcomes. Matching methods were also used to select comparable treatment and control plots. In this way, the study ensured comparability not only o f intervention and non-intervention villages, but also of households and plots, thereby reducing the influence of confounding factors. The outcomes measured included both quantitative and qualitative indicators o f perceived changes, enabling assessment of program impacts on a wide array of outcomes, including household income. 'Inassessing the impacts o f an intervention, the "counterfactual" is conceptually defined as the outcome that would have occurred had the intervention not occurred. Since the true counterfactual is not directly observable, proxy counterfactual observations are needed. The most reliable method, when feasible, is the random assignment of treatment and control Observations in an experimental design that assures the groups are statistically identical inall respects other than the intervention. When random assignment is infeasible, alternative quasi-experimental means may be used, such as selecting comparable intervention and noninterventiongroups by matching relevant observable characteristics. This is the approach used inthis study. 15 5. The remainder o f this review is organized as follows: Section I1presents the problem context; Section I11 provides background on SLM programs operating in Niger; Section IV provides a brief discussion of the methodology; Section V presents the main findings and their implications; and Section VI presents recommendations. Readers are referred to the full version o f the report on this study (Report 11) for a more detailed discussion o f the methodology and findings. 16 11. Poverty and LandDegradation inNiger 6. The interlinked problems of poverty, vulnerability, land degradation, and low agricultural productivityare severe in ruralNiger.Niger is one o fthe poorest countries inthe world, ranking at the bottom of 177 countries for which the human development index was computed in 2006 (UNDP 2006). Three-fifths of the population lives in absolute poverty and poverty is even more severe in rural areas, where 85 percent of the populationlives (RN 2002). During the 1990s, malnutrition increased, with the share of stunted children increasing from 32 to 40 percent between 1992 and 2000, and the share of underweight children increasing from 36 to 40 percent inthe same period (RN 2002). Rates of infant and child mortality inNiger are the highest inWest Africa, at 126 and 280 deaths per thousand, respectively, in2000, and the rates are evenhigher inrural areas (RN 2002). 7. Agricultural productivity in Niger is low and highly variable from year to year due to the dry and highly variable climate, Since 1980, average yields of millet-the most important food crop-have averaged only about 400 kilograms per hectare, while sorghum yields have been even lower (Figure 11.1). Crop yields have trended upward since the droughts and famines of the 1970s and 1980s, as rainfall has increased. Nevertheless, per capita food production has remained stagnant, with large inter-annual variations, as rapid population growth of more than 3 percent per year has offset increasedfood production(Figure 11.2). Figure 11.1:Yields of Major Crops in Niger, 1980-2006 Kilograms per hectare 1,000 900 700 -Cowpeas---Groundnuts------Millet _ ~ ~ Sorghum n . Source: FA0 2008. 17 Figure11.2: Per CapitaFoodProductioninNiger, 1990-2006 Index (1990 = 100) 140 2o 0 I 1990 1992 1994 1996 1998 2000 2002 2004 2006 Source: FA0 2008. 8. Land degradation is an important factor contributing to low agricultural productivity,poverty, and other social and environmental problems. Many people perceive land degradation-along with population growth and drought-to be the main cause of poverty and vulnerability (RN 2002). Soil fertility depletion is a major problem in croplands. It results from very low and declining use of inorganic fertilizers, such as occurred following the devaluation o f the FCFA in the 1990s; limited availability and use of organic fertilizers; and declining fallow periods due to rapid population growth (MDA/FAO 2002; Abdoulaye and Sanders 2005; Pender et al. 2008). Henao and Baanante (2006) estimate that Niger i s among the more rapid soil-depleting countries in Africa, with an average depletion rate of nitrogen (N), phosphorus (P), and potassium (K) of 56 kilograms per hectare per year in 2002-4. Also contributing to soil nutrientdepletion, low productivity, and other problems are high rates o f soil erosion caused by both water and wind. Net soil loss rates in parts of southwesternNiger have beenestimated to be in the range of 26 to 46 tons per hectare per year-well above the rate of soil formation (Chappell 1996; Warren, Batterbury, and Osbahr 2001). Soil water erosion not only affects farmers' crop production, but also contributes to serious sedimentation problems in the Niger River Basin; and wind erosion contributes to offsite damages to crops, other vegetation, buildings, equipment, and human health. Land degradation in rangelands, resulting from expansion of cropland into rangelands, overstocking o f remaining rangelands, droughts, and tenure insecurity, is a major concern contributing to low and variable productivity of livestock production (RN 2000, 2003). Depletion of forests and woodlands is also a major concern: between 1990 and 2005 Niger lost nearly 26 percent o f its forest and woodland habitat (Butler2006). 18 9. Not all of the evidence on land degradation in Niger and elsewhere in the Sahel is negative, however. There is accumulating evidence of a "re-greening" occurring in Niger and elsewhere inthe Sahel since the early 1980s (Figure 11.3). Figure11.3: OverallTrends inGreennessofVegetationinthe Sahel, 1982-2003 15w 1O"W 5cw O"E 5% W E 15% 20'E 25"E 3O'E I I 1 I I I 15'w 10'W 5'W O'E 5cE lO*E- 15 N 10 Pi I I I 15cE 2WE 25"E 0 - 400 830 Kilometers NDVIchange(9) Source: Herrmann, Anyamba, and Tucker 2005. Note: NDVI indicates normalized difference vegetation index. Data are expressed as percentage changes in the average NDVI and are based on monthly advanced very high resolution radiometer (AVHRR) NDVI time series data. 10. Much, but not all, of this trend can be explained as a result of increased rainfall in the Sahel since the early 1980s. Even after controlling for the effects o f increased rainfall, Herrmann, Anyamba, and Tucker (2005) found unexplained increasesingreennessinlarge areas of the Sahel, including large parts of the Tahoua and Maradi regions o f Niger (Figure 11.4). This area o f unexplained re-greeninginNiger i s centeredinthe area where Projet IntegrCKeita (PIK), a large natural resourcemanagementproject, operatedfrom the early 1980s. FigureII.4: OverallTrends inthe ResidualNDVI,1982-2003 15RW 1V'W 5'W O"E S3E 10"E 1VE 20% 25rE 30"E 15 N 15 N - I O N 10 N 1YW IOQW S'W 0'E 5 % tO'E 153E 20% 25E 0 400 800 Ktlometers Stope residuats (x1o2) Source: Herrmann, Anyamba, and Tucker 2005. Note: Data are based on regressions of vegetation greenness (AVHRRNDVI) onthree-monthly cumulative rainfall. 19 11. These findings are consistent with other recent "success stories" on improvingland managementin parts of Niger (for example, Mortimore et al. 2001; Mortimore 2005; Adam et al. 2006; and Larwanou, Abdoulaye, and Reij 2006). For example, Mortimore et al. (2001) found ' that despite, or perhaps because of, decreasing availability of natural woodland in Maradi, tree densities on farms were increasing because of the widespread practice by farmers of protecting natural on-farm trees o f value. This indigenous practice was promoted by Projet de DCveloppement Rural de Maradi (PDRM), the forestry services, and national legislation to protect certain types of trees. A similar development has been observed in the Zinder region, where natural, farmer-managed regeneration of trees is estimated to have occurred on about 1 million hectares of land in three of the region's departments (Larwanou, Abdoulaye, and Reij 2006). Gao (Faidherbiaalbida)trees are the dominant tree protected by farmers inZinder due to their benefits as a leguminous tree, their value as fodder and wood, and the fact that they lose their leaves during the rainy season and, hence, don't compete with crops for water and light during their growing period. Farmer protection of gao trees is not a new practice inZinder; for example, in the early 20th century, the Sultan of Zinder banned the cutting o f these trees (de Montgolfier-KouCviand L e HouCrou 1980). Nevertheless, the practice has reportedly increased in response to the droughts and famines of the 1970s and 1 9 8 0 ~population growth and the ~ resulting intensification of farming systems, changes inforestry policies allowing farmers secure tenure to the trees on their land, and development programs promoting tree planting and natural regeneration (Larwanou, Abdoulaye, and Reij 2006). For Niger as a whole, Adam et al. (2006) estimated that at least 3 million hectares of landhave beenrehabilitatedthrough farmer-managed natural regeneration. 12. This evidence suggests that developmentprograms, combinedwith national policy changes and local responses to adverse conditions, have contributed to re-greening and improvedland managementin Niger, although the specific impacts of these programsare notyet clear. Inthe next section, development programs promotingSLM inNiger are reviewed, along with the associatedlandmanagementpractices utilized. 20 111.SustainableLandManagementProgramsandPracticesinNiger 13. Since the early 1980s, more than 50 programs with SLM components have been executed or are ongoinginNiger. Usinga sample of 31programs,at least207 billionFCFA are estimatedto have been invested,in nominalterms (Annex 1). Most of these investments occurred inthe regions o f Tillabery, Tahoua, DOSSO, Maradi, and Zinder. While some programs were region-specific, others have beenimplementednationally or inseveral regions, such as the previously mentioned PAC, PGRN, and LUCOP, along with Projet d'AmCnagement des For& Naturelles (Natural Forest Management Project; PAFN), Projet de Promotion de 1'Irrigation PrivCe Phase 2 (Private Irrigation Promotion Project Phase 2; PIP2), and others. The largest completed programs in terms of funding include PGRN (25.6 billion FCFA), PIK (at least $75 millionQ, Projet d'Appui au DCveloppement Rural de Dosso (Rural Development Support Project in Dosso; PADER) (9.7 billion FCFA), and PAFN (8.5 billion FCFA). The largest ongoing programs for which we have cost estimates include PIP2 (30.2 billion FCFA), PAC (24.9 billion FCFA), and Appui pour la SCcuritC Alimentaire par la Petite Irrigation(Support for Food Security by Small Scale Irrigation; ASAPI) (18.6 billion FCFA). One large, ongoing program excluded from Annex 1due to lack of current information i s the previously mentioned PSPR, which supports a broad range of poverty reduction activities throughout Niger. PSPR was initiated in 2001 with a budget of 8.1 billion FCFA, received under the Highly IndebtedPoor Countries (HIPC) initiative (RN2002). Hence, the figure of 207 billion FCFA cited above is an underestimate of the amount spent by programs, since the costs o f both PSPR and PIK are not included. 14. Most of the investmentsin SLM programshave occurredsince the mid-1990s. Only seven of the programs listed in Annex 1 began before 1995; all o f the large multi-regional programs begun after 1995; several of the completed programs began after 2000; and, of the ongoing programs, all but one beganafter 2000. 15. Many of the programs are multipurpose development projects, for which natural resource management activities (NRM) are only one component. For example, PAC and PSPR are large poverty reduction programs investing ina wide range of productive activities and socioeconomic infrastructure, as well as inNRMactivities. Some, such as PADER and Projet de Developpement Rural de Tahoua (Rural Development Project in Tahoua; PDRT) focused on general rural development, while others focus on particular economic activities, such as promoting small-scale irrigation (for example, ASAPI and PIP2), but include NRMcomponents. However, some projects, such as PAFN and PGRN, have focused more specifically on NRM. Considering the emphases of the different programs, and based on consultations with program officials, roughly one-third of the total program expenditures included inAnnex 1targetedNRM activities. 16. Most of these programs have promoted physical water harvesting and soil and water conservation (SWC) measures, tree planting, and other vegetative measures to We do not have an estimate of the total cost in FCFA of the PIK project. Reij and Steeds (2003) report that $75 million was spent on this project between 1984 and 1999 ($63 million plus food rationsvalued at $12 million), indicatingthat this was also a large project. 21 rehabilitate lands and provide useful products and services, and investments to improve water availability. The major water-harvesting and SWC measures promoted include improved planting pits (zui), used on private cropland; half-moons and stone bunds, used on private and communal lands; banquettes (embankments with trenches used for harvesting water and tree planting); small dikes; and water spreading dams used mainly on communal lands. These and other NRM measures discussed in this report are described in Annex 2.4 Vegetative measures used mainly on communal lands include promotion of tree nurseries and tree plantations, vegetative bands and windbreaks, areas protected for natural regeneration, and sand dune fixation. On private cultivated lands, as well as some communal lands, regeneration of naturally occurring trees has beenpromoted by programs and i s also a traditional practice in some areas. Construction of wells and micro-dams is also being supportedby various projects. 17. Based on estimates providedby the developmentprograms, program activities have supported rehabilitation of more than 250,000 hectares of land, including more than 60,000 hectares treated with physical SWC measures, 65 million trees plantedon an estimated 160,000 hectares, vegetative bands and dune fixation on 25,000 hectares, and 8,000 hectares o f community areas protected for natural regeneration (Table 111.1). This is consistent with the estimate by Adam et al. (2006) of the area o f land rehabilitated in regions of Niger included in their study, not counting their larger estimated area of farmer-managed natural regeneration of trees on cultivated lands (more than 3 million hectares). For more detaileddescriptions ofthe NRMpracticespromotedinNiger, see MDAPAC (undated). 22 TableIII.l: Outputsof SelectedDevelopmentProjectsinNiger NRM Type of NRMactivity Quantity Physical measures Soil and water conservationand water-harvesting measures, such as zai, half-moons, and stone bunds (hectares) >47,000 Banquettes (hectares) 17,000 Water-spreadingdams 840 Vegetative measures Small nurseries 53 Trees planted 65 million Tree plantations(hectares) 162,000 Vegetative bands (hectares) 23,000 Wind breaks or live fences (kilometers) 410 Dunefixation(hectares) 2,000 Protectedor reservedareas Community protectedareas (hectares) 8,000 Livestock corridors (kilometers) 704 Water sources Wells 3,000 Micro-dams 51 Source: Reportsfrom programofficials. Note: See Annex 2 of this report for descriptionsofthe various NRMactivities. Data on tree plantationsare estimated basedon the numberoftrees planted, assumingthat, on average, 400 trees are plantedper hectare. 18. The reported outcomes of these investments include increased vegetation, reduced erosion, rehabilitation of degraded land, increased agricultural yields, more fodder for livestock, improved water availability, improved food security, improved welfare of vulnerable groups such as women and landlesspeople, and reducedpoverty, among others. Evidence from case studies supports these claims invarious contexts. For example, studies of an SLMproject supportedby IFADinthe I1161aDistrict ofthe Tahouaregionfoundthat, inaddition to earning a 20 percent rate o f economic return, the project and the SWC practices it promoted (mainly zai) led to increased yields, the rehabilitation of 9,000 hectares of degraded land, the promotion of the development of the land market for barren and crusted land, increased local labor demand, and diffbsion of zai to other areas (Box 111.1). Reij and Steeds (2003) report that between 1984 and 1999, PIK restored 20,000 hectares of strongly degraded land; planted 17 million trees (by 1991); fixed 1,300 hectareso f sand dunes; stabilized 1,400 kilometers of stream banks; built 40 micro-dams, two large dams, and hundreds of small water spreading dams; trained more than 100,000 people in SWC techniques; and constructed large numbers of wells, schools, health clinics, and roads. The increase inincomes generatedby the project was reported to be more than $6 million per year compared with a total project cost o f $75 million fiom 1984 to 1999, suggesting a lower rate o f return than that estimated for the IFAD project in IllCla.' Another study estimated a 30 percent reduction in water and wind erosion inthe Tahoua region as a result of the intervention of various projects, and that use of SWC structures have reduced The basis for these estimates was not provided by Reij and Steeds (2003), who argue that a more detailed impact assessment and cost-benefitanalysis ofthis project is needed. 