IFPRI Discussion Paper 02133 August 2022 Quantitative Analysis to Inform Priorities for International Agricultural Research Keith Fuglie Keith Wiebe Steven Prager Timothy B. Sulser Nicola Cenacchi Camila Bonilla-Cedrez Dirk Willenbockel Environment and Production Technology Division INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE The International Food Policy Research Institute (IFPRI), a CGIAR Research Center established in 1975, provides research-based policy solutions to sustainably reduce poverty and end hunger and malnutrition. IFPRI’s strategic research aims to foster a climate-resilient and sustainable food supply; promote healthy diets and nutrition for all; build inclusive and efficient markets, trade systems, and food industries; transform agricultural and rural economies; and strengthen institutions and governance. Gender is integrated in all the Institute’s work. Partnerships, communications, capacity strengthening, and data and knowledge management are essential components to translate IFPRI’s research from action to impact. The Institute’s regional and country programs play a critical role in responding to demand for food policy research and in delivering holistic support for country-led development. IFPRI collaborates with partners around the world. AUTHORS Keith Fuglie (keith.fuglie@usda.gov) is a Senior Economist in the Economic Research Service, U.S. Department of Agriculture, Washington, DC. Keith Wiebe (k.wiebe@cgiar.org) is a Senior Research Fellow in the Environment and Production Technology Division at the International Food Policy Research Institute (IFPRI), Washington, DC. Steven Prager (steven.prager@gatesfoundation.org) is currently a Senior Program Officer in the Agricultural Development Program with the Bill and Melinda Gates Foundation, Seattle, WA. At the time of his involvement with this study, he was a Principal Scientist at the Alliance of Bioversity International and CIAT in Cali, Colombia. Timothy B. Sulser (t.sulser@cgiar.org) is a Senior Scientist in IFPRI’s Environment and Production Division, Washington, DC. Nicola Cenacchi (n.cenacchi@cgiar.org) is a Senior Research Analyst in IFPRI’s Environment and Production Technology Division, Washington, DC. Camila Bonilla-Cedrez (camila.bonillacedrez@wur.nl) is currently a Researcher in the Animal Production Systems Group at Wageningen University & Research in Wageningen, The Netherlands. At the time of her involvement in this study, she was a Postdoctoral Fellow at the Alliance of Bioversity International and CIAT in Cali, Colombia. Dirk Willenbockel (d.willenbockel@ids.ac.uk) is a Research Fellow at the Institute of Development Studies, University of Sussex, Brighton, UK. Notices 1 IFPRI Discussion Papers contain preliminary material and research results and are circulated in order to stimulate discussion and critical comment. They have not been subject to a formal external review via IFPRI’s Publications Review Committee. Any opinions stated herein are those of the author(s) and are not necessarily representative of or endorsed by IFPRI. 2 The boundaries and names shown and the designations used on the map(s) herein do not imply official endorsement or acceptance by the International Food Policy Research Institute (IFPRI) or its partners and contributors. 3 Copyright remains with the authors. The authors are free to proceed, without further IFPRI permission, to publish this paper, or any revised version of it, in outlets such as journals, books, and other publications. mailto:keith.fuglie@usda.gov mailto:k.wiebe@cgiar.org mailto:steven.prager@gatesfoundation.org mailto:t.sulser@cgiar.org mailto:n.cenacchi@cgiar.org mailto:camilla.bonillacedrez@wur.nl mailto:d.willenbockel@ids.ac.uk Table of Contents Abstract ........................................................................................................................................................ vi Acknowledgments and Disclaimer .............................................................................................................. vii List of Acronyms ......................................................................................................................................... viii Executive Summary ....................................................................................................................................... 1 Scope of Study ............................................................................................................................................ 10 Geographic scope of the analysis ........................................................................................................... 10 Technological scope of the analysis: agricultural productivity ............................................................... 11 Stages or levels of research resource allocation and prioritization: programs and projects ................. 12 Methodology ............................................................................................................................................... 13 Three questions for research prioritization ............................................................................................ 13 The parity or congruence model: Matching R&D resources to economic value .................................... 14 Weighted parity: prioritizing agriculture research for vulnerable populations ...................................... 15 Simulation model of research impacts ................................................................................................... 16 Results of the Parity Model ......................................................................................................................... 23 Value of agricultural commodities in all target countries and by region ............................................... 23 Using evidence of impact of past R&D to inform research priorities ..................................................... 25 Illustration: Using the parity model to allocate research funds among crop improvement programs .. 27 Economic importance of farming systems.............................................................................................. 29 Economic importance of crop and animal pests and diseases ............................................................... 33 Results of the Simulation Model ................................................................................................................. 38 Multi-dimensional impacts from increases in commodity productivity ................................................. 38 Nutritional impacts from increases in commodity productivity ............................................................. 42 Impacts from increases in aquaculture productivity .............................................................................. 45 Impacts of regional improvements in agricultural productivity ............................................................. 45 Impacts of improvements in natural resource management ................................................................. 48 Some Summary Observations ..................................................................................................................... 50 References .................................................................................................................................................. 53 Appendices .................................................................................................................................................. 56 Appendix 1: The Parity Model of Agricultural Research Resource Allocation ........................................ 56 Appendix 2. Regional Results of the Parity Model for Commodities ...................................................... 60 Appendix 3: Benefit-Cost Analysis of Crop Improvement Projects for Roots, Tubers & Bananas (RTB) 66 Appendix 4: Regional Results of the Simulation Model for Commodity Scenarios ................................ 67 List of Tables in the Executive Summary Table ES 1. Multi-dimensional impacts of increases in agricultural productivity .........................................7 Table ES 2. Agricultural productivity and nutritional change in target countries ........................................ 8 Table ES 3. Sub-Saharan African farming systems ranked by poverty-weighted agricultural value ........... 9 List of Tables in Body of Report Table 1. Summary of productivity investment scenarios ........................................................................... 17 Table 2. Natural Resource Management (NRM) productivity scenarios .................................................... 19 Table 3. Total and weighted values of commodities & commodity groups for all target countries .......... 24 Table 4. Evidence on impact from past agricultural R&D in developing countries .................................... 26 Table 5. Parity allocation of research funds to commodity programs ....................................................... 27 Table 6. Research resource allocation and project selection within the RTB program .............................. 28 Table 7. Some characteristics of Tier 1 farming systems in target countries ............................................. 30 Table 8. Total and weighted values of agricultural production in Tier 1 farming systems ......................... 31 Table 9. Total and weighted values of agricultural production in Tier 2 SSA farming systems .................. 33 Table 10. Total losses from pests & diseases for nine crops in target regions ........................................... 35 Table 11. Most significant crop pests in target area ................................................................................... 36 Table 12. Most significant crop pests by region ......................................................................................... 36 Table 13. Economic significance of 14 animal diseases in Sub-Saharan Africa and South Asia ................. 37 Table 14. Multi-dimensional impacts of increasing commodity productivity in the target countries ....... 41 Table 15. Agricultural productivity and nutritional change in target countries ......................................... 43 Table 16. Impacts of increasing productivity in aquaculture ...................................................................... 45 Table 17. Multi-dimensional impacts of increasing productivity in a region.............................................. 