23 run-off and increased water infiltration (Arnbouta, Moussa, and Daouda 2000). Adam et al. (2006) report improvements in tree density and availability, increased diversity of vegetation, increased production of timber and nontimber tree products, reduced water and wind erosion, increased infiltration and availability of water, improved soil fertility, increased availability of cultivable land, increased food production and availability, increased production o f cash crops, increased stock o f small ruminants, increased access to livestock fodder sources, reduced conflicts between farmers and herders, development of social organizations, increased capacity of local people to manage natural resources, improved living and economic conditions for women, reduced emigration, and higher incomes and reduced poverty in most of the project villages studied inthe Maradi, Tahoua, and Tillaberi regions, compared with little improvement inthe non-interventionvillages studied. 19. As noted earlier, however, these studies are beset by various methodological limitations, some of which are addressed in this assessment. The next section briefly discussesthe methodology ofthe current study, including its advantagesand limitations. 24 Box III.1: Project-led Land Rehabilitation inthe Ill6la District of Niger Between 1988 and 1997, an IFAD-funded SWC project operated in the I1161a district of the Tahoua region. The project promoted water-harvesting and SWC measures-including zui, half-moons, and stone bunds-and measured the impacts on yields over several years (Hassane, Martin, and Reij 2000). Millet yields on farmer-managed plots treated with zui and manure averaged 513 kilograms per hectare, and yields with half-moons and manure averaged 535 kilograms per hectare during 1991-6; yields on plots without these practices averaged only about 120 kilograms per hectare, and district yields averaged 301 kilograms per hectare (Hassane, Martin, and Reij 2000). Similar average yields for the period 1992-6 were obtained on a researcher-managed experimental plot with half-moons and manure, while higher yields o f 832 kilograms per hectare were obtained on the experimental plot with zui and manure. An experimental plot also investigated the impacts of these practices without manure and found substantially lower yields (546 kilograms per hectare for zui only and 285 kilograms per hectare for half-moons only), although still much higher than yields without the practices (26 kilograms per hectare). The yield advantages o f these SWC and manure practices were greater inyears with better rainfall, but even in low rainfall years the treated plots obtained at least 290 kilograms per hectare, on average, compared with less than 100 kilograms per hectare for the untreated plots. Hassane, Martin, and Reij (2000) argue that the yield advantage o f these treatments is greater than indicated by the comparisons with untreated plots because in most cases, prior to treatment, the land was unusedbarren and crusted land that would otherwise have yielded nothing. As a result o f these practices, a market developed for these barren lands, and 40 percent of surveyed farmers reported purchasing such land. Some farmers sold cultivated plots with sandy soils to buy and rehabilitate barren plots, which have higher production potential once rehabilitated due to the higher clay content o f the soil (Hassane, Martin, and Reij 2000). Limitations in the availability of such land were reported as an important factor hindering the spread o f these practices, along with labor and financial constraints and lack o f means to transport manure. Other reported impacts included increased demand for labor (including women's labor), promotion of local labor groups, and reduced emigration of labor. An impact evaluation conducted in 1998 estimated that about 9,000 hectares of degraded land had been treated by the project, representing 15 percent of cultivated area (Reij and Steeds 2003). The most widely adopted practice, consistent with the yield advantages reported above and lower investment costs, was zui, which accounted for most of the treated area. Spontaneous diffusion o f zui was found in other parts o f the IllCla district and elsewhere in Niger. Most adoption occurred on private cropland. Only 540 hectares were treated collectively. The promotion of these practices on communal lands was abandonedby the program due to landtenure problems. The on-farm incremental economic benefits were estimated to be $65 per hectare per year after the first year compared with the project investmentcost of $250 per hectare (Reij and Steeds 2003). The estimated rate o f economic return on the project was 20 percent (Hassane, Martin, and Reij 2000). 25 IV. Methods 4.1 Data Sources and Sample Selection 20. This impact assessment is based on analysis of secondary data collected for every village in Niger, together with and a community- and household-levelsurvey conductedin 139 villages with 1,218 households in the DOSSO, Maradi, Tahoua, Tillaberi, and Zinder regions. The analysis sought to assess the impacts of SLM programs on the adoption of land management practices, crop yields, and household income using sampling, matching, and econometric methods to control for potential biases that could result from comparing outcomes from noncomparable observations. 21, The first step of the study involvedassembling a databasefor all of Niger's villages, which was used to assess the coverage of SLM programs, and as a sample frame for the surveys. This database includes information on the geographical location, agroclimatic zone, dominant soil type, population, distance to urban centers (Niamey and the nearest town), access to a village market, infrastructure and services, and presence of major past or current SLM programs in the village. With the assistance of officials from many development programs operating inNiger, informationwas collected on the intervention villages o f 12 major programs, including the largest ones listed in Annex 1 (with the exception of PIP2, PSPR, and a few others). 22. The presenceof major programspromotingNRMactivitieshas been greatest inthe Tahoua and TillabCri regions, and least in the Agadez and Diffa regions. Based on the secondary data, 926 villages (less than 8 percent of all the villages inNiger) participated in the 12 major`programs. More than two-thirds of participating villages were located in the Tahoua and Tillaberi regions, while only 22 of these villages were located in the in Diffa and Agadez regions (Figure IV.1). 23. The Diffa and Agadez regions were excluded from the sampling frame for the surveys because few villagesinthese regionsparticipatedin any of the SLMprograms.This and other basic decisions inthe design of this assessment were basedon consultations with PAC officials, the Ministry of Agricultural Development, and other stakeholders in Niger. The agroclimatic conditions inthe sample villages are similar to those found inDiffa, however, so in that respect the sample is representative o f that region, although Diffa i s more remote from Niamey than the other regions and hence is different in terms of market access. The sample is less representative of Agadez, which i s in a more arid zone than the other regions o f Niger, so results do not reflect the situation or expectedimpacts of SLMprograms inthat region. The completedprojectsincorporatedinthe database includeProjetAgro-Sylvo-Pastoral(Agro-Forest-PastoralProject; PASP), PIK, PGRN, PAFN, and PDRT. The ongoing projects include ASAPI, Projet d'Appui pour le DkveloppementLocal (PADL), LUCOP, PAC, Programme de LutteContre 1'Ensablementdu fleuve Niger (Program to Fight Sedimentation of the Niger River; PLCE), andPSPR. 26 FigureIV.l: LocationsofVillagesParticipatingin SLMProgramsinNiger Source: International Crops Research Institute for the Semi-Arid Tropics (ICRISAT) village database, including intervention villages from 12major NRMprogramsinNiger. 24. To assess the factors affecting access to programs, descriptive statistical analysis of the secondary database was used to identify the characteristics of program versus nonprogramvillages. This analysis revealedthe extent to which program placement was biased toward areas of better agroecological conditions or market access, or toward areas where soil conditions are worse (these issues are discussedinthe next section). 25. Usingthe village database, a stratified random sample of 139 villages- comprising 70 program intervention villages and 69 comparable nonprogram villages-was selected for the surveys. Initially, 72 program villages were selected using random sampling o f the program villages stratified by program and region. The programs included in the stratification were LUCOP, PAC, PASP, PGRN, PIK, PSPR, and villages with multiple programs. PASP, PGRN, PIK, and PSPR have operated in the largest number o f villages, whereas LUCOP and PAC were selectedbecausethey were representativeof participatory programs and because PAC sponsored this study and LUCOP was closely affiliated with it. For each selected program village, a matching nonprogram village was identifiedfrom the same administrative region and agroclimatic zone. Matching villages were found using propensity score matching (PSM) (Rosenbaum and Rubin 1983) based on distance to the nearest town. There were 71 matching nonprogram villages (one of which was the best match for two different program villages), resulting in a total target sample o f 143 villages (Figure IV.2). Surveys were not conducted in four of these villages, due to problems of inaccessibility and misclassification of program status, resulting in 139villages inthe sample. 27 Figure IV.2: Locations of Selected Program and NonprogramVillages I Koa-yrapram Program 0 = 400 IGlomebrr S Source: ICRISATvillage database, basedon selectedvillage sample. 26. Community- and household-level surveys were conducted in each sample village early in 2007. At the village level, a group o f usually 10 to 15 respondents was assembled, representing village leaders and people who differed by gender, age, occupation, and social status. The village questionnaire asked about access to markets, infrastructure and services, programs and organizations operating in the village, local prices of inputs and outputs, land tenure systems, and management of village common lands, including the costs of and returns to any investments made in the common lands, institutional arrangements for managing such resources, and perceptions of changes in resource and welfare conditions and the reasons for change. The household survey collected information about household demographic characteristics; endowments of physical, human, natural, financial, and social capital; plot tenure and quality characteristics; plot-level land management decisions, investments, inputs and outputs; sources of information about landmanagementtechnologies; household participationin community land management activities; and perceptions of changes in welfare and resource quality indicators andthe reasons for change. 27. Within each programvillage, six program participantsand six nonparticipantswere randomly selected for inclusion in the householdsurvey, while in nonprogramvillages, six households were randomly selected. These three separate groups of households were selected becausenonparticipants inprogram villages are more likely to be affected by the presence of the 28 program-especially in terms of community-level infrastructure and activities-than are households innonprogram villages.' 28. Comparisons between program participants and comparable nonparticipants within program villages reflect the direct impacts of program participation, while comparisons between nonparticipants incomparable program and nonprogram villages reflect the spillover impacts on nonparticipants inprogram villages. Comparisons between program participants and comparable households in comparable nonprogram villages reflect the total impact of the program, considering bothdirect and indirect spillover effects. 4.2 DataAnalysis 29. The comparability of households in these different groups was addressed by using matching and econometric estimators to estimate the impacts of the programs on household-level land management practices, crop production, and income. Matching estimators were used to select households and plots from different groups that were as similar as possible in terms of their relevant characteristics-such as their endowments of assets and education, access to markets and infrastructure, and plot tenure and quality characteristics-in the year 2000 (before most of the projects being assessed had begun). Two different matching estimators were used: PSM (as previously discussed) and a nearest-neighbor matching estimator corrected for bias due to imperfect matching (NN) developed by Abadie et al. (2004). Each of these estimators has advantages and disadvantages, so both were usedto check the robustness of the conclusions. Ordinary least squares (OLS) and instrumental variables (IV) regression estimators were also used to investigate impacts o f SLM programs and practices, to test for biases, andto investigate the robustnessof conclusions using these models.' 30. The economic costs and benefits of land management practices at the household level were estimated using the results of the econometric analysis, and at the community level based on informationcollectedin the communitysurvey and through secondary data. Data collected on labor, materials, other inputs and outputs, and the prices of inputs and outputs were used, as well as estimates ofthe value of nonmarketed outputs, such as the change invalue of tree stocks and the contribution o f fodder materials browsed from communal areas to the value o f animals and animal products produced (details are providedinReport I1ofthis study). 31. In addition to estimatingthe economic impacts of SLM programinvestmentsusing quantitativeindicators,the impactsof these investmentswere assessed usinga wide variety of indicators of social and environmentalchange reported by households and community membersin the surveys. Indicators investigated included perceived changes inthe availability of timber, fuel wood, fodder, pasture areas, cultivable land, and beneficial wild animals and This approach is similar to that used by Feder, Murgai, and Quizon (2004) in their evaluation of the impacts of farmer field *schoolsestimation in Indonesia. IV can be used to test for and address problems of selection bias due to unobservable differences, if suitable instrumentalvariables can be identified(Heckman et al. 1998). OLS does not correct for such problems, but is more efficient than IV estimationifsuch problems are tested for and rejected. Inthe IV models adoptedfor this study, statisticaltests supported the use ofthe OLS model, so IV modelresultsare not reported.Details ofthe estimationmethods, statisticaltests, and resultsare discussedinReportI1ofthis study. 29 plants; the presence of pest and predators; the availability and quality of water; conflicts over natural resources; soil erosion; flooding during the rainy season; cropland quality; crop and livestock productivity; food security; revenues; and the general welfare o f the people. These qualitative indicators (based on perceptions) are subject to the same potential biases as those reported in other studies, but some can be compared with the results for quantitative indicators (for example, crop productivity and income) so that the robustness of the conclusions can be assessed using different types of indicators. 4.3 Strengths and Limitations 32. To summarize the above, the strengths of the methods used in this analysis were in (i) assembling information on a large number of SLMprograms and their locations; (ii) that using information to select a large and statistically representative sample for the survey; (iii) using matching techniques to ensure the selected program and nonprogram villages, households and plots were as comparable as possible; (iv) having the ability to investigate the direct and spillover effects of programs by selecting both nonparticipants in program villages and households in nonprogram villages to compare against direct program participants; (v) controlling for possible selection biases allowing changes in outcomes to be attributed to the programs; (vi) estimating the economic costs and benefits of program interventions and land management practices; and (vii) combining qualitative with quantitative data to deepen the insights derived from the study. 33. The reviewwas also subject to severallimitations. Amongthese are the following: The absence of suitable baseline data. Ideally, baseline information should be collected prior to the program for both program and nonprogram villages and households for use in assessing changes in the different groups over time. Unfortunately, such baseline information was not available. As a result, recall information was usedfor major endowments and other key factors inthe base year of 2000. This i s not a perfect solution, given potential errors in recall, but this problem was addressed by focusing on major endowments and other slowly changing factors that could be readily recalled. The data collected for this analysis could serve as a baseline for future impact assessment studies of SLMprograms inNiger. Lack of remote sensing and other historical data on changes in land use and management. This analysis could be strengthenedthrough the use o f remote sensing data and other historical sources to trace land use and other changes in the villages studiedandrelate those to the survey findings. Budget and time constraints precluded the pursuit ofthis additional informationduring the course ofthe analysis. This could, however, be readily pursued in the future because the village locations were geo- referenced. In the context o f TerrAfiica, analytical studies on SLM practices best suited to address climate change adaptation and mitigation are underway. They will use remote sensing data and GIs, both at the regional and the country levels (including Niger). Data collected for the present study and the above mentioned studies will contribute to the development of the national SLM Information Systems for the Country Strategic Investment Framework (CSIF) on SLM (the CSIF will include geo-referenceddata ina geographic informationsystem). 