47 Table 18. Impacts of improvement natural resource management ........................................................... 49 List of Figures Figure 1. Global map showing countries and regions of interest ............................................................... 11 Figure 2. Prevalence of poverty and child stunting in target regions ......................................................... 16 Figure 3. Major global (Tier 1) farming systems ......................................................................................... 30 Figure 4. African (Tier 2) farming systems .................................................................................................. 32 Figure 5. R&D allocation parity rule for commodities ................................................................................ 41 Figure 6. Changes in RNI ratios from increases in commodity productivity ............................................... 44 Figure 7. Sources of calories in the average diet under commodity productivity scenarios ...................... 44 Figure 8. R&D allocation parity rule from regional increases in agricultural productivity ......................... 47 List of Appendix Tables and Figures Table A2.1. Commodity values and weighted commodity values for ESA ................................................. 60 Table A2.2. Commodity values and weighted commodity values for WCA ............................................... 61 Table A2.3. Commodity values and weighted commodity values for S ASIA ............................................. 62 Table A2.4. Commodity values and weighted commodity values for SE ASIA ........................................... 63 Table A2.5. Commodity values and weighted commodity values for LAC ................................................. 64 Table A2.6. Commodity values and weighted commodity values for CWANA ........................................... 65 Table A3.1. Ex ante benefit-cost analysis of Roots, Tuber & Banana (RTB) research projects ................... 66 Table A4.1. Multi-dimensional impacts of increasing agricultural productivity in ESA .............................. 67 Table A4.2. Multi-dimensional impacts of increasing agricultural productivity in WCA ............................ 68 Table A4.3. Multi-dimensional impacts of increasing agricultural productivity in S ASIA .......................... 69 Table A4.4. Multi-dimensional impacts of increasing agricultural productivity in SE ASIA ........................ 70 Table A4.5. Multi-dimensional impacts of increasing agricultural productivity in LAC .............................. 71 Table A4.6. Multi-dimensional impacts of increasing agricultural productivity in CWANA ....................... 72 vi Abstract Investors in international agricultural research seek sustainable agri-food technologies that can potentially serve multiple objectives, including economic growth, food security, and sustainable use of natural resources. We employ quantitative economic models to examine the potential multi- dimensional impacts of agricultural productivity gains in the Global South. These models take into account behavior responses to agricultural technological change, i.e., how productivity changes may affect decisions on what to produce, trade, and consume. We consider and compare potential impacts of productivity growth in different technologies and regions and assess implications along several impact dimensions, including economic and income growth, the population at risk of hunger, adequacy of micronutrients in human diets, land and water use, and greenhouse gas emissions. Evidence on the economic significance of major crop and farm animal pests and diseases is also summarized. Potential impacts of technologies that increase agricultural productivity vary widely by commodity, farming system and region. These results can help inform decision-making about an optimal R&D portfolio that takes into account the multiple objectives of agricultural R&D investments and illuminate potential tradeoffs among objectives that may result from different R&D spending decisions. Keywords: agricultural productivity; CGIAR; impact assessment; parity model; IMPACT model; farming systems; natural resource management; agricultural pests and diseases vii Acknowledgments and Disclaimer This work was supported by the U.S. Agency for International Development and the U.S. Department of Agriculture. The findings and conclusions in this report are those of the authors and should not be construed to represent any official USAID, USDA, or U.S. Government determination or policy. viii List of Acronyms AVRDC Asian Vegetable Research & Development Center CGIAR Global partnership that unites international organizations engaged in research about food security CIAT International Center for Tropical Agriculture CIMMYT International Maize & Wheat Improvement Center CIP International Potato Center CIRAD French Agricultural Research Center for International Development CoSAI Commission on Sustainable Agricultural Intensification CRP CGIAR Research Program CWANA Central, West Asia & North Africa DSSAT Decision Support Systems for Agrotechnology Transfer crop growth model ESA East & Southern Africa FAOSTAT Food and Agricultural Organization global agricultural statistics FPU Food Production Unit (IMPACT model unit of analysis) GBAD Global Burden of Animal Disease GBCL Global Burden of Crop Loss GDP Gross Domestic Product GHG Greenhouse Gasses GLDC Grain Legumes and Dryland Cereals GLOBE Global dynamic computer general equilibrium model ICARDA International Center for Agricultural Research in the Dry Areas ICRISAT International Crop Research Institute for the Semi-Arid Tropics ICWAS Crop Water Allocation and Stress Model IFPRI International Food Policy Research Institute IGHM Global Hydrology Model IITA International Institute for Tropical Agriculture IMPACT International Model for Policy Analysis of Agricultural Commodities and Trade IPM Integrated Pest Management IRRI International Rice Research Institute IWSM Water Basin Simulation Model LAC Latin America & Caribbean NARS National Agricultural Research Systems NOTILL Zero tillage NRM Natural Resource Management NUE Nitrogen Use Efficiency OIE World Organization for Animal Health R&D Research & Development RNI Recommended Nutrient Intake RTB Roots, Tubers and Bananas SASIA South Asia SDG1 Sustainable Development Goal 1: No Poverty SDG2 Sustainable Development Goal 2: Zero Hunger SEASIA Southeast Asia SPIA Standing Panel on Impact Assessment SSA Sub-Saharan Africa SSP2 Shared Socioeconomic Pathway 2: Middle-of-the-Road USAID United States Agency for International Development WCA West & Central Africa WUE Water Use Efficiency 1 Executive Summary This report provides a quantitative analysis to inform decisions about resource allocation in international agricultural research. It focuses on research aimed at raising agricultural productivity and examines how changes in productivity in different crops, regions, and farming systems are likely to affect economic growth, poverty reduction, human nutrition, and natural resource use at the global and regional levels. By quantifying the multidimensional impacts of productivity growth (for different commodities, regions, and farming systems), the report sheds light on how to move toward an efficient research portfolio that achieves high returns across multiple objectives. This has implications for USAID and other donor agencies regarding their R&D funding allocations to the CGIAR, Feed-the-Future Innovation Labs, and other mechanisms supporting international agricultural research. Decision-making for research resource allocation occurs at multiple levels, or stages. At one level, donor organizations may need to allocate funding across multiple Centers or programs, or choose specific projects to fund. If general funding is provided to Centers or programs, then a second level of decision making is for Center or program leaders to select the projects to support with available program funds. At each stage, judgements are made about technical and institutional feasibility and the likelihood for success and widescale impact. The quantitative assessments in this report are designed to assist decision-makers with these judgements, especially on the potential of commodity research to achieve widescale impact across multiple objectives. For a comprehensive approach to allocating resources for international agricultural research, we suggest a “staged” approach to decision-making. At the first stage, donor organizations could use relatively simple “parity rules” to allocate R&D resources among agricultural commodities at the global and/or regional levels. These parity rules can be adjusted to consider the multiple goals for R&D as well as give some considerations to the relative costs of developing and disseminating new technologies to smallholder farmers or other potential users. At a second stage, research program managers could lead more detailed exercises in ex ante benefit-cost analysis to guide project selection within commodity programs. At each stage, socioeconomic information (the parity rules and benefit-cost analysis) would bring demand-side considerations into research priority setting and technology design. The report gives illustrative examples of how this approach could be applied across CGIAR commodity research programs. A major component of this report focuses on the first, higher-level program stage of research resource allocation. The primary audience for this analysis is the donor community: the governments, international organizations, and private foundations that fund international agricultural research. The suggested guide for research resource allocation at the program level is the “parity rule.” The parity rule posits that funding should be allocated across research programs in proportion to the benefits anticipated from the technologies developed by those programs. Of course, the goals for R&D are multidimensional – they include not only raising smallholder incomes, but also improving nutrition, reducing poverty, enhancing social inclusion, mitigating and adapting to climate change, and conserving natural resources. As a complement to the program focus, more detailed modeling helps address these multidimensional goals by employing more sophisticated socioeconomic and biophysical simulation models to derive estimates of how increases in agricultural productivity are likely to affect several of these dimensions. These simulations reveal potential tradeoffs among R&D