30 Lack of objective indicators of social and environmental impacts. As with other studies, only subjective perceptions o f social and environmental impacts were investigated, and these could be subject to bias. Measurement and valuation o f some impacts-such as changes in vegetative cover, water availability and quality, carbon sequestration, or soil erosion--could be pursued in future studies (using remote sensing data in some cases), but these approaches were infeasible for this study, which focused on economic impacts, especially on crop production and household income. Focus onprivate costs and benejits. As agreed at the design stage, this work did not look at broader externalities inthe context o f a larger assessment o f public costs and benefits. This dimension would have required a landscape approach supported by a more integrated analysis o f the integration o f land and water management. This actually should be further explored in the near future to support o f the provision o f additional guidance for scaling-up, going beyond the plot level. Factors causing the success (or failure) of SLMprograms. The analysis inthis study mainly focuses on whether or not the S L M programs were successful, in terms o f various outcome measures, and less on why programs succeeded or failed in particular circumstances. Some indications o f the determinants o f success are suggested by the analysis (for example, land management practices that are not profitable inparticular contexts are unlikely to be adopted or sustained). However, a more in-depth analysis o f why practices and approaches that are likely to be profitable are not more widely adopted has not yet been done (for example, assessing the role o f land tenure institutions in affecting success). Further research drawing on the data collected for this study, combined with more in-depth case-study work in selected sites could help to address these issues. Other knowledge products prepared in the context of TerrAfrica, such as development of a local monitoring and evaluation system relatedto SLMactivities and a public expenditure reviewrelated to S L M are also helping to address this knowledge need. 31 V. Findings 5.1 Coverageof SLMPrograms 34. Despite the large investmentsin SLM programs, their coverage of communities in Niger has beenlimitedand has focused in areas of more favorable agro-climatic conditions and market access. As noted above, 12 o f the largest programs have operated in fewer than 8 percent o f the villages o f Niger. Program villages are more common inthe south o f the country (Figure 1V.l) and are therefore more likely to be in a relatively high rainfall zone than non- program villages (Table V.1). Program villages tend to be closer to Niamey than non-program villages and are also somewhat closer to the nearest town. Program investments are also more common in areas with degraded soils (Table V.1). Thus, it appears that SLM programs more often operate in locations with greater land management problems, but where the potential impacts of interventions are likely to be more favorable. Table V.1: Characteristics of Major SLMProgram and Non-program Villages in Niger Characteristic No program(YO) Any program (YO) Lengthof growingperiod 125 to 150days 0.3 0.3 100to 125 days 8.3 16.4 75 to 100days 50.2 41.3 50 to 75 days 37.9 40.6 Less than 50 days 3.3 1.3 Soiltype Clay 17.9 14.4 Sandylloamy 1.4 1.7 Sandy 69.2 65.6 Degraded 8.8 14.5 Alkaline 2.8 3.9 Distanceto Niamey(kilometers) 474.0 295.0 Distanceto nearest town (kilometers) 40.2 34.3 Source: ICRISATvillage database, includinginterventionvillages from 12major NRh4programsinNiger. 35. Program operations are also more common in areas with lower poverty indicators. Programs are more common in the southern belt of the country, which is also where poverty indicators tend to be lower. Within the community survey sample, program villages tend to have better access to infkastructure and services and are more likely to use agricultural inputs than non- program villages, even though the non-program villages were selectedto be a similar distance fiom the nearest town (Table V.2). It i s not clear fiom these results, however, whether lower poverty and better access i s a result of program involvement, or whether programs have tended to focus on better endowed communities. 32 Table V.2: Access to Markets, Infrastructure,Services and Inputs in Program and Non-program Villages inthe CommunitySurvey Sample Characteristic No program Any program Distanceto nearest town (kilometers) 27.1 30.7 Access to infiastructure and services Share of villages (%) Pavedroad 22.5 47.1 Laterite road 40.8 60.3 Input supply shop 9.9 22.1 Irrigation 32.4 52.9 Cerealmill 33.8 52.9 Cerealbank 9.9 55.9 Savingskredit cooperative 22.5 35.3 Use of inputs (ever used) Inorganic fertilizer 50.7 64.7 Improvedseeds 18.3 52.9 Insecticides 38.0 48.5 Animal vaccines 60.6 63.2 Improved forages 7.0 11.8 Tree seedlings 8.5 42.6 Source:InternationalFood Policy ResearchInstitute (IFPRI) and ICRISAT community survey 36. Results from the household survey suggest that for some programs, households in program villages were already wealthier than nonparticipants in program villages or households in comparable non-program villages prior to the program. For example, households participating in PAC owned more irrigated land, livestock, farm equipment, and household assets than either nonparticipants in program villages or households in non-program villages in 2000/1, prior to the PAC program (Table V.3)' Econometric analysis of the factors associated with participation in PAC confirms that PAC participants tended to be somewhat better off prior to the program, having more farm equipment and higher education, and being more likely to be a president of a farmers' association in 2000/1.'0 Similarly, participants in PSPR owned more irrigated land and livestock than nonparticipants in 2000/1, prior to the commencement of that program.l1These programs (especially PAC) appear to have involved wealthier households to some extent (although the participating households were still quite poor on average). The fact that participants in these programs also had more wealth in 2006/7 may therefore have been due to their greater initial wealth and not to the effects of the programs. Also, participants inother government programs (aside from PGRN) and inNGOprograms were wealthier on average in many respects than nonparticipants in 2000/1 (as well as at present).12 Nevertheless, without taking into account when each of these programs began, it is not possible The analysis investigatedthe impacts of participation in PAC, PGRN, and PSPR because these large programs were the most adequately representedin the sample. Other programs were aggregatedinto "other governmentprograms" and"NGO programs" becauseof limited numbersofobservationsof each specific program. loSee ReportI1for resultsof this econometric analysis. The econometric analysis of factors associated with participation in PSPR, however, did not find statistically significant associations between these wealth indicators and program participation, once other factors were controlled for. The analysis found that households with less farm equipment in 2000/1 were more likely to participate in PSPR (see Report I1 for further details). l2The econometric analysis found that households with more householdassets in 2000/1 were more likely to be participantsin other governmentprograms; thoughother wealth indicatorsdid nothaveastatistically significanteffect. 33 to determine whether the greater wealth of participants in 2000/1 was due to program effects or to a bias toward wealthier households on the part of the programs. In contrast, participants in PGRN had fewer household assets in 2000/1 (and currently) than nonparticipants, which may reflect pro-poor targeting o f PGRN, but once again this is uncertain because by 2000 that program had already been operating for several years.l3 Table V.3: MeanCharacteristicsofProgramandNon-programHouseholds Non- Other House- program govern- holds house- ment NGO innon- holdsin PAC PSPR PGRN program program program program house- house- house- house- house- Variable villages villages holds holds holds holds holds Number ofobservations 407 312 103 143 92 123 178 Landowned(hectares) 200011 9.84 8.47 10.70 10.70 8.91 9.69 13.98*** 200617 9.54 8.83 11.36* 11.06 9.60 9.85 14.75 *** Irrigatedland(hectares) 200011 0.13 0.22 0.88*** 0.69* 0.16 0.61 1.00*** 200617 0.11** 0.25 0.90*** 0.76** 0.29 0.69* 1.05*** Value of livestock(thousandFCFA) 2000101 105.3 98.1 142.3*** 121.0* 117.4 135.2*** 138.4*** 2006107 164.6** 137.6 217.0*** 168.1** 152.7 191.4*** 201.7*** Value of farm equipment(thousandFCFA) 200011 32.7 34.8 56.6*** 32.3 33.2 49.1 * 66.5*** 200617 57.2 56.2 70.8 54.9 43.9 72.0* 87.7*** Value ofhouseholdassets-transportation equipment, durable goods, buildings (thousandFCFA) 200011 156.3 168.3 238.0* 203.4 64.6*** 278.3*** 323.0*** 200617 216.1 235.5 314.3 288.1 99.2*** 340.8** * 419.0** * Source: IFPRIandICRISAThouseholdsurvey *, **, *** indicate that differences from non-programhouseholdsinprogramvillages are statistically significant at the 10, 5, and 1percentlevels, respectively. 5.2 CommunityLandManagementInvestments 37. The three most common types of community land management investments include tree plantations, areas protected for natural regeneration, and areas where communal SWC and other investmentswere made (without tree planting or protection). 5.2.1 Treeplantations 38. Tree plantations are the most common community land management investment promoted by SLM programs. Tree plantations were found in 90 percent of program villages l3The associations of wealth indicators with participation in NGO programs and PGRN were statistically insignificant in the econometricanalysis. 34 but inonly three of the sample non-program villages (Table V.4). Program involvement appears to be essential for communities to invest in tree plantations, probably because o f the technical assistance required to properly manage the trees. In addition, the programs provide financial support for the cost of seedlings and equipment. Some programs also compensateparticipants for their labor with cashor "food for work," while others, suchas PAC andLUCOP, do not. Table V.4: Mean Values of CommunityLand ManagementInvestments Tree Protectedareas Other planta- Nan- communal Variable * tions All Program program investments Number ofvillages with investmenta 53 14 9 5 19 Number ofinvestments 103 16 9 7 37 Area (hectares) 46.8 29.9 28.0 32.4 14.9 Trees planted(number) 13,714 n.a. n.a. n.a. n.a. Trees surviving(%) 53.7 n.a. n.a. n.a. n.a. Plantationswith acacia Senegal(%) 78.6 n.a. n.a. n.a. n.a. Stock oftrees (number) 9,467 n.a. n.a. n.a. n.a. Value oftree stock (thousandFCFA) 17,200 n.a. n.a. n.a. n.a. Area with half-moons(%) 34.0 31.2 44.4 14.3 27.0 Area with banquettes(%) 7.8 6.2 11.1 0.0 0.0 Area with other SWC (%) 13.6 6.2 11.1 0.0 8.1 Value oflabor investment(thousandFCFA) 1,255.8 53.1 70.4 30.9 149.5 Value ofother investmentcosts (thousandFCFA) 141.7 1.3 2.2 0.0 52.0 Value oftrees cut in2006 (thousandFCFA) 4.2 5.0 8.9 0.0 0.0 Value ofproductscollectedin2006 (thousandFCFA) 1.9 0.03 0.06 0.0 4.1 Value of fodder for goats browsingarea in2006b 655.3 516.4 874.6 55.9 0.0 Value of fodder for camelsbrowsingareain2006 2,195.4 171.4 304.6 0.0 0.0 Totalvalue of benefitsin2006 (thousandFCFA) 2,856.9 692.8 1,188.2 55.9 4.1 Total investmentcost (thousandFCFA) 1,397.5 54.4 72.6 30.9 201.5 Percentage ofareas with positivebenefitsin2006 19.4 50.0 66.7 28.6 5.4 Percentage ofareas with benefitsin2006 greater than investmentcost 11.7 43.8 66.7 14.3 0.0 Percentage of areas with value of stock in2006 seater thaninvestmentcost 92.2 n.a. n.a. n.a. n.a. Source:Calculatedby authorsbasedon IFPRVICRISAT community survey Note: n.a. indicatesnot applicable. a. Information was collected on 125 plantations in 62 villages. However, the number of trees planted was not providedby respondentsfor 22 ofthe plantations. The table thereforeonly includes 103plantations. b. Browsingby goats and camels was estimatedinnumbers of goat-years or camel-years using the average number of animals browsing, multipliedby the number of days browsingper year andby the number of hours browsingper day, dividedby the number of hours per year (24 x 365). The value of a goat-yearwas estimatedto be 9,500 FCFA inregionswheregoatmilkis consumed(theDiffa,Maradi, Tahoua, andZinder) and3,900 FCFAinregionswhere goat milk is not consumed(Dosso andTillaberi). The value of a camel-yearwas estimatedto be 94,000 FCFA. The basis for these estimatesis providedinReport11. 35 39. Average reportedtree survival rates are fairly low in plantations(only 54 percent), though substantial variation exists across plantations. Survival rates are significantly higher for certain types o f trees (that is, Azadirachta indica (neem) andAcacia seyal) and inplantations promoted by PAC compared with other programs. SWC measures, especially half-moons, are commonly used intree plantations to help collect water and increase tree survival. Surprisingly, however, no significant impact of SWC measureswas found on reported survival rates, basedon a regression analysis of determinants of survival rates.14 Further investigation o f these results i s warranted, but these preliminaryresults suggest that certain types of trees and approaches should bepromoted, while improvements may beneededinthe targeting of SWC measures. 40. Acacia Senegal is the most commonly planted tree in plantations, yet little gum Arabic is beingharvested.Acacia Senegalwas plantedinnearly 80 percent oftheplantations in the sample and acacia seyal in 12 percent, yet gum Arabic was collected from only two plantations in 2006. This i s partly due to the recent nature of most acacia Senegal plantations, given that nearly two-thirds of the sample plantations were plantedafter 2002. Yet even among mature plantations, very little harvesting of gum Arabic was reported. This indicates that the economic potential of these plantations is not being realized. A 2003 study o f gum Arabic in Niger found numerous constraints affecting this commodity, including lack of technical knowledge on managing the trees or harvesting the gum properly, conflicts among farmers concerning access rights to the trees and gum, poor practices in handling and storing the gum leading to poor quality, lack of organization of the marketing sector, high taxes and risks of confiscation that underminemarketing, poor transport and communication infrastructure, lack of experience with the international market, and absence of finance for marketing (Minist2re de l'Hydraulique, de 1'Environnementet de la Lutte Contre la Ddsertification 2003). Although some efforts are being made to address these constraints, most of them apparently continue to limit the production and marketing of gum Arabic (UNFCCC 2008). Further investigation of the reasons for limitedexploitation of gum Arabic and appropriate remedies is needed. 41. The main benefits of community tree plantations are the value of the wood and fodder produced. Based on the estimated numbers of trees planted, tree survival rates, and the market values of wood inthe survey communities, the averageestimatedvalue of wood ina six- year-old plantation (the mean age in the sample) is about 17.2 million FCFA per plantation, or about 368,000 FCFA per hectare, considering the average plantation size o f about 47 hectares (Table V.4). This is equivalent to about 7 percent of the total value o f household assets inthese villages on average, representing a substantial contribution to the wealth o f households in villages with plantations. The average value of fodder in tree plantations, based on the type, number, and amount of time animals spent browsing in the plantations, the estimated share of annual fodder provided by browsing, and the estimated value resulting from growth of and products from those animals, was about 2.8 million FCFA per plantation (or 61,000 FCFAha) in 2006 (Table V.4). While this estimate is conservative-it assumed that net economic benefits were only derived through browsing by goats and camels, since goats and camels browse on tree leaves and animals like cattle and sheep mainly graze on grasses that would have been available without the plantation-it still predicts that tree plantations have quite large benefits for livestock. Using an alternative method to estimate the value of fodder production4onsidering the types and densities of trees planted, the annual production of browsing material per tree, and l4These findings are basedon regressionresults for tree survivalrates, providedinReport11. 