objectives, depending on the R&D emphasis given various commodities or commodity groups. For example, productivity 2 improvement in certain commodities may have large impacts on income growth but little impact on conserving agricultural land use or reducing GHG emissions. If a donor’s objective is primarily focused on climate change mitigation, it could give relatively more of its R&D funding to commodities that reduce net emissions. If a donor’s objective is primarily to raise incomes or reduce poverty, then it might give relatively more attention to commodities that have the greatest impact on that objective. The value of this quantitative analysis is that it can help to clarify these choices and trade-offs in R&D funding decisions. The geographic focus for the international agricultural research considered in this report is a set of 103 low- and middle-income countries in Asia, Africa, and Latin America. Excluded are high-income countries in these continents and large or primarily temperate agricultural producers (namely, China, Brazil, and the Southern Cone countries of South America are excluded from the analysis). Results are presented for the whole set of 103 “target countries” as well as disaggregated among six regions (corresponding to the regional definitions used by the CGIAR in defining its programs). One set of results uses a parity model to suggest an allocation of R&D funding across commodity programs. The parity model uses the current value of commodity production as a proxy for future benefits from research to improve productivity of that commodity. The commodity value is also weighted by national prevalence of poverty and child stunting to emphasize commodities of greater importance to these populations. The parity model is simple to implement but relies on strong assumptions to infer an optimal or efficient allocation of R&D resources. A second set of results use an economic simulation model of the global agrifood economy. This model is considerably more complex but can make a more systematic assessment of how changes in agricultural productivity are likely to affect income, nutrition, and use of environmental resources. The economic simulation model looks forward to 2030, taking into account how changes in population, per capita income, and other factors are likely to affect the market demand for agricultural commodities. It also explicitly models how producer and consumer behavior might respond to changes in yields, prices, and market demand over the next decade. The economic simulation model provides a means to validate or refine the research allocations suggested by the simple parity model. However, neither the simulation model nor the parity model takes into account the cost of R&D needed to achieve a given increase in yield or productivity. Instead, the report draws on evidence from past returns to R&D investments by CGIAR centers to offer some judgements about the likelihood of progress in the future. At the present time, explicit consideration of research costs, technology adoption pathways, and research benefits is probably best done to inform research project selection within commodity programs. The research managers, scientists, and socio-economists working within commodity programs are more likely to possess the specialized knowledge and information necessary to make informed judgements about technology needs and opportunities and the costs of technology development and dissemination in specific national and regional contexts. Eventually, systematic evidence from ex ante and ex post benefit-cost analysis of research projects could be built up to provide more sophisticated benefit-cost analysis for R&D funding allocation at the program level as well. Table ES 1 provides a summary of key results of the multidimensional impacts of increasing agricultural productivity in the 103 target countries. The first column of figures shows the current value of production for six commodity groups and the aggregate value of these commodities by region. If the economic impacts of commodity research are strongly correlated with the current economic significance of the commodities, and if the R&D costs of raising productivity are roughly similar among commodities, 3 then the parity rule indicates an efficient R&D allocation rule that is likely to maximize economic returns across the commodity R&D portfolio. The poverty-weighted commodity values and value-shares (the 3rd an 4th column of figures in the table) give greater weight to commodities important to countries with a high prevalence of extreme poverty. The parity model applied to the poverty-weighted value, for example, gives greater importance to root and tuber crops, pulses and oilcrops at the expense of other commodity groups. This is because these crops are particularly important in Sub-Saharan African countries, where poverty rates are relatively high. Table ES 1 also summarizes results from the simulation model. The table shows multiple dimensions of impact associated with a 25% increase in productivity. A 25% increase in the productivity in fruits & vegetables is projected to generate the highest income gains, followed by productivity gains in cereal grains. The productivity gain in fruits & vegetables is projected to only make small gains on the other dimensions of impact, however. Productivity improvement in cereal grains is projected to have much larger impacts on decreasing the population at risk of hunger and reducing agricultural use of land, water, and GHG emissions. For each of these impact dimensions, the “parity rule” applied to the simulation results indicates how R&D might be allocated across commodity programs to maximize the impact of the R&D portfolio on that objective. A decision-maker wishing to tackle several of these objectives at once would need to balance the portfolio allocation accordingly. Aggregating the impacts of a 25% commodity productivity improvement by region shows that the largest income gains and GHG emissions reductions would likely occur in South Asia. Nutrient deficiency in human diets is recognized as a major cause of malnutrition with significant social and economic costs. Macro and micronutrient deficiencies have especially severe consequences for stunting growth in young children, which has both immediate and long-term consequences for human health and well-being. While Table ES 1 shows impacts of higher agricultural productivity on caloric adequacy, Table ES 2 extends this analysis to protein, iron and zinc. For zinc, the simulation model shows how productivity increases in the six commodity groups are likely to affect the population at risk of dietary inadequacy of this micronutrient. For protein, iron and other micronutrients, there is insufficient information to make this assessment. Instead, impacts are measured by changes in the “recommended nutrient intake (RNI) ratio.” An RNI of 1.00 means that the total nutrient availability in the diets of a population just equals the recommended nutrient intake for that population. To assure adequate nutrient availability for the entire population would require an RNI substantially greater than 1.00 to accommodate the unequal distribution of nutrients among individuals in the population. The simulation model indicates that raising cereal productivity would have by far the largest effect on improving RNI ratios for zinc, iron, and protein. This is due to the predominance of cereals in the diets of very poor and undernourished people, and thus the importance of grains as the primary source of these nutrients. Although overconsumption of carbohydrates and rising rates of obesity are growing problems in some countries, that issue might be better addressed by other public policies, food policy in particular, rather than by agricultural research policy. Agricultural research policy, on the other hand, can have major impacts on reducing undernutrition. Whether a parity rule suggests an optimal or efficient allocation of R&D funding depends in part on whether the costs of technological development and dissemination are roughly similar across commodities. One indicator of future returns to research is evidence on returns from past investments in R&D. The report summarizes findings from recent reviews of ex post impact of CGIAR commodity research. R&D for most crops has earned moderate to high rates of return. Exceptions are banana/plantain and ruminant livestock, where there is only limited evidence of R&D impact so far (one 4 exception, however, is forage crop improvement, which has earned moderately high returns through its impacts on productivity in beef and dairy). With a few exceptions, research investments in crop improvement appear to offer significant potential to raise productivity. Agricultural research programs may be structured not only along commodity lines but also around farming systems. Farming or food systems research is more integrative and aimed at adapting suites of technologies and practices to local needs. Table ES 3 illustrates how the parity model can be applied to guide R&D allocation among farming systems research programs. For 12 major farming systems in Sub- Saharan Africa, the table shows the total value and poverty-weighted value shares of farm production in that system, along with information on the total population, crop area, and number of ruminant livestock in each system. This information could be used to help select the priority farming systems for R&D investment and/or to guide the allocation of R&D resources among farming systems (for example, by using the poverty-weighted value share as a rule of thumb for its R&D budget share). Again, a second stage of priority setting employing benefit-cost analysis could be used to help select R&D projects within each farming systems research program. Research programs organized around commodities or farming systems are multidisciplinary and develop a range of technologies. Crop and animal breeding improve yield potential as well as yield stability by incorporating tolerance or resistance to abiotic and biotic stresses. Innovations in agronomic and husbandry practices assure quality crop and animal nutrition and efficient and sustainable use of natural resources. Improved management and control of pests and diseases and better on-farm harvesting and storage practices help protect against yield losses. Quantitative analysis of potential research impacts can provide critical information for efficient project selection within research programs. And, when combined with estimates of the research project costs and time needed for successful technology development and diffusion, such analysis can be used to help select the projects offering highest expected returns across multiple objectives (like those illustrated in Table ES 1). The report provides quantitative assessments of potential impacts of two classes of technologies – natural resource management (NRM) and integrated pest management (IPM) – that have relevance for international agricultural research. A key feature of these technologies is their focus on not only raising agricultural productivity but also enhancing environmental health and sustainability. Like new crop varieties or improved animal breeds, NRM and IPM need to be tailored to local conditions. For NRM, economic simulation models are used to estimate potential impacts in target countries of: i) No-till farming (NOTILL) using cover crops and mulching in maize and wheat to reduce soil erosion, improve soil water holding capacity, and increase soil carbon and organic matter; ii) Improved Nitrogen Use Efficiency (NUE), or the share of available soil nitrogen that is harvested in cereal grains, through innovations in fertilizers and fertilizer management, to enhance crop yield and reduce nutrient loadings on the environment; iii) Improved Water Use Efficiency (WUE), or the share of applied irrigation water that is used by crops, using improved irrigation technologies and water management methods to improve and stabilize crop yield and conserve water. The simulations find that successful adaption and adoption of NOTILL, NUE and WUE can significantly increase agricultural productivity, improve nutrition (by making food more affordable), and conserve natural resources. One issue facing agricultural technologies and NRM technologies in particular is the 5 Jevons Paradox: whether improved resource-use efficiency may actually lead to more use of the resource because the efficiency gains have made the resource cheaper. The economic simulation model, which takes into account behavioral changes that might give rise to the Jevons Paradox, finds that improvements in WUE in the target countries are likely to provide substantial water savings. However, improved NOTILL and NUE may slightly expand land area in cereal crops even as they reduce soil erosion and/or nutrient runoff and leaching. For IPM, the report draws on recent global or regional assessments of crop and animal losses from pests and diseases to identify those posing the greatest threat to smallholder producers in the target countries. For crops, expected annual loss estimates are derived for more than 100 pests and diseases afflicting 10 major crops – wheat, rice, maize, soybean, cassava, potato, sweet potato, yam, banana, and plantain. Combined losses from all of the pests and diseases assessed for these crops are estimated to exceed $113 billion per year in the target countries, with more than $65 billion per year in rice losses alone. The most economically significant crop pests in the target countries include rice bacterial blight, cassava mosaic disease/brown streak virus, and stem borers and sheath blight in rice. Rice losses dominate in South and Southeast Asia, losses in cassava, banana and maize are most significant in Sub- Saharan Africa, while maize losses are most significant in Latin America. For livestock, the loss assessments for 14 pests and diseases affecting cattle, goats, sheep and poultry in Sub-Saharan Africa and South Asia are estimated to cost farmers more than $20 billion per year due to animal mortality and morbidity. Endoparasites and ectoparasites affecting these species are the most economically significant, accounting for more than half of these economic losses. These results illustrate large potential impacts from R&D on agricultural productivity, food security and environmental sustainability. However, this information is insufficient to judge which specific R&D projects directed toward breeding, NRM, IPM, or other technologies are likely to generate the highest return. That also depends on the costs and time frame for technology development and dissemination. The NRM simulations, for example, assume that NOTILL, NUE and WUE technologies can be locally adapted and widely disseminated in the target countries over the next decade. However, ex post impact studies have not found much evidence of impact from past investments in international NRM research. IPM research, on the other hand, has a stronger record of achievement. Looking ahead, ex ante benefit-cost analysis is a powerful tool to improve R&D project selection within commodity or farming systems for international agricultural research programs. This report describes two examples of where benefit-cost analysis has been used in CGIAR crop improvement programs to guide project selection. One example is from ICRISAT, where, in the 1990s, agricultural scientists and economists conducted a benefit-cost analysis of more than 50 potential R&D projects in groundnut, sorghum, pulse crops and NRM to guide selection of the most promising projects. A second more recent example is from the CGIAR research program on Roots, Tubers, and Bananas (RTB). This interdisciplinary effort first conducted a global survey of key informants to identify the most important productivity constraints facing smallholder growers of cassava, yam, potato, sweet potato, banana and plantain across Asia, Africa, and Latin America. They then considered a number of potential solutions to these problems and selected 33 projects for more in-depth benefit-cost analysis. Again, this effort helped identify the R&D projects judged most likely to have the largest impact on productivity and livelihoods in these regions. Importantly, these analyses also established baselines against which to monitor and assess progress. Ex post impact assessments will reveal if productivity gains are falling short or exceeding expectations, which in turn can suggest adjustments and modifications to the R&D 6 investment portfolio. By using ex ante quantitative analysis in program-level and project-level R&D allocation and ex post impact assessment to assess progress, funders and supporters of international agricultural research can have greater confidence that resources are being put to their best possible use. 7 Table ES 1. Multi-dimensional impacts of increases in agricultural productivity* Current Gross Value of Production (2017-19 avg) Simulations of impacts of 25% yield gain (output/ha or output/animal) over 2010 levels in commodity group by 2030^ Production value Poverty-weighted production value Income Pop at risk of hunger (caloric adequacy) GHG emissions Land use Irrigation water use Commodity Value (b$) Parity rule (%) Value (b$) Parity rule (%) Value ∆ (b$) Parity rule (%) Value ∆ (mil. pop) Parity rule (%) Value ∆ (106T) Parity rule (%) Value ∆ (106 ha) Parity rule (%) Value ∆ (b m3) Parity rule (%) Cereals 312.4 24.6 48.3 22.3 340.2 21.6 -159.5 49.8 -132.5 33.0 -3.8 29.2 -1.2 41.4 RTB 124.6 9.8 35.7 16.5 124.7 7.9 -37.4 11.7 -17.0 4.2 -1.1 8.5 -0.4 13.0 Oilcrops 34.1 2.7 9.9 4.6 137.8 8.8 -37.8 11.8 -46.4 11.6 -4.3 33.1 1.6 0.0 Pulses 35.4 2.8 8.9 4.1 6.5 0.4 -36.0 11.2 -50.5 12.6 -2.7 20.8 -0.7 23.5 Fruit & Veg 318.5 25.1 49.9 23.0 684.7 43.6 -37.8 11.8 -3.1 0.8 -0.1 0.8 -0.6 22.1 Cash Crops 66.5 5.2 12.7 5.8 80.9 5.1 -11.8 3.7 -15.5 3.9 -1.0 7.7 0.1 0.0 Livestock 379.5 29.9 51.4 23.7 197.4 12.6 4.3 0.0 -135.9 33.9 1.5 0.0 0.1 0.0 SUM 1,271.0 100.0 216.8 100.0 1,572.2 100.0 -316.0 100.0 -400.9 100.0 -11.5 100.0 -1.1 100.0 Region Simulations of impacts of 25% yield gain over 2010 levels in all commodities by 2030 SASIA 334.9 32.3 65.1 39.1 771.1 49.7 -112.7 33.3 -148.3 33.5 -4.0 30.1 -2.7 66.2 SEASIA 198.0 19.1 5.9 3.6 204.2 13.2 -69.3 20.5 -86.3 19.5 -4.5 33.8 1.3 0.0 ESA 83.2 8.0 32.4 19.4 60.3 3.9 -37.7 11.1 -39.9 9.0 -0.7 5.3 -0.8 18.7 WCA 102.1 9.8 40.3 24.2 246.9 15.9 -61.4 18.1 -47.6 10.7 -2.1 15.8 -0.4 9.7 LAC 125.7 12.1 4.3 2.6 12.8 0.8 -21.5 6.4 -50.3 11.3 -0.9 6.8 -0.2 5.5 CWANA 193.0 18.6 18.6 11.1 256.3 16.5 -35.9 10.6 -70.9 16.0 -1.1 8.3 0.5 0.0 SUM 1,037.0 100.0 166.6 100.0 1,551.6 100.0 -338.5 100.0 -443.3 100.0 -13.3 100.0 -2.2 100.0 *The analysis focuses on 103 low- and middle-income countries ("target countries") located in six global regions (see Figure 1). Excluded are high-income countries, China, Brazil, and Southern Cone countries of South America. Impacts of productivity simulations on agriculture in the target countries produce economic impacts in these countries as well as world-wide through price and trade effects. The figures in the table only include impacts on the set of target countries. Current gross value of production is 2017-19 average annual quantities produced valued at 2015 prices derived from FAOSTAT. ^Value ∆ = change in impact value compared with the projected value in 2030. The simulations assume a 25% increase in yield over this reference yield. The 'parity rule' is the relative size of this impact compared with all other groups (it is the % of the sum of impacts across groups). RTB=roots, tubers, and bananas. Green-shaded cells show the highest desired outcomes; red-shaded cells indicate undesirable outcomes. 8 Table ES 2. Agricultural productivity and nutritional change in target countries Simulations of impacts on human dietary nutritional adequacy from 25% yield gain (output/ha or output/animal) in commodity group Zinc adequacy (Zinc) * Iron adequacy * Protein adequacy *^ Commodity Change in pop at risk (mil.) RNI ratio Change in RNI ratio (%) RNI ratio Change in RNI ratio (%) RNI ratio Change in RNI ratio (%) Reference scenario (no yield change) 1.04 0.83 2.24 Cereals -39.89 1.08 3.63 0.87 3.89 2.32 3.24 Roots & tubers -9.20 1.05 0.69 0.84 0.83 2.25 0.50 Oilcrops -9.49 1.05 0.77 0.84 0.71 2.25 0.59 Pulses -10.27 1.05 0.84 0.84 0.83 2.26 0.73 Fruit & vegetables -6.56 1.05 0.61 0.84 0.81 2.26 0.61 Cash Crops -3.93 1.05 0.35 0.84 0.46 2.25 0.26 Livestock -1.29 1.04 0.19 0.83 -0.07 2.25 0.24 * For nutrients other than zinc, the relationship between dietary nutrient availability and population at risk from dietary inadequacy is not well established. For many nutrients, dietary adequacy of a particular nutrient may be dependent on other factors, such as access to sanitation, clean water, and availability of complementary nutrients in the diet. The RNI ratio provides a measure of overall nutrient availability in the diets of a population. It is the ratio between average per capita nutrient availability in diets and the recommended nutrient intake (RNI). An RNI ratio of 1.00 implies that on average per capita nutrient availability just equals the recommended daily intake of that nutrient. For a population with an RNI ratio of 1.00, it is likely that half the population (those consuming below the mean) will have inadequate nutrient availability and half the population (those consuming above the mean) will have adequate availability of the nutrient in their diet. As RNI rises, a larger share of the population will experience adequate nutrient availability in their diet. ^Protein adequacy depends not only on the quantity but also the quality of protein. The simulation model only considers the quantity of protein availability in the diet. Green-shaded cells show the highest desired outcomes; red-shaded cells indicate undesirable outcomes. 