36 the value of fodder reported inthe literature-the meanvalue offodder productionwas estimated to be about 31,000 FCFAha plantedafter five years, or about half of the fodder value estimated based on livestock values. Other estimated economic benefits, such as the value of trees cut and the value of gum Arabic, fruits, or other products collected from the plantations yielded smaller orders of magnitude on average (about 6,000 FCFA per plantation in2006). 42. The main cost of establishingtree plantations is the labor of planting seedlings and constructing water-harvesting and SWC measures. Valuing labor at the survey's median daily wage level, estimated labor investment costs averaged about 1.3 million FCFA per plantation. Additional costs were much less than this (averaging about 140,000 FCFA per plantation), but this does not include the costs of seedlings, materials, and equipment provided by SLMprograms. 43. Plantations are highly profitable to the communities on average, although large variation in returns exists across plantations. As noted above, the average annual benefit in 2006 was greater than the investment cost; however, very large variation exists around these averages. There were positive estimated benefits in 2006 for only about one-fifth of the sample plantations, and the benefits exceededthe investmentcosts for only 12percent of the plantations. Much of this is due to the recently established plantations, given that nearly two-thirds of the plantations with zero benefits in2006 were plantedafter 2002. Still, many older plantations also yieldedzero current benefits. The main sources of variation inthe profitability oftree plantations are variations in tree survival 'rates (ranging from close to zero to nearly 100 %, with a standarddeviation of 31%), labor investmentcosts (coefficient ofvariation equal to 4), and in the value o f fodder browsed by livestock fiom the plantation (coefficient o f variation greater than 5) -Details are available inReport. 44. The high variability of returns to plantations implies both a need and an opportunity to increase the economic returns to many plantations. Inaddition to promoting the harvestingof gum Arabic, as noted above, development ofthe livestock sector (for example, through investments in dairy goats or goat fattening) could increase the return to investmentsin plantations that provide browsing for such animals. Promotion o f planting and harvesting of other valuable tree and forest products (for example, fruits, honey) could also helpto increasethe benefits of plantations. Improved targeting of tree types used in plantations and management practices that increase survival rates (e.g., watering seedlings) could increase returns where survival rates are low. Improved targeting of labor intensive SWC investments might also increase the effectiveness o f such investments in promoting tree survival where they are used, andavoid unnecessarycosts wherethese are not effective. 45. Despite low current returns for many plantations,nearly all of them have a greater stock of wood value than the initial investment costs, suggesting that even for plantations with little current benefits, the investment was worthwhile in terms of the resulting contribution to community wealth. The average value o f wood stocks in plantations is much larger than the average community investment costs, and, for most plantations, this stock value will continue to grow untilthe trees are mature. 37 46. Considering the full costs of investments in tree plantations (including costs of seedlings, equipment and other items supported by programs), plantations represent an attractive investment option for SLM programs. Using the average economic benefits reported in Table V.4 and the full costs o f plantations from the literature, the internal rate o f return (IRR) for plantations is estimated to be 45 percent (Figure V.l), and the net present value (NPV) is estimated to be 307,000 FCFA per hectare, assuming a discount rate o f 10 percent (Table V.5). Using the more conservative estimate o f fodder value mentioned above, the IRR is reduced to 28 percent andthe NPV to 125,000 FCFA per hectare. These estimates do not include the potential value o f gum Arabic, the harvesting o f which could increase the IRR by several percentage points (to 33 percent inthe conservative fodder value case and 48 percent inthe less conservative case), assuming that 0.65 kilograms o f gum i s harvested per tree beginning in the fifth year after planting, and that farmers' net return from gum production is 100 FCFA per kilogram.l5 Figure V.l: Estimated Mean Internal Rate of Return to Community Tree Plantations and ProtectedAreas Percent 303 251.6 253 203 353 103 53 i) Plantattonswithout Plantatianswitt' Plantationsa th Prorroted gum A%bic g ~ iArab c n g u nArab c and protected area: value of c LOW 'odder value Source: Estimated by authors based on IFPRVICRISATcommunity survey and ralues fiom literature IsThese estimatesare based on figures from UNFCCC (2008); they are more conservative thanthe estimatesusedby Abdoulaye and Ibro (2006), who assumed that each acacia Senegal tree produces 1.5 kilogramsof gum beginninginthe fifth year, and that gum Arabic is worth 1,000 FCFA per kilogram. Although the price of gumArabic is at leastthis high, farmers' net returns from producing and marketinggum Arabic are likely to be much less than the market price, based on costs and marketingmargins. The margins estimated inUNFCCC (2008) are reportedas potentialmargins, after improvementsin marketing, so they may still be optimistic. Despite estimatinglower benefitsfrom gum Arabic than Abdoulaye and Ibro (2006), the estimatedrates of return for tree plantationsreported here are still higher because they assume an unrealistically low level of annual fodder production from plantations(only 15kilogramsper hectare). 38 Table V.5: EstimatedAggregate Benefitsand CostsofPlantationsand ProtectedAreas Promotedtree Promoted Item plantations protected areas NPV per hectare (10 percent discount rate) with no gum Arabic or C sequestration value (thousand FCFA per hectare) Low fodder value 125 118 Highfodder value 307 239 NPV per tree planted (FCFA per tree) Low fodder value 427 n.a. Highfodder value 1,048 n.a. Number o ftrees planted(million trees) a 65.3 0 Estimated total area o f promoted protected areas (hectares) n.a. 8,008 Estimated aggregateNPV (million FCFA) Low fodder valueb 27,878 947 Highfodder value 68,378 1,911 Value o f tree stock Estimated survival rate o f plantedtrees (percent) 53.7 n.a. Estimated discounted value per tree, assuming mean age o f six 1,219 n.a. years (FCFA)` Estimated discounted stock value per hectare (thousand FCFA 207 144 per hectare) Discountedaggregate value o f surviving tree stock (million FCFA) 42,735 1,155 Sum o fNPV and discounted stock value (million FCFA) Low 70,613 2,101 High 111,113 3,065 Source:Calculatedby authors. Notes: n.a. indicates not applicable. a. The total number o f trees planted and area of promoted protected areas are based on the estimates provided in Table V.4. b. The estimated total NPV for tree plantations is based on the estimated NPV per tree planted, multiplied by the number o f trees planted, since a better estimate o f the number o ftrees planted than area o f plantations was available. For promoted protected areas, the estimated NPV per hectare was multipliedby the estimated total number o f hectares o f such areas. c. The estimated value per tree is based on a regression for the price o f trees as a function o fthe age o ftrees (1041 + 196.1 x age - 1.594 x age'), usingthe mean age o f plantations for our sample (six years). The value o f tree stocks is discounted to the present value using a discount rate o f 10 percent over six years. d. The aggregate value o ftree stock in plantations is estimated as the total number o ftrees planted multiplied by the proportion surviving multiplied by the discounted value per tree. The estimated aggregate value' o f tree stock in protected areas is based on an estimate o f the value per hectare o f tree stocks inpromoted protected areas multiplied by the estimated total area of promoted protected areas. The value per hectare o f tree stocks in protected areas is estimated as the value o f annual benefits per hectare from promoted protected areas (42,000 FCFA per hectare, based on estimates inTable V.4) dividedby the value o f annual benefits per hectare from plantations (61,000 FCFA per hectare), multiplied by the discounted value per hectare o f tree stocks in plantations (207,000 FCFA per hectare), which results ina discounted value o f 144,000 FCFA per hectare for protected areas. 39 47. The total estimated net present value of tree plantations promoted by the major SLM programs studied ranges from 71 to 111 billion FCFA, greater than the amount estimlated to have been spent on NRM activities by these programs. More than 65 million trees have been planted by the programs studied (Table V.5). These are estimated to generate an NPV from the flow o f annual benefits of 427 to 1,048 FCFA per tree planted, for a total NPV of 28 to 68 billion FCFA (Table V.5).16 Adding the estimated discounted value of the surviving trees (valued at about 1,200 FCFA per tree) results in a total net present value of 71 to 111 billion FCFA. This is comparable to or greater than the amount roughly estimated to have been spent by the programs studied on NRMactivities-about 69 billion FCFA, assuming that one- third oftotal program expendituresof207 billionFCFA were spent onNRMactivities. 48. The estimated value of benefits from tree plantationswould be even greater if the value of carbon sequestered or other environmentalbenefitswere included. For example, it has been estimated that a Clean Development Mechanism (CDM) project begun in 2006, which i s planting nearly 18,000 hectares of acacia Senegal inNiger at a density of 400 trees per hectare, will sequester 1milliontons of C02equivalent (tCO2e) by 2017 andmore than2.3 milliontC02e by 2035 (UNFCCC 2008). The CDM Bio Carbon Fundis paying $4.20 per tCO2e that will be sequestered within the first eight years of the project. This value is low compared with the average value of carbon credits for CDM projects, which averaged $10.90 per tCO2e for developing-country projects in 2006 (Ambrosi 2007). Using the lower value ($4.20 per tC02e sequestered) as a conservative estimate of the value of carbon sequestration, and estimates from UNFCCC (2008) of the amount of carbon sequestered by acacia Senegal trees each year, including the carbon sequestration benefit increases the average IRR for tree plantations (including the potential value of gum Arabic) from 33.2 to 34.7 percent (using conservative estimates of fodder value) or from 48.1 to 49.1 percent (assuming less conservative fodder value estimates). These estimates indicate that the value of carbon sequestration contributes modestly to increasing the social profitability of tree plantations, but it is not a major determinantof socialprofitability. 49. Many 'other environmental or social benefits of tree plantations are reported by most communities having plantations. These include improvements in the availability of fodder, fuel wood, and useful wild plants; reduced soil erosion; and improved food security, revenues, and general welfare o fthe people (Figure V.2). These positive assessments are broadly consistent with the economic benefit-cost results discussed above, although the perceived impacts are positive in many cases even when no current economic benefits were reported. In part, this reflects the broader set of indicators for which data on perceptions were collected, but it may also reflect a tendency for people to overstate favorable outcomes for programs operating in their communities. Positive outcomes were least commonly reported concerning impacts of plantations on availability and quality of water. In a few cases, negative impacts o f plantations were reported for some indicators, including impacts on availability o f drinking water, availability of pasture areas, presence of predators or pests, conflicts over natural resources, and food security. These responses indicate that trade-offs may be present in some cases, although positive or no-impact responseswere muchmore common thannegative ones. l6This is basedon anestimatedaverage of 293 trees plantedper hectareinthe plantationsstudied. 40 FigureV.2: ReportedQualitativeImpactsof CommunityTree Plantations Share of sampleplantations (%) --p-m Major deterioration Minor deterioration Insignificant impact Minor improvement Major improvement - - __ - - __. - __ -________-_____ _ _ - _I_ .- _I_^_ _xI_ Source: Calculatedby authors. 41 5.2.2 Areas protectedfor naturalregeneration 50. Some communities have protected communal areas for natural regeneration of vegetation,in most cases promotedby an S L M program.Sixteensuch areaswere found in 14 o f the survey communities (10 percent), 9 o f which were promoted by a program (Table V.4). Program involvement appears to be less essential in establishing protected areas than it is for plantations, probably because o f the lower investment cost and level o f technical expertise required. 51. The main economic benefit to communities from protected areas-the value of fodder material produced-is much greater in areas promotedby S L M programsthan in other areas. Although program involvement is not essential for establishing protected areas, it appears to help increase their benefits, probably by helping to overcome collective action problems inestablishing and protectingthese areas. 52. Although the benefits of protected areas are lower than those of plantations, the costs are also much lower, resulting in much higher IRR for protected areas, especially those promotedby programs. The estimated mean IRR is more than 170percent for promoted protected areas using the lower estimate o f fodder value (Figure V.1) and 46 percent for non- promoted ones. Nevertheless, the NPV per hectare i s lower for promoted protected areas than it i s for plantations (at a 10 percent discount rate) because o f the lower annual benefits, although at high discount rates protected areas have a higher NPV. Thus, protected areas are an attractive investment, especially for communities whose ability to finance a tree plantation is limited. Nevertheless, plantations are a more profitable use o f land if the costs can be financed and the technical issues addressed. 53. Overall, protectedareas promoted by the programs investigatedhave generated a total estimated NPV of about 2 to 3 billion FCFA, considering the value of fodder productionandwood generated.This estimate is based onthe estimated area ofprotected areas promoted by programs multiplied by the estimated NPV per hectare o f protected areas (Table V.5). Although this is much less than the estimated total benefits o f tree plantations, it is still a substantial benefit. 54. As for tree plantations,community membersreportmany environmentaland social benefitsof protectedareas.Most communities with protectedareas report increased availability o f fodder; beneficial wild plants and animals; and improved food security, revenues, and general welfare (Figure V.3). Consistent with the estimates o f the economic benefits, these perceptions are generally more favorable for protected areas promoted by programs than they are for other protected areas. 42 FigureV.3: ReportedQualitativeImpactsof CommunityProtectedAreas Percent of sample protected areas 100% 90% 80% 70% 60% 50% 40% 30% Major deterioration Minor deterioration E>Insignificant impact Minor improvement Major improvement Source: Calculatedby authors. 43 5.2.3 OthercommunitySWC investments 55. SWC investmentswere also made in other communal areas (without tree planting or protection)by about 14 percentof villages, almost always promotedby SLM programs. These investments yielded very low estimated economic returns. Only 5 percent o f the sample investment areas yielded positive benefits in2006, and the benefits were relatively small even when positive benefits were reported (Table V.4). On the other hand, the investment costs o f these areas were higher on average than those for protected areas because o f the labor and materials required for constructing SWC measures. The mean IRR could not be calculated for these areas because o f negative returns, while the NPV is negative. 56. Environmental and social benefits are reported by communities for these investmentsin many cases, though less commonly than for plantationsor protectedareas. For example, reduced soil erosionwas cited as a benefit by about halfthe communities with such investments, andreduced flooding duringthe rainy season, increased availability o f cropland and pasture, and improved food security were cited by about one-third o f communities. Thus, these investments appear to be providing some perceived benefits to community members, even if such benefits are difficult to quantify economically and are less commonly reported than for other community investments. 57. Given the relativelyhighcosts and low economic benefitsof establishing.theseSWC investments in unprotected areas-and the higher returns of alternative investments- SLM programs should reconsider whether and when to promote these types of investments. 