9 Table ES 3. Sub-Saharan African* farming systems ranked by poverty-weighted agricultural value Farming system Value of agricultural production (m$/year) Poverty- weighted value share (%) Population in 2017 (millions) Total crop area (1000 ha) Number of cattle, goats & sheep (1000 SLU^) Major commodities** (listed in order of production value) 1 Maize mixed 42,719 19.8 214 35,751 49,372 Beef, maize, poultry, cassava, vegetables, cow milk, tree fruits 2 Agropastoral 40,857 18.3 176 52,311 87,684 Beef, vegetables, goat & sheep products, groundnut, maize, cow milk, sorghum 3 Cereal-root crop 43,654 17.6 98 41,838 45,685 Yam, vegetables, cassava, beef, rice, groundnut, cow milk, maize 4 Root & tuber crop 26,524 11.7 131 22,729 13,372 Cassava, yam, banana, tree fruits, maize, rice, cocoa 5 Tree-crop based humid 25,993 8.9 107 20,407 4,911 Cassava, yam, cocoa, banana, vegetables, tree fruits, rice 6 Highland perennial 17,570 8.4 99 14,590 25,181 Banana, beans, beef, maize, vegetables, cow milk, potato, coffee 7 Pastoral 11,391 4.0 53 11,873 52,393 goat & sheep products, beef, cow milk, millet, vegetables, groundnut, sorghum 8 Highland mixed 12,112 3.8 70 12,185 36,812 Beef, teff, vegetables, cow milk, maize, wheat 9 Fish-based 9,116 3.8 82 6,894 7,505 Cassava, tree fruits, banana, vegetables, beef, poultry, rice, yam 10 Large-scale irrigated 5,477 2.6 26 4,559 13,680 Rice, goat & sheep products, vegetables, tree fruits 11 Perennial mixed 4,434 1.0 16 1,043 4,469 Tree fruits, beef, cow milk, sugarcane 12 Arid pastoral oasis 737 0.1 5 150 3,705 goat & sheep products, beef, maize, cow milk, poultry, sesame, sorghum Total 240,584 100 1,077 224,328 344,770 *Sub-Saharan Africa combines the West & Central Africa (WCA) and East & Southern Africa (ESA) regions. ^SLU=standard livestock units for Africa (cattle weight=1.0; goat & sheep weight = 0.2). **Goat and sheep products include meat, milk, and wool. 10 Quantitative Analysis to Inform Priorities for International Agricultural Research Scope of Study The purpose of the report is to provide quantitative analysis to inform the allocation of financial resources to international agricultural research. This report presents statistical information and modeling results on the economic significance of agricultural commodities, farming systems, and agricultural pests and diseases in selected regions of the world. This information can assist in prioritizing investments in international and national agricultural research by identifying the commodities and productions systems where productivity improvement has the greatest potential to generate economic growth, reduce poverty, lower undernutrition, conserve natural resources and achieve other social goals. The quantitative analyses in this report are drawn from two analytical approaches. The first approach is the parity model. This model identifies the commodities and farming systems of greatest economic value and suggests “parity” – where research expenditure share is equated with production value share – can provide a good rule of thumb for an efficient allocation of research resources. “Weighted parity” extends this approach to give greater weight to commodities or farming systems more important to countries with higher rates of poverty, child stunting, or food insecurity. The second approach uses simulations from a model of the global agri-food economy. Simulations examine how changes in agricultural productivity affect output, resource use, and prices, and take into account how producers and consumers respond to these changes. It provides quantitative assessments on how agricultural productivity growth in various commodities and farming systems could impact incomes, poverty, undernutrition, and natural resource use in developing countries in the coming decade. The second modeling approach provides a means of validating (or contradicting) the relatively simple framework of the parity model. Geographic scope of the analysis The focus of this analysis is to inform investment in international agricultural research by USAID and other donors, particularly in the commodity improvement programs of the CGIAR and related national agricultural research systems. The quantitative assessment covers the countries and regions highlighted in Figure 1. These regions consist of East and Southern Africa (ECA), West and Central Africa (WCA), South Asia (SA), Southeast Asia (SEA), West, Central Asia, and North Africa (CWANA), and Latin America and the Caribbean (LAC) minus Brazil and the southern cone countries of South America. China, Brazil, and Southern Cone countries, although they have historically been significant beneficiaries of international agricultural research, are excluded from the assessment (along with higher-income CWANA countries) because these countries possess strong national agricultural research systems, have achieved upper-middle income status, and/or lie outside tropical zones that have been a primary focus of CGIAR agricultural research programs. 11 Figure 1. Global map showing countries and regions of interest LAC = Latin America & Caribbean; CWANA = Central, West Asia & North Africa; WCA = Western & Central Africa; ESA = Eastern & Southern Africa; S Asia = South Asia; SE Asia = Southeast Asia & Pacific. The gray-hatched areas are excluded from the quantitative assessment but are part of the simulation framework as documented in Robinson et al. 2015. For some measures, ESA and WCA countries are grouped together into Sub-Saharan Africa (SSA). Technological scope of the analysis: agricultural productivity This report focuses on research designed to increase agricultural productivity. This includes a range of technologies and scientific disciplines, including genetic improvement (breeding), soil and water management, crop and animal nutrition, control of pests and diseases, and integration of crop and livestock farming. One key dimension of research resource allocation is by commodity – determining how to allocate research resources among crop, livestock, and farmed fish commodities in order to have the largest impact on human welfare. A second dimension is by farming system, a geographic area defined by its agro-ecological conditions and commodity set. Research investments in farming systems complement, and help integrate, technologies emerging from commodity improvement programs. Farming systems research usually includes substantial attention to crop and animal husbandry practices to enhance stocks of natural capital (i.e., preserve soil, water, and biodiversity). A third set of analyses examines technologies specifically designed to conserve and enhance the quality of natural resources – specifically, the report gives attention to no till or conservation agriculture, improvements in nitrogen use efficiency, and increased water use efficiency in irrigation. A fourth dimension is management of biotic stresses or threats. Biotic stresses are the insects, microorganisms, and weeds that inflict or threaten significant economic losses in crop and animal production. The quantitative assessments included in this report provide guides for allocating research resources along each of these dimensions. One area not considered in this report is management of abiotic stresses. Abiotic threats to agriculture are those posed by environmental conditions and weather events – saline, acidic or otherwise problem soils, droughts, floods, heat waves, and frosts. Abiotic stresses are likely to grow in importance with climate change and increased frequency of extreme weather events. This is an important omission that should be addressed in future quantitative assessments of agricultural research priorities, especially when considering likely demands for new technology beyond 2030 or 2040. 12 Stages or levels of research resource allocation and prioritization: programs and projects Decisions about what research to fund and how much to fund it are made at various stages of the R&D process by different sets of decision makers. Two general stages or levels of R&D resource allocation are program-level and project-level. Program-level research refers to a long-term investment in a broad area of research and which has general goals. Each program consists of multiple projects, usually of fixed duration, which have specific targets or objections. Periodically, old projects are terminated and new projects begun, but programs continue over a longer time horizon. The CGIAR system is comprised of a number of research initiatives and centers, some of which focus on a commodity or related group of commodities. In fact, R&D for commodity improvement is probably the most natural ‘program’ level designation for agricultural research. A commodity improvement program will include scientists from multiple disciplines working on disciplinary or multi-disciplinary projects. Programs may also be defined around farming systems (or other geographically defined areas), resources areas (e.g., soil, water, labor), stages of a value chain (e.g., production, post-harvest processing, utilization), or social objective (e.g., research on policy, nutrition, gender equity). USAID-funded university Innovation Labs provide another example of program orientation of agricultural research. Decisions about research resource allocation at the program and project levels are typically made by different sets of decision makers. For the CGIAR, program-level funding decisions have historically been made by the system’s major donors. Project selection was then left to the research managers and scientists leading the programs. Quantitative modeling can help inform and guide decisions about program-level funding and project selection. Tools like the parity model, economic simulation models, and ex ante benefit-cost