5.3 HouseholdLandManagementPractices 58. Land managementpracticesused by farmers in Niger include (in descending order of use) organic fertilizer, inorganic fertilizer, zai, half-moons, stone bunds, tree planting, and mulching. Organic fertilizer (mainly manure) is applied to more than 40 percent o f plots and inorganic fertilizer to about 18 percent; none o f the other practices is used on more than 10 percent o f the nearly 3,000 householdplots inthe sample (Table V.6). When organic fertilizer i s applied, farmers apply about 300 kilograms per hectare on average, though very large variations in application rates occur across plots. Inorganic fertilizer is applied in small doses, averaging only about 20 kilograms per hectare when used. This is consistent with findings o f other studies showing that fertilizer "micro-dosing" is a common practice in Niger due to the high cost o f fertilizer and farmers' poverty and limited access to credit (Pender et al. 2008). Tree plantingby households on their own land is rare-trees have been planted on less than 4 percent o f the sample plots since 2001-and when trees were planted only about 12 were planted per hectare on average. 44 Table V.6: Plot-LevelLandManagement Decisions by ProgramType (mean values) Non- Other Households program govern- in non- households ment NGO Full program in program PAC PSPR PGRN program program Variable sample villages villages households households households households households Numberofplots 2,914 975 747 259 343 211 323 482 Percentof plotswithpracticein2006 Stone bunds 5.2 0.3*** 9.2 4.6*** 6.4' 0.5*** 13.3; 6.8 Half-moons 4.3 0.5*** 3.7 10.0*** 12.0*** 4.7 13.3*** 7.3'. Zai 9.5 2.8*** 10.6 14.7* 15.2** 9.0 26.3*** 15.6** Mulch 3.6 3.9 3.3 o.o*** 3.5 3.8 3.1 5.6' Organicfertilizer 42.7 42.1' 46.3 43.2 35.6*** 39.8' 47.4 42.1 Inorganicfertilizer 18.3 20.7*** 14.4 14.3 15.7 22.3** 19.8** 18.9** Any trees on plot 77.4 83.8*** 63.7 91.5*** 76.4*** 82.9*** 80.5*** 84.6*** Faidherbiaalbida 33.6 36.6** 31.7 32.4 14.9*** 46.4*** 31.9 30.1 Trees plantedsince 2001 3.5 4.0** 2.0 2.3 3.2 6.6*** 2.5 2.3 Numberoftrees plantedper hectare,ifany, since 2001 11.8 8.5 10.3 6.8 19.0* 14.1 8.1 9.6 ' Quantityof fertilizer used,ifany (kilogramsper hectare) Organicfertilizer 299.8 254.6** 376.4 297.4 295.1 124.5*** 439.6 280.0 Inorganicfertilizer 20.3 17.9* 12.0 19.2 47.4*** 10.9 19.7 27.6** Source: Estimatedby authorsbasedon IFPRVICRISAT householdsurvey. *, **, *** indicate that differences from non-programhouseholdsinprogramvillages are statisticallysignificant at the 10, 5, and 1percentlevels, respectively. 59. Despite limited tree planting, trees are commonly found on cultivated plots. More than three-quarters of the cultivated plots inthe sample had trees on them, and about one-third had faidherbiu albidu trees, which have traditionally been protected and promoted through farmer-managed natural regeneration (Adam et al. 2006). These data suggest that farmer- managed natural regeneration is occurring on a very large scale inNiger, consistent with the findings of Larwanou, Abdoulaye, and Reij (2006) for Zinder and of Mortimore et al. (2001) for Maradi, andwiththe "regreening" o fNiger found by Hermann, Anyamba, and Tucker (2005). 60. Many land management practices are more common in S L M program villages, especially (but not only) among participants in SLM programs. Zai, half-moons, stone bunds, and organic fertilizer are all more commonly usedby nonparticipants inprogram villages than by households in nonprogram villages, suggesting that the impacts of SLM programs on adoption of these land management practices "spill over" to nonparticipants in the program villages. However, direct participants in several programs adopted several practices at even higher rates than nonparticipants in the program villages, suggesting that direct participation in programs has an additional impact on adoption. For example, zai and half-moons are more commonly used by participants in PAC, PSPR, other government programs and NGO programs thanthey are bynonparticipants inprogramvillages. Inorganic fertilizer is more commonly used by participants in PGRN, other government programs, and NGO programs than by nonparticipants. Trees are more commonly found on plots operated by participants in all programs than those of nonparticipants inprogram villages, and participants in PGRN are more likely to have planted trees since 2001. Not all practices are more commonly usedby program participants or inprogram villages, however. For example, stone bunds are less common on the plots of participants in PAC, PSPR, and PGRN than on plots of nonparticipants in program villages; mulch is less commonly used by PAC participants than by nonparticipants; and use of 45 organic fertilizer, tree planting, and the presence of trees on plots i s more common in nonprogram villages than on plots of nonparticipants in program villages. These differences do not necessarily indicate program impacts, however, since they may be affected by differences in the characteristicsofthe villages, households, andplots ratherthanby program access. 61. Using matching methods to control for differences in village, household, and plot characteristics, adoption of several water harvesting and SWC practices is found to be promotedby SLMprograms(Table V.7). Considering only results that are robust to the choice of matching estimator, the findings include the following: Participants in PGRN are more likely to have half-moons on their plots than comparablenonparticipants inthe same villages. Participants inother government programs are more likely to have stone bunds, half- moons, andzai on their plots thancomparablenonparticipants inthe same villages. Participants in NGO programs are more likely to have zai on their plots than comparablenonparticipants inthe same villages. The magnitude of program impacts are as large as a 16 percentagepoint increase inprobability of using the practice (the impact of other government programs on probability of using zai), though in most cases the impacts are smaller than a 10 percentage point increase. Considering results that are significant using the NN estimator only, PAC participants are more likely to use stone bunds, half-moons, and zai than are nonparticipants in the same villages, and PSPR participants are more likely to use half- moons than are nonparticipants. Moreover, nonparticipants in program villages are more likely to use o f stone bunds, half-moons, and zai than are comparable households and plots in non-program villages, indicating that programs have positive spillover effects on adoption by nonparticipants in program villages. In contrast, nonparticipants in program villages are less likely to use mulching than are households in non- program villages. Perhaps other SWC practices, such as zai and half-moons, substitute for mulchinginprogram villages. 62. SLM programs have more mixed impacts on tree planting and protection in cultivated land. Participants in other government programs and in NGO programs are more likely to planttrees than are nonparticipants inthe same villages, suggestinga positive impact of these programs on tree planting. The evidence does not support the claim that farmer-managed natural regeneration is being promoted by SLMprograms (since the presence of trees in general and faidherbia albida in particular is not more likely in program villages or for program participants). 63. SLM programs do not have statistically significant impacts on the adoption of organic or inorganic fertilizer. Given that the focus of SLM programs is generally not the promotion of inorganic fertilizer use, this finding is not surprising. What i s more surprising is that the programs have little effect on organic fertilizer use, since it is often usedincombination withzai or half-moons incultivated areas (Hassane, Martin,andReij 2000). Apparently farmers are already well aware of the benefits of organic fertilizer, but they may be constrained by its availability, explaining why its promotionby SLMprograms couldhave limitedimpact. 64. By promotinglastinginvestments, SLM programs can have long-term impacts even after the program ends. The most obvious example of long-term impacts is tree planting; 46 evidence also suggests that half-moons are more common among former PGRNhouseholdsthan they are amongcomparable householdsinthe same villages, eventhough that program ended in 2001 (Table V.7). However, the evidence does not show greater investment inhalf-moons or any other land management practices since 2001 by former PGRN households (see Report I1of this study for results). Thus, the longer term impacts of this program appear to be due to its effect on the adoption of longer term investments during the periodthat the program operated rather than its impacts on household landmanagementdecisions after the program ended. 47 I I * * * E *0 8 vl 8 *2 * -9: E 8 I I * * * 39 2***m * 2 8 hl N 8 7 I I * *w m 9 7 w W 8 8 2 *w 8 I- 2 9 0 vl m 8 2 8 d m 8 b, 8 * N m 8 8 z ** 8 8 q00 * * 82 q 0 5.4 Impactsof Povertyon Adoption of HouseholdLandManagementPractices 65. Many land management practices are inhibitedby aspects of poverty. For example, farmers with less access to irrigation are less likely to adopt zai, stone bunds, and inorganic fertilizer, probably because the returns to such practices are greater if they are combined with irrigation (see Report I1for results). Households with smaller endowments o f male labor are less likely to invest in stone bunds and half-moons due to labor constraints. Households with less livestock are less likely to invest instone bunds or organic fertilizer, the latter finding most likely because o f the lack o f availability o f manure. Land tenants are less likely to invest inzai, stone bunds, or organic fertilizer than owner-operators, probably because o f short term tenancy that limits their ability to recoup the value o f such investments. Village chiefs and other leaders are more likely than other farmers to invest in several practices, such as mulch, stone bunds, and organic and inorganic fertilizer, possibly because they have better access to information about the benefits o f such practices. 66. Not all endowments of physical, natural, human, or social capital are associated with increased land investments or with all types of land investments. For example, households with more land are less likely to invest in stone bunds on a given plot, possibly because they have labor constraints or less need to invest. Households with more livestock are less likely to mulch, probably because crop residues are more valuable to them as fodder than as mulch. Households with more farm equipment are less likely to invest in stone bunds, probably because the opportunity cost o f their labor is greater than households with less farm equipment. Households owning more durable household assets are less likely to invest in stone bunds, to plant trees, or to use organic fertilizer, and more educated households are less likely to invest in half-moons, possibly also because o f higher labor opportunity costs. 67. These results indicate that some aspects of poverty can contribute to poor land managementand land degradation, but this is not universal. Since poverty can imply lower opportunity costs for labor, it may promote land management investments that have relatively low returns to labor, even though it may inhibit investments whose returns or feasibility are affected by complementary assets (like irrigation or livestock), limited awareness, or tenure insecurity. Such complex relationships between household endowments or other aspects o f poverty and land management decisions are consistent with theories o f agricultural household behavior in the presence o f imperfect markets, and are found in a large number o f studies (see Nkonya, et al. 2008 for a review o f the literature on this issue, and for some recent empirical results from Uganda). 5.5 ProductionImpactsand Profitabilityof HouseholdLandManagementPractices 68. Zai, organic fertilizer, and inorganicfertilizer significantlyincreasecrop production per hectare. Average crop yields on plots using all o f these practices are nearly double those without these inputs (Table V.8). Using matching methods to select comparable plots and households, the impacts o f these practices are statistically and quantitatively significant (Table V.9). For example, the results o f the PSM model imply that the mean value o f production per hectare is 24 percent higher on plots withzai thanon similar plots without zai. The results o f the 49 NNmodelare quite similar to the PSMresults. The estimated impacts of organic fertilizer use are similar in magnitude, while the impact of inorganic fertilizer i s somewhat smaller (a 16 percent increase). The estimated impacts of organic and inorganic fertilizer are also statistically significant and even larger inthe OLS model (a 33 percent increase due to organic fertilizer and a 19 percent increase due to inorganic fertilizer). Stone bunds and mulch also have positive predicted impacts, although the impact is only weakly statistically significant inthe PSM model (to the 10 percent level). None of the other practices has statistically significant impacts on mean crop production. Table V.8: Crop yields on plotswith andwithoutzai and organic and inorganicfertilizer Zai Organic fertilizer Inorganic fertilizer Number of plots Mean yield Without Without Without 1,135 285.1 With 216 398.7 With Without 786 361.5 With 219 449.9 With Without Without 87 294.7 With 13 206.9 With Without 130 534.5 With 33 558.3 Source:Estimatedby authors based on IFPRVICRISAT household survey. Note: Crop yield data include total for millet, sorghum, cowpeas, and groundnuts. Table V.9: Estimated Impacts of Land Management Practices on Value of Crop Production (percentage change) Land management practice PSM NN OLS Zai +24.4** +22.1** +5.5 Half-moons -1.9 -1.2 -1.8 Stone bunds +20.6* +16.9 +9.5 Mulch +61.4* +16.9 +12.2 Trees (all types) -10.6 -2.3 -9.8 Faidherbia albida trees -15.7 -9.9 NE Organic fertilizer +22.6*** +25.9* ** +33.1*** Inorganic fertilizer +15.7** +14.7* +19.4** Source:Estimatedby authors based on analysis o fIFPRIACRISAT householdsurvey. Note: PSM indicates propensity score kernel matching; NN indicates nearest-neighbor matching with bias correction; OLS indicates ordinary least squares regression; and NE indicates that the regression could not be estimated. Other explanatory variables inthe regression, besides land management practices, included village fixed effects, participation in S L M programs (PAC, PSPR, PGRN, other government programs, and NGO programs); gender, age, education, and social status o f the household head; number o f men, women, and migrants in the household; household dependency ratio; household endowments o f land (area owned and share irrigated), livestock, farm equipment, and household assets; and the status o f plot area, tenure (individual vs. collective, means o f acquisition), soil texture, and soil fertility. Fullregression results are reported inReport 11. *, **, *** meanstatistically significant at the 10, 5, and 1percent levels, respectively. 50 69. Zai, stone bunds, and organic and inorganic fertilizer all reduce the probability of low yields and have a more favorable impact on low-productivity plots. Graphs of the cumulative distribution function (cdf) of crop production per hectare for matched plots with and without these landmanagementpractices show that zai and both organic and inorganic fertilizer shift the entire distributions to the right, with greater impact on lower productivityplots (Figures A3.1 to A3.3 inAnnex 3). The cdf for plots with stone bunds lies to the right of the cdf without stone bunds for low-productivity plots but to the left of the cdf without stone bunds for high- productivity plots (Figure A3.4), indicating that stone bunds help to reduce the variance of production, even if the impact on mean production i s less clear. Other practices-half moons, mulching, and tree planting-show little systematic impacts on the shape or location of the distribution functions. Consequently, these practices have little apparent impact on either production risk or meanproduction. 70. Based on the estimated production impacts, and the costs of these practices, zai appear unprofitable, organic fertilizer use breaks even, and inorganic fertilizer is marginally profitable on average. As noted above, the matching results predict that zai increasethe value of crop productionper hectareby as muchas 24 percent on average. The value of this production increase, basedonthe meanvalue o fproduction inthe sample, is about 10,000 FCFA per hectare. Assuming, consistent with MDAPAC (undated) and Abdoulaye and Ibro (2006), that it costs 50,000 FCFA per hectare to establish zai and 24,000 FCFA per hectare per year thereafter to maintain them (valuing labor at the median market wage rate), zai are not pr0fitab1e.l~This finding is irrespective of the discount rate used to evaluate profitability, since the value of the meanincrease inproduction is less that the estimated annual cost of maintaining zai. The mean production impact of organic manure use in the analysis is worth about 11,000 FCFA per hectare, and the median labor cost of applying manure i s 11,000 FCFA per hectare. Hence, manure application breaks even on average. Inorganic fertilizer is marginally profitable on average, with an estimatedmarginal value-cost ratio (VCR) of 1.6.'' 