analysis are complementary approaches that help inform both levels of decision making. The analytical and data requirements are quite different, however. The parity model is the simplest to construct, use and explain. Economic simulation models are considerably more complex but can provide greater rigor and insights across multiple types of research impacts. Benefit-cost analysis includes not only an assessment of research impacts but also requires explicit assessments of the R&D costs and time frame for technology development and dissemination. Given these data and analytical requirements, benefit-cost analysis may be a more appropriate tool at present for project-level selection rather than program level allocation. For informing program level decision making, this report draws on the parity and simulation models to compare potential impacts of different research programs. To assist decision makers with making judgements about likely R&D costs or returns, the report synthesizes evidence from past returns to CGIAR research as a guide to where we might expect future returns to R&D to be high or low. Thus, the analyses presented in this report are directed primarily to donor organizations to inform program-level funding allocations. This analysis can be especially helpful in helping like-minded donors come to agreement on allocating pooled R&D resources across programs. In the CGIAR system, this would be tantamount to giving greater discretion for project selection within programs to leading scientists and research managers of those programs. However, donors could also encourage or require program managers to undertake benefit-cost analysis to help guide project selection within their programs. The report contains an illustration of how program-level and project-level quantitative assessments could be used together to allocate funding across programs and select projects within programs. This illustration draws on two examples of programs that have used ex ante benefit-cost assessments for project selection: the CGIAR programs on Roots, Tubers, and Bananas (RTB) and Grain Legume & Dryland Cereals (GLDC). 13 Methodology This section lays out the approaches used for the quantitative assessment of agricultural research priorities. It first describes three principal questions that need to be answered for an efficient allocation of research resources. It then describes the principles behind the parity model, and a variant, the “weighted parity” model, as a basis for allocating research resources across research programs. The parity model is simple to understand and straightforward to implement but involves a number of simplifying assumptions. A second modeling approach uses a simulation model to estimate how increases in agricultural productivity in target regions would likely affect incomes, poverty levels, nutrition, and natural resources use in the future. This approach serves to validate and extend the analysis from the parity model. While neither approach conducts a full benefit-cost analysis of program research expenditures (the focus is primarily on potential benefits and not on anticipated research costs), it could be combined with ex ante benefit-cost analysis (or other scoring approaches) to inform selection of the best projects within each research program. Three questions for research prioritization In his seminal book on Agricultural Research Policy, Ruttan (1982, p. 263) identified the following two questions that need to be addressed in order to allocate research resources efficiently: 1. What are the possibilities of advancing knowledge or technology if research resources are allocated to a particular commodity, problem, or discipline? and 2. What will be the value to society of the new knowledge or the new technology if the research effort is successful? Ruttan (1982) argued that answers to the first question – on the potential supply of new technology - can only be answered with any degree of authority by scientists who are on the leading edge of the research problem being considered. Further, answers to the second question – on the demand for new technology - require formal socio-economic analysis. The intuitive insights of research managers and scientists are no more reliable in answering questions of the societal value of research than the intuitive insights of research planners are in evaluating scientific or technical potential. Thus, research priority assessment is best addressed through multi-disciplinary perspectives. The quantitative analysis in this report focuses on Question 2, the potential socio-economic value of advancing agricultural productivity in developing countries. Important insights regarding potential returns to research can be gleaned from evidence on past returns to similar kinds of research. What has worked (or failed) in the past may well perform similarly in the future. We supplement the quantitative analysis on potential value of research with evidence on economic returns from past investments in CGIAR agricultural research and discuss implications for efficient research resource allocation. To Ruttan’s two questions above we add a third, which any funder or performer of public research must also assess: 3. Is there a compelling need or unique role for public international agricultural institutions to undertake this type of research? 14 In the broader innovation system, there are many organizations conducting R&D and providing innovations. National research institutions and universities, private companies, non-profit organizations, and farmers themselves produce innovations that can be widely applied locally, nationally, and beyond. For public-sector organizations, identifying research areas where “market failure” limits incentives for private R&D is one justification for public R&D investment in that area. Market failures arise when private research performers cannot adequately profit from their R&D investments, due to lack of protection for intellectual property rights or inability of farmers to afford or take on risks associated with adopting technological innovations. International organizations like the CGIAR often prioritize generating “international public goods,” or new knowledge and technologies that have wide applicability across national boundaries. Question 3 is about making sure that R&D investments by one’s own institution do not duplicate or crowd out R&D by other players in the innovation system. The parity or congruence model: Matching R&D resources to economic value The principal way that agricultural research improves public welfare is by increasing agricultural productivity. For lower income countries, sustained growth in agricultural productivity not only improves food security but also has profound effects on economic development. By making food more abundant at a lower price, agricultural research impacts the material and nutritional welfare of virtually everyone in the country and gives impetus to the process of structural transformation of national economies. The highest economic returns to agricultural research will likely be for technologies that have wide potential use. Research prioritization is about identifying these high impact areas. It stands to reason that commodities, production systems, and problem areas with the greatest economic significance should generally receive greater attention from (public or private) research or other investments. For example, the “parity rule” suggests that research budget should be allocated across commodities in proportion to their value share of production (Alston, Norton and Pardey 1995). While there are many reasons why such a simple allocation rule might not lead to the best use of research resources, it does provide an objective basis or starting point for discussions about research prioritization. Departures from the parity rule should have clear justification (below, we discuss how the parity rule can be amended into a “weighted parity” rule to take into account emphasis on targeting agricultural research to reduce poverty and child undernutrition). See Appendix 1 for a formal description of the parity model and how this is related to benefit-cost analysis of research. Determining the parity allocation involves estimating economic values and value shares. For commodities, the gross value of production is given by its annual harvest (in metric tons) multiplied by its farm-gate price (in $/ton). For this analysis, annual production (from FAOSTAT) in the target countries is averaged over 2017-2019. Prices are global averages from 2014-2016 measured in purchasing-power- parity dollars (also from FAOSTAT). Gross values are measured in millions of 2015$. Building on Wiebe et al. (2021), who developed parity shares for 20 food crops for which CGIAR Centers conduct breeding, the present analysis extends this to include most major crop, animal and fish commodities produced predominantly by smallholder farms in the target regions. Crops produced primarily through plantation systems or contract farming (e.g., rubber, sugar cane, tobacco) are excluded. Farming systems are where agricultural commodities, technologies, natural and human resources come together to produce livelihoods. Farming systems stand at the base of the agri-food value chain and the broader food system. Much of the agronomic, socioeconomic, and integrative research of the CGIAR and 15 national agricultural research systems (NARS), often in close alignment with extension programs, is organized around farming systems. Several CGIAR centers (CIAT, IITA, ICRISAT, and ICARDA) have mandates for specific farming systems. Farming systems research is often organized as a separate program from commodity research, but with strong complementarities to it. Thus, in addition to commodities, farming systems provide another useful framework for prioritizing agricultural research. To determine the parity allocation rule across farming systems, we estimate the gross value of production of crops and animal commodities for major farming systems in the target regions. This is achieved using spatially-disaggregated estimates of crop and animal production that are then aggregated up using farming systems boundaries from Dixon et al. (2021). For animal production, spatially-disaggregated data are for animal populations on farms. The output of meat, milk, eggs, and wool from animal populations are derived by multiplying animal numbers by the national average animal production (output per animal held in stock). Spatially-disaggregated data on crops and animals are