71. The production impacts and profitability of zai are underestimated to the extent that the plots on which zai were constructed were previously unused. As noted earlier, Hassane, Martin, and Reij (2000) claim that most zai plots in their study sites in the Tahoua region were barren, uncultivated lands prior to reclamation with zai. Unfortunately, rigorous evidence on this issue is lacking, and the IFPWICRISAT survey did not collect information on the prior uses of zai plots. Some systematic differences are found when comparing the soil characteristics of plots with and without zai. Plots with zai are less likely to have sandy soils (42 percent o fzai plots have sandy soils compared with 56 percent of non-zai plots) and more likely to have sandy-clay soils (34 percent of zai plots have sandy-clay soils compared with to 22 percent of non-zai plots). This i s because zai are more useful inheavier clay soils that are prone to crusting (MDAPAC undated). Still, the differences inplots with and without zai were not so "TheseestimatesarecomparabletothemedianlaborcostforzaiestimatedbasedontheIFPRIACRISAThouseholdsurveydata (40,000 FCFA per hectare). The estimatedcost of maintenanceis less thanthe cost of initial constructionbecause farmers usually reuse previouslyconstructedholes(AbdoulayeandIbro 2006). '*The marginal VCR measures the ratio of the value of additional output produced from an additional unit value of fertilizer input. It is generally argued that a VCR of greater than 2 is neededto achieve widespread adoptionof inorganic fertilizer, due to productionrisks andother considerations(CIMMYT 1988). The VCR was estimatedbasedon econometric regressions(available on request). The VCR estimated for inorganic fertilizer use is less than the VCR estimatedby Pender et al. (2008), which was greater than 3; reflectingthe more favorableagroecologicalandmarket access conditionsofthe villages intheir sample. 51 great as to make it impossible to find comparable non-zai plots to match the sample zai plots. Thus, despite the systematic differences mentioned above, such differences are not large enough to support the conclusion that all zai plots would be otherwise unusable. Furthermore, as noted by Hassane, Martin, and Reij (2000), unused land with the type of soil for which zai are most effective is limited, which implies that opportunities for restoring unused land using zai are similarly limited. 72. Part of the reason for the lower estimated impacts of zai and organic fertilizer reportedin this analysis comparedwith those in some previous studies may be particular conditions during the year studied or differences in the sample frame. In the survey, the average total yield of millet, sorghum, cowpeas, and groundnuts on plots without zai and organic or inorganic fertilizer is 285 kilograms per hectare, muchhigher thanthe six-year average district yields of 125 kilograms per hectare reported by Hassan, Martin, and Reij (2000), but similar to the average yield for the most favorable rainfall year in their study period (296 kilograms per hectare in 1994). Relatively favorable rainfall in 2006 (CILSS 2006) could account for some of the difference betweenthe results of this review and those of these earlier studies. Furthermore, the impacts found in the previous studies are not necessarily representative of the broader set of regions included in this study, and impacts estimated in those studies may therefore be larger than the average impacts occurring throughout most ofNiger. 73. Even if zai and manure are not profitable at the market wage rate in a particular year, farmers still may adopt such practices based on their risk-reducing advantage or because households have surplus labor at the time-of constructingzai. If households value their family labor at less than market wage rates because the investments are made during slack labor periods or because households face constraints in obtaining employment at these wages, investing in zai may still be worthwhile. Furthermore, the returns to zai in the study year may have been atypical. Hence, zai cannot be considered unprofitable for all households or in years other than the sample year; rather, it can only be determinedthat zai appear to be unprofitable basedon the meanreturns and labor costs estimated for the sample households inthis review. 5.6 Impactsof SLMProgramson Crop ProductionandIncomes 74. Although SLM programs contribute to adoption of several land management practices,most programs have limited impacts on the value of crop production, and PAC participation has a negative association with crop production. Using either matching estimator, as well as OLS and IV regressions, PAC participants obtained 30 to 35 percent lower mean values of crop production per hectare (depending on the estimator) than nonparticipants in PAC villages (Table V.10). The estimated impacts of participation in other programs are generally smaller and less robust to estimation method. PSPR participants have higher value of crop production per hectare than nonparticipants in PSPR villages, but the difference is only statistically significant using the NN estimator. Participants in NGO programs also have higher value of crop production per hectare than nonparticipants in NGO program villages, but this result i s only weakly significant using the NN estimator. In contrast, PGRN participants have lower values of crop production per hectare than nonparticipants in PGRN villages, but this is only statistically significant using the PSM estimator. Nonparticipants in program villages have 52 higher values of crop production than similar households in nonprogram villages, but this difference is only statistically significant using the PSM estimator. The lack of robustness of most of these results offers limited confidence that any o f these programs, with the exception of PAC, is having an impact on crop production. 75. The negative associationof PAC participationwith crop productionmay be due to the promotion of income-generating activities other than crop production. Despite the significantly lower crop production of PAC participants compared with nonparticipants, participants' per capita income i s not significantly lower (Table V.10). This suggests that PAC participants are pursuing other activities that compensate for their lower crop income. It may take more time for the full impacts ofparticipation inPAC activities on household incomes to be realized, particularly in the case of those involving substantial initial investments and capacity building. Further research on this issue i s needed to investigate why PAC participants are obtaining lower crop production, how they are compensating for this, whether this is a short- or long-term phenomenon, andwhat to do about it ifit is a longer term problem. Table V.10: Estimated Impacts of Programs on Value of Crop Production and Household Income per Capita percentage change Value of crop production Householdincomeper capita Comparison PSM NN OLS PSM NN OLS PAC participantsvs. nonparticipantsinPAC villages -29.8* -34.7*** -30.8** -19.0 t7.8 -22.1* PSPRparticipantsvs. nonparticipantsinPSPR villages +23.2 t32.2** t13.5 t14.1 t16.0 -5.7 PGRNparticipantsvs. nonparticipantsinPGRN villages 43.5*** -14.5 -17.4 +10.0 +9.2 t4.9 Other government programparticipantsvs. nonparticipants -4.2 t9.6 +17.1 +8.4 t19.6 t3.6 NGOprogramparticipants vs. nonparticipants t24.7 +23.1* t3.5 -1.5 +15.4 +6.4 Programnonparticipantsin programvs. nonprogram t14.8** +7.4 NF! +14.9*** t22.6*** NE Source: Estimatedby authorsbasedonanalysis o f IFPRIhCRISAThouseholdsurvey. Note: PSM indicates propensity score kernel matching; NN indicates nearest-neighbor matching with bias correction; OLS indicatesordinary least squares; andNE indicatesthat the impact was not estimateddue to the use of village fixed effects. Other explanatoryvariables inthe OLS regressionare the same as those reportedinthe note to Table V.9. Fullregressionresults are.reportedinReport11. *, **, *** meanstatistically significant at the 10, 5, and 1percentlevels, respectively. 76. The lack of positive and robust impacts on crop production resulting from participation in SLM programs likely stems, in part, from the fact that many of the household-levellandmanagementpracticespromotedby the programshavelimitedimpact on crop productivity. For example, although several programs increase adoption o f half-moons 53 and stone bunds, these practices have limited estimated impacts on mean crop production. Furthermore, even for practices that do have a positive impact, such as zai, program impacts are generally fairly small (increasing the probability of the use of zai by less than 10 percentage points in most cases). And the programs have not promoted increased use of either organic or inorganic fertilizer, whichcould have larger impacts on production. 77. Another reasonfor the limited impact of program participation on crop production i s that many of the activities of SLM programs focus on community-level activities, not increased crop production. Promoting community activities, such as planting trees and constructing half-moons incommunal areas, and other activities, such as developing cerealbanks or water supplies, may not significantly increase crop production or even the incomes o f direct participants. This does not meanthat community membersdo not benefit; rather it meansthat the benefits spreadto non-participants as well. 78. Despite the limited impacts on crop production, program activities appear to contribute to higher incomes in program communities more generally. Using both matching estimators, incomes of nonparticipants in program villages are found to average as much as 23 percent higher than incomes of comparable households in non-program villages. This suggests that the programs are having broader community-level impacts on incomes inprogram villages, even ifthe incomes o f direct program participants are not substantially increased. 79. The impacts of SLMprogramsvary acrosswealthier and poorer households, across the country's different regions and agro-climatic zones, and by type of program. For example, the negative association of PAC with crop production occurs only for the wealthier households inthe sample (Table V.11). This suggests that wealthier householdsmay be attracted to PAC-promoted activities that are not related to crop production, such as cereal banks, input- supply shops, livestock fattening, dairy production, food processing, and others. NGO programs have a significant positive impact on crop production inthe Dosso region, perhaps because they focus on promotinghigher value crops inthis more favorable agricultural region. Consistentwith this, NGOprograms haveapositive impact onhousehold incomes inthe agro-climatic zone with the highest rainfall.lg Table V.ll: Estimated Impacts of Programs on Value of Crop Production and HouseholdIncome per capita byWealthofHousehold Percentage change Value of crop production Household income per capita Program or Pooresthalf Wealthiest half Poorest half Wealthiest half program type of sample of sample o f sample of sample PAC -10.6 41.1** -19.7 -21.1 PSPR +5.0 +14.4 +0.8 -12.4 PGRN -24.2 -16.8 -12.2 +52.0 Othergovernmentprograms +46.3* +8.9 -11.7 +17.4 NGOprograms +1.5 +22.9 +24.5 -6.9 Source: Estimatedby authorsbasedon analysis of IFPRVICRISAThouseholdsurvey. Note: Resultsare basedonordinary leastsquaresregressions. These results are reportedinReport I1ofthis study. 54 *, **, *** indicates that results are statistically significant at the 10, 5, and 1percent levels, respectively. 5.7 Other ImpactsReportedby ParticipantsinSLMPrograms 80. Household participants in SLM programs generally report that programs have favorable impacts on a wide variety of environmental, social, and economic indicators. More than half the program participants in all SLM programs (with the exception of PGRN) reported increases in the availability of useful trees (Figures V.4 to V.8). Most participants in PSPR, other government programs, and NGO programs reported increased availability and quality of pasture and fodder, increased availability and quality of cropland, increased agricultural productivity and food security, and improved welfare o f the people. It is clear that not all of the benefits of SLM programsare measured by impacts on crop productionand income. 81. Positiveimpactson these indicatorswere less commonlyreportedby participantsin PAC and PGRN, but they were still positive for a sizable fraction of participants in these programs. For example, less than 50 percent of PGRN participants reported that the program had significant positive impacts for every indicator (Figure V.6) and more than 50 percent of PAC participants reported significant positive impacts only for one indicator-availability of useful trees (Figure V.4). The dominant response across all indicators for these two programs (especially PGRN) was "no significant impact." Incontrast, more than 50 percent o f participants inPSPR, other government programs, and NGO programs reported significant positive impacts on many indicators (Figures V.5, V.7, and V.8). It is not clear why the perceived impacts of PAC and PGRN are generally less favorable, but this finding is consistent with the less-favorable impacts of PAC on crop production and income that were found. PAC may be able to learn from the morepositive results reported for some other programs; further study ofthis issue is needed. 82. Favorable program impacts were less commonly reported for several other indicators,such as the presence of predatorsor pests, the availability and quality of water, and conflicts over natural resources. Most participants in all types of programs reported no program impact for these indicators. For some indicators (presence of predators and pests and availability and quality of water) a small percentage of program participants reported negative program impacts. As with the qualitative impacts reported by community-survey respondents, these household-level responses indicate that some potential trade-offs can result from SLMprogramactivities,althoughtheseappearto beuncommon. FigureV.4: PerceptionsofPAC Impactsfrom PACParticipants Share ofprogram partic@ants(%) 55 100% T 90% j- I 80% 1I 70% 5 I 60% 50% -$ II 40% 4 30% -I 20% ~ 10% : i 0% "4- Major deterioration Minor deterioration Insignificant impact Minor improvement Major improvement Source: Calculated by authors. 56 FigureV.5: Perceptions ofPSPRimpactsfromPSPRParticipants Share ofprogram participants (%) 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Major deterioration Minor deterioration- *Insignificant impact Minor improvement Major improvement " _ - - ~ " ___. --.----..--- _ _ "- - I" ~ ^^___ I"__^"" I "" _I" Source: Calculatedby authors. 57 FigureV.6: PerceptionsofPGRNImpactsfromPGRNParticipants Share ofprogram participants (%) Major deterioration Minor deterioration Insignificant impact Minor improvement Major improvement _____-,__l_l. _. __" .,..,,.._........._ " _,.._._"l" ___I_I ll-..l.l..l".""._ ............. ... ource: Calculated by authors. 58 FigureV.7: PerceptionsofProgramImpactsby Other Government ProgramParticipants Share ofprogram participants (%) Source: Compiledby authors. 59 Figure V.8: Perceptions of Program Impacts by Nongovernment Organization Program Participants Share ofprogram participants PA) Majordeterioration Minordeterioration I lnsigtiifieantimpact Mitiorimprovement Major improvement Source: Calculated by authors. 60 5.8 Summaryof MainFindingsandImplications 83. The mainfindings andtheir implications ofthis study canbe summarized as follows: Programcoverage andparticipation Main finding: SLM programs serve only a small minority of communities and households, better serving communities having more favorable rainfall and market access and wealthier households. Implication: Much more investmentin SLMprograms is neededto reach most of Niger's areas andpeople. Communitylandmanagementinvestments Mainfindings: Community tree plantations and protected areas earn high returns on average and are sufficient to pay for the costs of SLM programs, but the returns are highly variable. These investmentshave positive impacts on a wide range of social and environmental indicators, though there are tradeoffs insome cases. Implications: More investment by SLM programs to promote community land management investmentsis justified, but efforts to better target and increase effectiveness o f investmentsare also needed. The case for such investments goes beyond their economic impacts, given other environmental and social benefits. But potential tradeoffs should also be considered. Householdlandmanagementpractices Mainfindings: SLMprograms are contributing to increasedadoption of severalhousehold level landmanagementpractices, especially SWC practices and tree planting, but the impacts of these practices on crop productionand income are limited. Greater production and income impacts would bepossible if SLMprograms promoted soil fertility managementto a greater extent. Various aspectsofpoverty can inhibit some types of landmanagementpractices and promote others. Implications: Increased emphasis on soil fertility management and better targeting of SWC measures i s needed. No one-size-fits-all approach to promoting SLM will work; approaches have to be adaptedto local contexts and farmers' constraints and capabilities. Impactsof SLMprogramson outcomes Maiifindings: Most programs have statistically insignificant impacts onthe value of crop productionor income of participants compared to non-participants inthe same communities. However, incomes are higher for all householdsinprogram communities thancomparable households innon-program communities, indicatingmore general community level impacts. 61 Wealthier PAC participants have lower value of crop production thancomparable non- participants, but their incomes are similar. Manyparticipants inall programs report favorable impacts on many social and environmental indicators, though positive impacts were less commonly reported by participants inPAC and PGRNthanother program participants. Implications: Better selection and targeting of household level land management practices is needed to improve impacts on crop production and income. Further research is needed to measure and value other benefits of program interventions and to investigate why PAC participants are obtaining lower crop productionand have less favorable impressions of impacts thanparticipants inother programs, whether this problem is continuing over a longer term, and what meliorative actions are needed. 84. These mainfindings and implications, along withmore detailed ones, are developed inTable v.12. 62 Table V.12: Summary of Findingsand Implications Findings Implications Program coverageandparticipation Mainfindings: SLMprograms serve only a small Much more investment in SLMprograms is needed minority of communitiesand households, better to reach most ofNiger's areas andpeople. serving communitieshaving morefavorable rainfall and market access and wealthier householh, 1. Despitelarge investments by SLM programs inNiger, only asmall fraction ofcommunities andhouseholds have participated. 2. SLMprograms tend to operate inareas o fhigher Programs appear to be targeted to locations rainfall, better accessto markets and services, where they are more likely to be effective, but many andmore degraded soils. less favorable-areas are being left out. 3. For some programs, households participating in More effort is needed to reachthe poorer SLM programs tend to be wealthier than households within program communities. nonparticipants, even before the programs began. Communityland managementinvestments Mainfindings: Community treeplantations and More investment by SLMprograms topromote protected areas earn high returns on average community land management investments is and are suflcient topayfor the costs of SLM just@ed but efforts to better target and increase programs, but the returns are highly variable. eflectiveness of investments are also needed The These investments havepositive impacts on a casefor such investmentsgoes beyond their wide range of social and environmental economic impacts,given other environmental and indicators, though there are tradeoffs in some social benefits. Butpotential tradeoffs should also cases. be considered. 4. Community tree plantations are the most SLM programs are essential for investments in common landmanagement investment promoted community tree plantations, likely due to the by SLMprograms; they are found inmore than financial and technical requirements and the need 90 percent of program villages and are rare in for support. non-program villages. 5. The main economic benefit generated bytree Tree plantations are not achieving their full plantations is the value o f fodder and wood economic potential. produced; little gum Arabic is being harvested Increased efforts are needed to identifyand despiteplanting of acacia Senegal or acacia address the constraints to harvesting and marketing seyal inthe vast majority o f plantations. o f gum Arabic. 6. The survival rate o ftrees inplantations is low Increased efforts to identify and address (averaging only 54 percent) and highly varied. causes o f low tree survival rates are needed. Survival rates are higher for certain tree species The promotion o fbetter survivingtrees is and inPAC plantations, but there is little needed, as is the emulation o f the PAC approach to measurable impact o f SWC measureson survival promotion and assistance. rates. 0 Better knowledge o fhow to target SWC measuresinplantations is needed. 7. The mean internal rate o freturn to tree 0 Tree plantations are an attractive plantations is inthe range of 27 to 45 percent, investment, even without gum Arabic harvestingor depending on how the value o f fodder is carbon payments. estimated. IfgumArabic were harvested, the Increased investmentintree plantations is IRRwould increase by several percent (to 33- economically justified. 48 percent); ifthe value o f carbon sequestration were included, the returns would increase a few additional percent (to 36-50 percent). 8. Variability inreturns to tree plantations is very Increased efforts are needed to identify and 63 Findings Implications high, however, with most ofthose sampled address low tree survival rates and constraints to providing zero current income benefits in2006. increasing returns fiom fodder production and other products includinggum Arabic. 0 Investments in livestock improvement, other high-value trees could increase returns. 9. Some communities are protecting communal 0 Improvements incommunal rangeland areas for natural regeneration, inmost cases management either through plantations or protected promoted by an SLM program, although this is areas are uncommon outside o f program much less common than tree plantations. communities; hence, continued degradation inmost such areas is likely without substantial expansion o f SLM programs. 10. Protected areas yield significant benefits 0 Protectedareas are an attractive investment relative to the costs (mainly interms o f for S L M programsto promote, especially if additional fodder production), yielding very financial or other constraints limit the potential o f highrates ofreturn, especially for those areas plantations. promoted by programs (> 170 percent); despite 0 Plantations are a better investment if earninga higher IRR,the net presentvalue o f finance is available at a reasonable interest rate and protected areas is less than the value for tree they are no other prohibitingconstraints. plantations at a 10 percent discount rate. 11. Overall, the total estimated economic benefits 0 Programs have paid for their SLM fiom tree plantations and protectedareas activities, only consideringthe returns to tree promoted by major programs are inthe range o f plantations and protectedareas. 73 to 114 billion FCFA, which is greater than 0 Increased investment by donors and the the expenditures by these programs onall SLM Government o fNiger in SLM activities is activities (roughly estimated to be 69 billion economically justified. FCFA). 12. Many communities have also invested in SWC 0 SLM programs could obtain higher activities inunprotectedcommunal areas economic returns by reducing their promotion o f without tree planting, usually promoted by a SWC measures inunprotectedareas and infavor o f SLM program; the costs o fthese areas are more investmento fplantations or protectedareas. significant, andthe economic benefits o fthese investments are very low. 13. Inaddition to economic benefits, community 0 The economic costs andbenefits do not tell landmanagement investmentsare perceivedby the whole story; other impacts should be communities to generate many environmental considered. and social benefits. 0 The value ofcommunity landmanagement investments i s likely to be greater than estimated if other benefits could be valued. 14. The reported benefits are generally greater for 0 Although the full benefits o f SWC plantations and protectedareas than for SWC investments inunprotectedareas are likely to be investments inunprotectedareas, though greater than their estimated economic value, they benefits are also reported for a significant are still likely less thanthe benefits o fplantations minority o fthese areas. andprotected areas. 0 There is a needto use appropriate combinations o ftechnologies, such as combining effective and well- targeted SWC measureswith tree plantations; stand-alone investments in SWC technologies are less likely to generate large benefits. Household land managementpractices Mainfinding: SLMprogramsare contributing to Increased emphasis on soilfertility management increased adoption of several household level land and better targetingof SWC measures is needed. managementpractices, especiallySWCpractices No one-size-fits-all approach topromotingSLMwill 64 Findings Implications and treeplanting, but the impacts of thesepractices work; approacheshave to be adapted to local on cropproduction and income are limited. contexts andfarmers' constraintsand capabilities. Greaterproduction and income impacts would be possible if SLMprogramspromoted soilfertility managementto a greater extent. Various aspects of poverty can inhibit some types of land management practices andpromote others. 15. The most common (active) landmanagement 0 Except for organic and inorganic practices found inthe analysis are (indecreasing fertilizer, none o fthe land management practices order o f importance) use o f organic fertilizer, found is likely having a substantial impact on inorganic fertilizer, zai, half-moons, stonebunds, aggregate crop production inNiger due to their tree planting, and mulching; none o fthese limitedadoption. practices is used on more thanhalfo fthe plots, andall except organic and inorganic fertilizer are usedon less than 10percent o fplots. 16. Protection o ftrees on cultivated land is 0 The results support the view that farmer- arguably the most common land management managednatural regeneration is widespread. practice: more thanthree-quarters o fplots have Environmental rehabilitation may depend trees, and about one-third have gao (Faidherbia more on how farmers manage trees intheir albida) trees, which have traditionally been cultivated landsthan on community investments in protectedinmany areas. tree plantations. 17. SLM programs are increasingadoption o f 0 SLM programs are effective inpromoting several SWC practices, includingzai, half- SWC practices; whether this should continue moons, and stone bunds. depends on the impacts o fthose practices. 18. Some S L M programs promote tree plantingby 0 It is important to consider how SLM participating households on their own land, but programs affect tree plantingand management on program impacts on farmer-managed private as well as communal lands. regeneration o ftrees appear to be limited. 0 Programs such as PAC that have limited impact on tree planting on farms can learn from others that are effectively inpromotingthis practice. 0Furtherresearchonthe impacts of SLMprograms on farmer-managed natural regeneration is needed. 19. SLM programs are having insignificant impacts 0 Although effectively promoting SWC and on adoption o f organic and inorganic fertilizer. tree planting, SLM programs are paying inadequate attentionto soil fertility. 20. Some programs are having long-term impacts 0 Programs can have longer term impacts by by promoting long-term investments such as in promoting longer term landmanagement half-moons and tree planting; however, the investments. evidence for PGRN, a program which ended in 0 Promoting increased adoption o f S L M 2001, suggeststhat program impacts on practices beyond the life o f a program is difficult adoption o f such measures occurred only and may require a different program approach, duringthe program's implementation. such as teaching principles o f SLM innovations rather than promoting specific practices. 21. Manyaspectso fpovertyreduce adoption o f ' 0 Actions to address these aspects o f some land managementpractices: for example, poverty can help to improve some aspects of land lack o f accessto irrigation i s associatedwith management. lower usage o f severalpractices: lack o faccess to male labor reduces adoption o f some labor- intensive practices; lack o f livestockreducesuse o f organic fertilizer; and landtenants investless inseveralpractices having longer termbenefits. 22. Nevertheless, severalaspects o fpoverty are 0 N o one-size-fits-all approach to SLM is 65 Findings Implications associatedwith greater adoption o fsome land- likely to be successful for all farmers. managementpractices, probably because poorer 0 Poverty-reduction strategies such as households have lower opportunity costs for promoting asset accumulation and education will labor or other inputs:for example, households not necessarily solve land-degradation problems with less landor farm equipment invest more in and insome cases may worsen them. stonebunds, those with less livestock are more Targeted efforts to promote SLM likely to mulch, those with less householdassets practices that are suited to the endowments of are more likely to use several Practices, and less- different types o fhouseholds are needed, along educatedhouseholds investmore instonebunds. with broader poverty reduction efforts. 23. Zai, organic fertilizer, and inorganic fertilizer 0 Promoting SWC measuresthat have little all increase mean crop production andreduce impact (such as half-moons) is unlikely to be the risk oflow production; stonebundsalso successful on a large scale unless impacts can be reduce risk of low production, although they improved. have a less-clear impact on mean production; Lack o f impacts on crop production other practices (half-moons, mulching, tree through tree plantingis a favorable sign, indicating planting) have insignificant impacts on crop that tree plantingor protection incropland (at production. current densities) can provide valuable products without reducing crop production. 0 Increased promotion o fzai and soil- fertility management practices has the potential to substantially increase crop production and reduce low production risks. 24. Inthe sample, zai were not profitable to farmers 0 Zai have less potential for widespread on average, organic fertilizer breaks even, and adoptionthan do soil-fertility management inorganic fertilizer is marginally profitable. practices. 25. The profitability o fzai is likely to be much 0 Zai are more likelyto be adopted inareas greater where they are necessaryto cultivate where unusedlandwith crusted soils canbe otherwise unusable land. rehabilitatedthrough the use o fzai; however, such land appears to be limited inavailability. 26. Despite the limitedprofitability o fzai and 0 Expectedprofitability is not the only organic fertilizer found inthis analysis, farmers important considerationto farmers inadopting may still adopt these practices to reduce risk land-management practices, although it is becausereturns are higher inparticular contexts important, especially helpingto reduce poverty. or years, or because o f surplus family labor Promoting these or other practices must reducestheir labor costs. take local contexts and farmers' decision criteria into account; top-down promotion o f land management practices is likely to fail. Impacts of SLMprograms on outcomes Mainfindings: Mostprograms have statistically Better selection and targeting of householdlevel insignificant impacts on the value of crop land managementpractices is needed to improve production or income ofparticipants compared impacts on cropproduction and income. Further to non-participants in the same communities. research is needed to measure and value other However, incomes are higherfor all benefits of program interventions and to investigate householch inprogram communitiesthan why PACparticipants are obtaining lower crop comparable households in non-program production and have lessfavorable impressions of communities,indicating moregeneral impacts thanparticipants in otherprograms, communitylevel impacts. WealthierPAC whether thisproblem is continuing over a longer participants have lower value of crop term, and what ameliorative actions are needed. production than comparablenon-participants, but their incomes are similar. Many participants in allprograms reportfavorable impacts on many social and environmental indicators, thoughpositive impacts wereless 66 Findings Implications commonly reported byparticipants in PAC and PGRNthan otherprogramparticipants. 27. Most programs have limited impacts on crop 0 PAC participation may reduce crop production; PAC, which has negative impact, is production by promotingother activities. the exception. 28. Despite a negative association with crop 0 This supports the view that the negative production, participants inPAC do not have impact o fPAC participation on crop production is significantly lower incomes compared with due to the promotion o f other activities that nonparticipants. compensate for lost crop income. 29. The negative association o fPAC participation 0 The opportunities for PAC participants to with crop production is found only for the shift to noncrop activities appear to be greater for wealthier households; the impacts of PAC on wealthier households. income are insignificant for wealthier as well as The impacts o fPAC participation on poorer households, however. income levels or distribution are not very large; they might increase over time, however. 30. Participation inother programs has limited 0Other investmentsbesides SLMprograms will impacts on crop production or overall income continue to be needed to address poverty. compared with nonparticipants inprogram communities. 31. Incomes o f households inprogram 0 S L M programs can play an important role communities (both interms of participants and inhelpingto addresspoverty, eventhoughthey nonparticipants) are significantly higher than are unlikely to solve it alone. incomes o f comparable households in nonprogram communities (as much as 23 percent). 32. Household participants inSLM programs report 0 Economic indicators such as crop favorable impacts of these programs on a wide production and income do not tell the whole story- range o f social, environmental, and economic -other impacts o fprograms need to be taken into indicators. account; increased efforts to measure and value such impacts are also needed. 