available for all developing countries for 2010 and are aggregated into seven major farming systems (IFPRI 2019). For Africa, data are available for 2017, and are aggregated into a more refined set of 12 farming systems (IFPRI 2020). Again, output quantities are valued using 2015 prices. The study also provides estimates of the economic significance of major crop and animal pests and diseases. Such estimates are valuable for project selection within commodity research programs. Ideally, estimates of losses, or potential benefits from pest and disease control, should be coupled with judgements about the costs and likelihood of success of developing and disseminating the control technologies and practices. Then, ex ante benefit-cost analysis could form the basis for selecting which projects to prioritize for funding within a program area. Estimates of average yield losses from pests and diseases for five crops (rice, wheat, maize, soybean, and potato) are from a global survey by Savary et al. (2019), supplemented with estimates of crop losses from pests & disease of cassava, yam, potato, sweet potato, and banana/plantain from research priority assessments conducted by the CGIAR Research Program for Roots, Tubers, and Bananas. These estimates of crop losses (% of yield) are multiplied by the gross production value to get an estimate of the economic value of losses for each pest and disease. Weighted parity: prioritizing agriculture research for vulnerable populations The analysis in this report extends the simple parity rule to give explicit attention to the Sustainable Development Goals of alleviating poverty (SDG1 - no poverty) and undernutrition (SDG2 - zero hunger). It achieves this by employing a weighting scheme (‘weighted parity’) that emphasizes those commodities and farming systems most important to regions where extreme poverty and child stunting are most prevalent. To focus priority on poverty alleviation, commodity values in a country are weighted by the prevalence of poverty, measured by the poverty headcount indexes at $1.9/capita/year. Poverty-weighted values of production give greater emphasis to commodities important in countries and regions with greater prevalence of poverty. Similarly, to focus on undernutrition and hunger, commodity values are weighted by the prevalence of child stunting in a country. Child stunting (low height for age) for children five and under is used as the indicator for undernutrition. Stunting is indicative of chronic undernutrition and in children can be the cause of permanent mental and physical impairment. Thus, each commodity, farming system, and crop pest loss is given three values: (i) gross value; (ii) gross value weighted by poverty prevalence; and (iii) gross value weighted by the prevalence of child stunting. 16 Figure 2 summarizes these poverty and undernutrition indicators by target region. One implication of these measures is that focusing on SDG1 (poverty) will give greater emphasis to commodities, farming systems, and agricultural pests important to Sub-Saharan Africa (WCA and ESA), where poverty incidence is highest. Focusing on SDG2 (using child stunting as the hunger indicator) will shift some of this emphasis to South and Southeast Asia, where stunting remains high even as poverty rates have declined in recent years relative to Africa. It turns out that a focus on stunting results in similar value shares as the unweighted value shares because stunting is more uniformly distributed across the target regions than extreme poverty. Thus, the more relevant comparison is between the unweighted value shares and the poverty-weighted values. Figure 2. Prevalence of poverty and child stunting in target regions WCA=West & Central Africa; ESA = East & Southern Africa; SA = South Asia, SEA = Southeast Asia, LAC = Latin America & Caribbean, CWANA = Central & West Asia and North Africa. Source: World Development Indicators. Simulation model of research impacts Simulating effects of productivity growth on markets and households Changes in agricultural productivity can directly and indirectly affect incomes, poverty levels, nutrition, as well as use of natural resources. Impacts will depend on the magnitude of the change, where the change takes place, and which commodities are targeted. Testing alternative investment options can provide insight as to the relative size of their impact, as well as potential complementarities, trade-offs, or reinforcement effects, and therefore represents a useful tool to support decisions on how to allocate research funds. For this study, the International Model for Policy Analysis of Agricultural Commodities and Trade (IMPACT) was used to explore alternative future scenarios. The IMPACT model is actually a system of models based on a partial equilibrium, multimarket economic model that simulates national and international agricultural markets. Because an increase in agricultural productivity can lead to lower local and global commodity and food prices, it affects behavior and incomes of not only producers but also consumers. Moreover, productivity increases in one country or region can affect welfare in other regions through changes prices and trade. The IMPACT model accounts for how famers and households 17 alter their decisions on what to produce and consume, and how commodities are utilized and traded as incomes and prices change. Links to climate models, water models, and crop models support the integrated study of changing environmental, biophysical, and socioeconomic trends, allowing for in- depth analysis of a variety of critical issues of interest to policymakers at national, regional, and global levels. More information can be found at https://www.ifpri.org/project/ifpri-impact-model, and from the main documentation of the model (Robinson et al. 2015). Three sets of scenarios were explored in this study, focusing on increased productivity for different commodity groups, increased productivity for different regions, and improved natural resource management (Table 1). Table 1. Summary of productivity investment scenarios Scenarios Targets (crops/regions/techs) Scenario description Scenario names Commodity productivity scenarios Cereals Increase yield in one commodity group in target countries by 25% over 2010 level, while holding yields of other commodity groups and in non-target countries unchanged Cereals+ Oils OLS+ Pulses Pulses+ Roots and tubers RT+ Fruit and vegetables FV+ Other crops (Cash crops) Cash+ Livestock Livestock+ Regional productivity scenarios CWANA Increase yield in all commodity groups in a region by 25% over 2010 levels, while holding yields in other regions unchanged CWANA+ ESA ESA+ LAC LAC+ SA SA+ SEA SEA+ WCA WCA+ Natural resource management (NRM) productivity scenarios Nitrogen use efficiency Increase the share of soil N recovered in harvest of grain in 37.5% of rice, wheat, and maize area in target countries NUE Zero tillage, soil cover & crop rotation Increase the adoption of no till from negligible levels to 35% of wheat and maize crop area in target countries NOTILL Water use efficiency Increase the share of applied irrigation water utilized by crops in target countries by 15% over reference level, at river basin level WUE Reference scenario Baseline with zero growth No productivity changes over 2010 level REF All scenarios target a selection of 103 low- and middle-income countries (see Figure 1). Climate conditions are held fixed at recent historical levels. Income and population changes in 2030 are based on the IPCC middle-of-the-road GDP and population scenario (SSP2) (O’Neill et al. 2014). The commodity productivity scenarios target one commodity group at a time (Table 1). The implementation in IMPACT is described for Cereals below and it is similarly applied to all other crop- groups:  In the 103 countries of analysis, we assume productivity-increasing investments such that cereal yields will increase by 25% between 2010 and 2030. Yields for the remaining crop groups in those same 103 countries, and for all crop groups in the rest of the world (55 countries), are held constant.  The goal of this step is to isolate and compare the impact of productivity growth in particular commodity groups in the target countries. However, final yields may differ from the target levels https://www.ifpri.org/project/ifpri-impact-model 18 due to endogenous effects captured within the model, mainly through price changes as markets adjust to different scenario conditions. Through this set-up we explore how investing in productivity of different commodity groups may affect production, income, hunger indicators, as well as environmental parameters (land and GHG emissions). Ultimately, we are interested in observing which commodity investments are more likely to lower hunger, improve income and favor environmental indicators, whether the same investment scenario may benefit all dimensions, and if not, which scenario performs better for which indicators. The regional productivity scenarios, from CWANA to WCA (Table 1), are built in the same general way but they ask a different question; they explore the impact of targeting investments to each of the six ONE-CGIAR target regions (CWANA, ESA, LAC, SA, SEA, WCA), one at a time. Each scenario takes all countries within a single region, and all the commodity groups in that region are assumed to experience a 25% increase in yields between 2010 and 2030. Yield growth rates for all commodities in all the other countries of the world are held constant. The natural resource management (NRM) scenarios each represent a basket of technologies designed to reduce or prevent natural resource degradation while raising agricultural efficiency and productivity. These scenarios are drawn from the work described in Rosegrant et al. (2014, 2017), where the goal was to explore the consequences from adoption of specific technologies and practices consistent with the principles of sustainable agricultural intensification. Here, we reproduce the same changes in productivity estimated by Rosegrant et al. (2014, 2017), although we scale back the projected 2050 adoption levels assumed in Rosegrant et al. to where they would be in 2030, to be consistent with the other scenarios. Water use efficiency (WUE) in irrigation is the proportion of applied water in irrigation systems that is used for crop growth. Improved water application technologies and management methods can significantly increase WUE. Rosegrant et al. (2017) developed simulations by assuming that development and adoption of WUE innovations could increase WUE at the river basin level by 15 percentage points by 2030.1 We apply this assumption to all irrigated cropland in the 103 target countries for the WUE scenario. Nitrogen use efficiency (NUE) is the share of available soil N (from all sources, including fertilizer) that is taken up in crop harvest. NUE in world cereal production is typically only about 50%, but can be improved through innovations in crop genetics, agronomic practices (proper amount, type, placement, and timing of fertilizer applications), and use of improved fertilizers and soil amendments. For a comparable amount of fertilizer application, higher NUE raises crop yields. By changing N efficiency and stress paraments in the DSSAT crop growth model, Rosegrant et al. (2014) estimated the impact of improved NUE on yields of rice, wheat, and maize in different regions of the world. To derive global impacts of NUE, Rosegrant et al. (2014) assumed a global adoption rate of NUE efficiency technologies of 75% in 2050. For our scenario, we scale back adoption to 37.5% in 2030. NOTILL is a set of soil and tillage management practices designed to reduce erosion and enhance soil quality, including soil water holding capacity. It includes zero tillage, continuous soil cover from crop residues and mulches, and crop rotations, though research is needed to adapt it profitably to specific 1 In IMPACT, basin efficiency is defined as the ratio of beneficial water depletion (i.e., crop evapotranspiration) to total irrigation water depletion at the FPU level (Robinson et al. 2015). 