33. Favorable impacts were less commonly 0 Less-favorable impacts for PAC interms reportedby participants inPAC and PGRN o fhousehold perceptions are consistent with the thanbyparticipants inother programs, although lower crop production estimations, suggesting a sizable fiaction o f these participants do report cause for concern. positive impacts. 34. Less-favorable impacts were reported for some 0 Potentialtrade-offs o f SLM program indicators, and in some cases adverse impacts activities need to be kept inmindand investigated. were reported for some indicators (for example, A broader approach to SLM, such as the presence o fpests or predator and the watershed management, may be neededto availability and quality o f water); reported adequately consider and address trade-offs. adverse impacts are uncommon, however. 0 Mitigating negative impacts appears possible given that these are not commonly reported and not often considered to be significantly important. Source:Compiled by authors. 67 VI. Recommendations 85. .Recommendations based on the findings of this study and other information were developed following discussions held in working groups during a broad-based national stakeholder workshop heldinNiamey during July 2008. Eleven recommendationswere made for improvingthe effectiveness of actions relatingto SLM inNiger. These recommendations concern development strategies and actions; policy and program monitoring and knowledge management; and capacity building, researchandtechnical assistance: Recommendationsrelatedto developmentstrategiesand actions: - Develop a political consensus on intensifyingSLM integrated with the national Strategy of Accelerated Development and Poverty Reductionand the RuralDevelopment Strategy. -- Support decentralizedmanagementof naturalresources. Increase investments in SLM programs in the context of the Rural Development Strategy through the community development plans (PDCs) and enlarge the zones of intervention, in particularbyworking at watershed levels. - Continue and extend investments promoting SLM technologies with high returns (e.g., plantations, protection areas for natural regeneration, organic and inorganic fertilizer, agro- forestry). - Develop a global, integrated and efficient approachin SLMprograms including bothpublic and private goods dimensions and increased consideration of the climate risks (both mitigation and adaptation). - Continue investing in other public goods and services that will have a large impact on poverty reduction and rural development, basedon the priorities identifiedinthe PDCs. Recommendations related to policy and program monitoring and evaluation, and knowledgemanagement: - Implement a system of participatory monitoring and evaluation and knowledge managementat all levels with increasedharmonization andalignment among programs under the framework o f the national strategy for rural development (SDR) supported by a geographic informationsystem. - Replace the sectoral approach to knowledge management with a multi-sectoral approach; -includingknowledge of forestry, pastoralism, socioeconomics, andother relevant fields. Synthesize and conserve context-specific knowledge on SLM and disseminate findings in readily usable form to stakeholders,~ Recommendationsrelatedto teaching, capacity building,researchandtechnicalassistance: -- Strengthenteachingof SLMinuniversity courses. Conduct action-oriented research and provide technical assistance to address key issues for increasing the effectivenessof SLMinvestmentsinthe context of investments programs. - Increase knowledge sharing of good practices and results across the countries and at sub- regional levels 68 86. Detailed suggestions of how these recommendations should be implemented, by whom andwhen are summarizedinTable VI.1. 69 e e e . e h 0 3 b Y P) * . 0 3 .5 e.e c4 .-CC 'Q) :B e c a A w a (0 0 .I 3 e .e 3 cc C cF c4 b 3sr 3 i2 c + E c e a 'C 0 - 2 .-ca E *e 4 e 4 e .e 3 4 cr b B8 2b B8 b m m 0 0 0 0 mI cu 3 4 0 0 cuI m 0 I 2 cu 0 cu m 2 P a, .I 2-J a" m" B6 2 d 0 0 0 w o c u H cu m Z Z 3 3cui 0 Z" 0 W 0 F v? W 00 I 3 :- a" e I d i! 0 e d a e a, d WY 9 c.l 0 0 (.? 2 3 ?-. II c. I 0 Ccl 0 0 0 0 0 3 N N n E 0 0 0 3 2- 2 II- 13 Annex 2. Descriptionof CommonLandManagementPractices Banquettes (Photo A2.1). Banquettes are long embankments and trenches used for harvesting water and for tree planting for windbreaks and other purposes. These are commonly found incommunal areas constucted with support of SLMprograms. Dikes. Dikesare nonpermeableandthus serve to retainwater; consequently they can also cause a problemof water logging and therefore soil degradation. Maintenance of dikes is time-consuming and, combined with the problem of water logging, limits the use of this technology except for the rice production. Vegetative strips. These are strips of land, planted with perennial grasses. The grass cover protects the soil against wind and water erosion. The grass strips can stand alone or be usedin combination with stone bunds. Ifcombined, the grass strips should be planted on both sides of the stone bund. Grass strips combat erosion, and the grasses can also be used to produce handicrafts, such as baskets or to make roofs. The disadvantages of the grass strips are that the seeds can infest the fields, in which case extra weeding is required. Snakes can also hide inthe grass strips. Moreover, seeds have been distributed freely in some villages in the past. This seems to have caused problems of dependency and unwillingness on the part of farmers to buy the seed today. Farmers complain that seeds are not available totlay. Half-moons,or "demi-lunes"(Photo A2.2). Half-moonsare earthen ridges inthe.shape of a half moon, with a diameter of four meters. Earth in front of the half-moon is dug away at a depth o f 15-25 cm. This earth i s then used to create the ridge, which i s built along the slope of the plot to retain water. In years of low rainfall the technique i s in particularly useful to promote the infiltration of water inside the half-moon, where the crop is planted. Often the half-moon is combined with the direct application of organic fertilizer.'Half-moons are consideredto be labor-intensive, and uptakei s also low. Living hedges. Planting (living) hedges helps to preserve the soil and protect crops. Hedgesform a barrierto reduce wind and water erosion, which also can lead to improved yields through increases in organic carbon and total nitrogen in the topsoil. If planted densely enough, living hedges can protect crops from grazing livestock. They can stand alone or incombination with earthendikes or stone bunds.Perennial species are preferred for hedging, as are species that offer additional advantages, such as providing forage, roofing material, or remunerative benefits. In Mali, Jatropha curcas is a widely grown species used in living hedges, and women collect its seeds to make soap (which they sell). The possibility to produce (1amp)oil and biodiesel from Jatropha seeds i s underexploited by farmers, primarilythey lack the required machinery to press the oil. Stone bunds, rock bunds, contour bunds, or stone lines (Photo A2.3). All four terms refer to technologies involving the placing o f stones in a continuous line. Stone lines are the simplest technique and in many villages are known as a traditional technique. The term "stone lines" i s usedto describe short rows o f stones (a few to a few tens of meters) in places where runoff water concentrates and tends to form rills. The practice reduces 77 runoff and erosion and retains organic matter on the field. Stone or rock bunds and contour bunds refer to several layers of stones placed along a contour line involving multiple stones placed on top of each other. Stone bunds are usually longer, higher, and more effective in combating erosion than stone lines. The bunds reduce the force of the water runoff and encourage infiltration. This leads to the creation of a micro-climate that favors the establishment of natural vegetation after the rainy season. Stone bunds can be considered an improvement over earthen dikes, which may have the disadvantage of high maintenance costs and instability under intense rainfall. Stone bunds and lines also have the advantageofnot causingwater-logging problems, as impermeable earthen dikes do. Zai or "tassa" (Photo A2.4) translate as "improved planting pits." Traditionally, planting pits were tiny pits made with a hoe to breakthe surface crust infarmers' fields before the onset of the rains. Improved zai are much larger than traditional planting pits (20-30 centimeters across and 10-25 centimeters deep compared, roughly, with 10 centimeters across and 5 centimeters deep) to collect and store more rainfall and runoff. Manure or other organic matter i s usually added inthe pitsto improve soil fertility. 78 Photos of Selected Practices PhotoA2.1: BanquettesUnder Construction PhotoA2.2: Half-MoonsBeingConstructed by CommunityWorkers 79 Photo A2.4: Zai 80 Annex 3. CumulativeDistributionPlotsof Yieldswith andwithout Practices FigureA3.1: MatchedPlotsWith and FigureA3.2: MatchedPlotsWith and WithoutZai Without Organic Fertilizer 4 8 8 lnvaluopha10 12 14 0 5 10 15 Lnvalmpha I- c.d.f. of without zal -c.d.f. of with zal -c d f of withoutorganicfeltilk- c d f of with organicfertill Source: IFPRVICRISAThouseholdsurvey Source: IFPRVICRISAThouseholdsurvey Note: Kolmogorov-Smirnovtest for equalityof Note: Kolmogorov-Smirnovtest for equality of distributions: p=0.080* distributions:p=O.OOO* ** FigureA3.3: MatchedPlotsWith and FigureA3.4: MatchedPlotsWith and Without InorganicFertilizer Without StoneBunds 1- '1 -E.8- 1.6- a 394- ;' V .2- 4 8 8 lnvalcropha10 12 14 -c.d.f.of -c.d.f. of Without stone bund -c.d.f.of with stone bund I without inorganicfeltill- c.d.f. of with lnorgenicfertil tK Source: IFPRVICRISAThouseholdsurvey Source: IFPRVICRISAThouseholdsurvey Note: Kolmogorov-Smirnovtest for equalityof Note: Kolmogorov-Smirnovtest for equality of distributions:p=O.OOO*** distributions:p=O.OlO** 81 References Abadie, A., D. Drukker, J. Leber Herr, and G.W. Imbens. 2004. "Implementing Matching Estimators for Average Treatment Effects in Stata." StataJournal 4 (3): 290-3 11. Abdoulaye, T., and G. Ibro. 2006. Analyse des Impacts Socio-economiques des Investissements dans la Gestion des Ressources Naturelles :Etude de Cas dans les Regions de Maradi, Tahoua et Tillabery au Niger. Niamey et Amsterdam : Centre Regional d'Enseignement Specialise en Agriculture et 1'UniversiteLibre d'Amsterdam. Abdoulaye, T., and J. H. Sanders. 2005. "Stages and Determinants o f Fertilizer Use in Semiarid African Agriculture: The Niger Experience." Agricultural Economics 32: 167-79. Adam, T., C. Reij, T. Abdoulaye, M. Lanvanou, and G. Tappan. 2006. Impacts des Investissements dans la Gestion des Resources Naturalles (GRN) au Niger: Rapport de Synthese.Niamey, Niger: Centre Rdgionald'Enseignement Specialise en Agriculture. Ambouta, K.J.M., B. Moussa, and 0.Daouda. 2000. Rdiabilitation dejachere dCgradCe par les techniques de paillage et de zai` au Sahel. In : Floret C. and Pontanier R, (Eds.), La jach2re en Afrique Tropicale, Rcile, Amdnagement, Alternatives. Paris : John Libbey Eurotext, 2000 : 75 1-759. Bandra, P. and F. Batta, 1998. `Soil and Water Conservation (SWC) in Burkina Faso' ODI. VoisinsMondiaux. Butler, R. A. 2006. "A Place Out of Time: Tropical Rainforests and the Perils They Face." http://www.mongabay.com (January 6,2006, accessed May 2006). Chappell, A. 1996. "Modelling the S atial Variation o f Processes in the Redistribution o f Soil: Digital Terrain Models and 1 3 6 sinSW Niger." Geomorphology 17 (1/3): 249-62. CILSS (Comite Permanent Inter-Etats de Lutte contre la Secheresse dans le Sahel). 2006. AGHRYMET Monthly Bulletin No. M07/06. http://www.agrhymet.ne/mens-agrl septernbre06/Septembre%202006~UK~final.pdf (accessed October 2007). CIMMYT (International Center for Maize and Wheat Improvement). 1988. From Agronomic Data to Farmer Recommendation: An Economics TrainingManual. Completely revised edition. Mexico City: CIMMYT. Cook, T. D., and D. T. Campbell. 1979. Quasi-Experimentation: Design & Analysis Issuesfor FieldSettings.Boston: HoughtonMifflin. de Montgolfier-Koudvi, C., and H. N. L e Houdrou. 1980. Study on the Economic Viability of Browse Plantations in Africa. In H.N. Le Houdrou, ed., Browse in Afiica: The Current State of Knowledge. Addis Ababa, Ethiopia: International Livestock Centre for Africa. FA0 (Food and Agriculture Organization of the UnitedNations). 2008. FAOSTAT Databases. http://faostat.fao.org/ (accessed May 2008). 82 Feder, G., R. Murgai, and J. B. Quizon. 2004. "Sending Farmers Back to School: The Impact of Farmer Field Schools inIndonesia." Review of Agricultural Economics 26: 45-62. Hassane, A., P. Martin, and C. Reij.2000. Water Harvesting, Land Rehabilitation andHousehold Food Security in Niger: IFAD's Soil and Water Conservation Project in Ille`la District. Rome and Amsterdam: International Fund for Agricultural Development and Vrije Universiteit. Heckman, J., H. Ichimura, J. Smith, and P. Todd. 1998. "Characterizing Selection Bias Using Experimental Data." Econometrica.66: 1017-99. Henao, J., and C. Baanante. 2006. Agricultural Production and Soil Nutrient Mining in Africa: Implications for Resource Conservation and Policy Development. Muscle Shoals, AL, USA:InternationalFertilizer DevelopmentCenter. Herrmann, S. M., A. Anyamba, and C.J. Tucker. 2005. "Recent Trends inVegetation Dynamics inthe African Sahel and Their RelationshipTo Climate." Global Environmental Change 15: 394-404. Kapoor, K., and P. Ambrosi. 2007. State and Trends of the Carbon Market 2007. Washington, D.C.: World Bank Institute, World Bank. Larwanou, M., M. Abdoulaye, and C. Reij. 2006. Etude de la Regeneration Naturelle Assistee dans la Region de Zinder (Niger). Washington, D.C. :International ResourcesGroup. LUCOP (Lutte Contre la Pauvretd). 2005. Rapport D'evaluation de L'impact. Presentation et Analyse des Donnees du Suivi Intense. Campagne 97/98,98/99,99/00,00/01,01/02 et 02/03:Niamey: LUCOP. MDAiFAO (Ministere du Developpement Agricole and the Organisation des Nations Unies pour 1'Alimentation et 1'Agriculture. 2002. Plan d' Action National sur la Fertilite des Sols et la Collecte des Eaux de Ruissellement (PANSICER). Niamey, Niger, and Rome: MDAEAO. M D A P A C (Ministere du Developpement Agricole and Programme d'Actions Communautaires). Undated. Recueil des Fiches Techniques en Gestion des Ressources Naturelles et de Productions Agro-Sylvo-Pastorales.Niamey, Niger :MDAPAC. Ministere de l'Hydraulique, de 1'Environnement et de la Lutte Contre la Desertification. 2003. Strategie Nationale de Relance de la Production et de la Commercialisationde la Gomme Arabique au Niger. ftp://ftp.fao.orgldocrep/fao/OO6/y5033f/y5033f05 .pdf. (Accessed May 2008). Mortimore, M. 2005. Dryland development: Success stories from West Africa. Environment, JanuarylFebruary issue: 10-20. Mortimore, M.,M. Tiffen, Y. Boubacar, and J. Nelson. 2001. Synthesis of Long-Term Change in Maradi Department, Niger, 3960-2000. Drylands Research Working Paper No. 39e. Crewkerne, U.K.:Drylands Research. 83 Nkonya, E., J. Pender, C. Kaizzi, E. Kato, S. Mugarura, H. Ssali, and J. Muwonge. 2008. Linkages between land management, land degradation and poverty in sub-Saharan Africa: the case of Uganda. IFPRI Research Report No. 159. Washington, DC: International FoodPolicy Research Institute. Pender, J., and J. Ndjeunga. 2008. Assessing Impactsof SustainableLandManagementPrograms on Land Management and Poverty in Niger. Report I1(full study report). International FoodPolicy Research Institute, Washington, D.C. Mimeo. Pender, J., T. Abdoulaye, J. Ndjeunga, B. Gerard, and E.Kato. 2008. Impacts of Inventory Credit, Input Supply Shops, and Fertilizer Micro-Dosing in the Drylands of Niger. IFPRI Discussion Paper No. 763. Washington, D.C.: International Food Policy Research Institute. Reij, C. and D. Steeds. 2003. "Success Stories in Africa's Drylands: SupportingAdvocates and Answering Skeptics." Paper commissionedby the Global Mechanism of the Convention to Combat Desertification. http://www.etfrn.org/ETFRN/workshop/degradedlands/ documents/reij.doc (accessedMay 2008). RN (Republic of Niger). 2000. National Action Program for Combating Desert@cation and Natural Resources Management (PAN-LCD/GRN). Niamey, Niger: The Prime Minister's Office, National Environment Council for a Sustainable Development, and National EnvironmentProgramfor a SustainableDevelopment. . 2002. Poverty Reduction Strategy. Niamey, Niger: Office of the Prime Minister, PermanentSecretariat of the PRSP. . 2003. Strategie de Developpement Rural. Niamey, Niger : Cabinet du Premier Ministre, SecretariatPermanentde la SRP. Rosenbaum, P. R., and D. B. Rubin. 1983. "The Central Role of the Propensity Score in ObservationalStudies for Causal Effects." Biometrika 70 (1): 41-55. United Nations Development Programme (UNDP). 2006. Human Development Report 2006. htt~://hdr.undp.ordhdr2006/statistics/ UNFCCC (United Nations Focal Convention on Climate Change). 2008. Project Design Document Form for Afforestation and ReforestationProject Activities (CDM-AR-PDD), Version 03. Bonn: CleanDevelopment Mechanism(CDM) ExecutiveBoard, UNFCCC. Warren, A., S. Batterbury, and H. Osbahr. 2001. "Soil Erosion in the West African Sahel: A Review and an Application of a "Local Political Ecology" Approach in South West Niger." GlobalEnvironmental Change 11: 79-95. 84