19 locations and cropping systems. Rosegrant et al. (2014) estimated potential yield impacts from successful NOTILL development in wheat and maize from DSSAT by simulating minimal soil disturbance during seeding. Like the case with NUE, we scale back the assumed 2050 adoption rate of 70% by assuming a 35% adoption rate in 2030. Table 2 summarizes the average yield impacts of NUE and NOTILL adoption in the 103 target countries, though yield impacts vary by region. By raising crop productivity, these technologies also affect food prices, nutrition, and land use. The analysis derives implications of the WUE, NEU and NOTILL scenarios for income growth, nutrient adequacy in diets, land, and water use, and GHG emissions from agriculture. More details on these metrics are provided in the following section. Table 2. Natural Resource Management (NRM) productivity scenarios NUE (Nitrogen Use Efficiency) Yield in 2030 (t/ha) Adoption area (%) Average yield (t/ha) REF NUE REF NUE (on 100% of area) Maize 2.41 3.82 0.0 37.5 2.94 Wheat 2.78 4.01 0.0 37.5 3.70 Rice 2.60 5.23 0.0 37.5 3.12 NOTILL (zero tillage, soil cover, crop rotations) Yield in 2030 (t/ha) Adoption area (%) Average yield (t/ha) REF NOTILL REF NOTILL (on 100% of area) Maize 2.41 4.07 0.0 35.0 2.99 Wheat 2.78 5.23 0.0 35.0 3.84 NUE and NOTILL are assumed to raise crop yield while reducing natural resource degradation. The yield impacts are averages for the entire target area of 103 countries. See Appendix X for region-specific yield assumptions. Source: Rosegrant et al. (2014). Multidimensional metrics of impacts: income, food security, micronutrients, natural resources and GHG emissions Income growth The IMPACT model is linked to the global dynamic computable general equilibrium model, GLOBE- Energy. The role of GLOBE within the framework of the project is to assess the macroeconomic income and welfare effects associated with the alternative scenarios (Willenbockel et al. 2018). GLOBE captures the effects of agricultural growth on the rest of the economy. In practice, the outputs from a first run of the scenarios in IMPACT are used as input into GLOBE. GLOBE then simulates the changes in GDP that may be expected when the productivity of agriculture increases, and the effects are transmitted to the wider economy. In turn, the changes in GDP act as exogenous inputs back into IMPACT, thereby affecting agricultural production, demand, and ultimately food security. Food security: Population at risk of hunger Estimates of population at risk of hunger is the main food security metric produced through IMPACT simulations. It is the share of a population consuming below a minimum caloric requirement. Estimation uses FAO methodology (FAO 2008), which is based on a strong empirical relationship between per capita 20 food availability and the share of undernourished within a population. The methodology postulates a distribution of per capita caloric intake around the mean per capita caloric availability and integrates this density function up to the minimum caloric requirement. This gives the population share consuming below the minimum requirement. Increases in mean per capita caloric availability shifts the distribution and reduces the estimated share of the population consuming below the minimum requirement. Micronutrient modeling Micronutrient modeling follows the nutrient accounting framework established in the IMPACT model described in Beach et al. (2019). The approach translates the GENuS database (Smith et al. 2016) into commodity-level nutrient content coefficients that represent average availability through per capita consumption. The core components of this accounting framework provide per capita nutrient availability at the country level (which can be aggregated up to larger geographies with population weights) and ratios of this availability to country-specific recommended nutrient intakes (RNI) across modeled scenarios. Data availability issues restrict our focus to protein, iron, and zinc. See Wiebe et al. (2021) for an assessment of how productivity growth in selected food crops in the Global South might affect RNI of the four nutrients included in this study (carbohydrates, protein, iron, and zinc) plus total fiber, phosphorus, potassium, folate, and vitamins A, B6, C and E. For this analysis we extend the nutrient availability numbers from the Beach et al. (2019) approach into an additional metric intended to be more easily interpretable for policymaking. We used data and models established by Wessells and Brown (2012) to construct an estimate of the share of a country’s population at risk of inadequate supply of dietary zinc2. The construction of zinc estimates from Supplementary Table S2 in Wessells and Brown (2012) provides a consistent model of population level availability of dietary zinc compared to physiological requirements (“% [of] mean physiological requirement”), which is comparable to the RNI ratios from the Beach et al. (2019) approach. We estimate an elasticity (-0.84) of the relationship between the Wessells and Brown’s (2012) “estimated % of population with inadequate zinc intake” in Supplementary Table S2 with IMPACT’s RNI ratio and use this to project the change in country level population at risk of inadequate dietary zinc intake. Natural resources: land and water In IMPACT, cropland is estimated as harvested area, which is to say total area planted and harvested within a year (which may include multiple harvests on the same land in a year). The total land supply over time is driven by exogenous trends on the availability of area for agriculture, as well as endogenous responses to changes in area demand, which in IMPACT is a function of changes in commodity prices (Robinson et al. 2015). Estimates of water use rely on the communication between the core IMPACT multimarket model and the IMPACT water basin simulation model (IWSM) and crop water allocation and stress model (ICWASM) (Robinson et al. 2015). Briefly, a global hydrology model (IGHM) simulates rainfall, evapotranspiration, and runoff in each basin. These hydrologic outputs are fed into IWSM, which manages water basin storage, and optimizes irrigation water distribution in a watershed. The information on irrigated water supply enters ICWASM, and the model provides the IMPACT multimarket model with water stress- 2 We also investigated the potential for developing this metric based on inadequate iron intake. The complexities of interacting co-determinants for iron adequacy for different segments of the population would make this a much more complicated model. 21 induced crop yield reductions for both irrigated and rainfed crops. In these steps the model keeps track of blue and green water use across rainfed and irrigated systems. Greenhouse gas emissions The focus of this analysis is on direct on-farm agricultural emissions and emissions from land use change and the ensuing loss of carbon storage in soils and forests. For each scenario, we estimate CO2 emissions from loss of forested area due to expansion of cropland, methane emissions (CH4) from rice cultivation and enteric emissions from livestock, and emissions of nitrous oxide (N2O) from application of manure and synthetic fertilizer to cropland. The results are reported in terms of change from the reference scenario. The calculations are based on methodology developed for the Rosegrant et al. (2017) report and for a recent IFPRI-led report for the Commission on Sustainable Agriculture Intensification (CoSAI) (Rosegrant et al. 2021). We provide some description of the process below and refer to the methodology sections of the two mentioned reports for additional details. Emissions are estimated by post-processing the outputs of the IMPACT model. To estimate N2O emissions we used the IPCC Tier 1 default factors for direct N2O emissions arising from mineral N fertilizer application to managed soils (0.01 kg N2O-N per kg N fertilizer applied) and to irrigated rice (0.003 kg N2O-N per kg N fertilizer applied). These factors were multiplied by the N fertilizer consumption projected in IMPACT for each country and each crop/commodity (see Appendix F and H in Rosegrant et al. 2017). It is important to note that our estimates exclude the indirect N2O emissions from nitrogen leaching and runoff, and from atmospheric nitrogen deposition. Estimates of CH4 emissions from rice cultivation derive from the combination of IPCC Tier 1 and 2 emission factors (as in Yan et al. 2009), with the crop yields projected by IMPACT. The CH4 emissions from rice production are first calculated for a unit of area and then multiplied by the rice production areas projected by IMPACT. To estimate CH4 emissions from ruminants we multiplied the animal numbers projected in IMPACT (both slaughtered cattle and dairy animals) by the per-head emission value obtained from the enteric fermentation section of FAOSTAT (see also Appendix F in Rosegrant et al. 2017). Finally, CO2 emissions were estimated from changes in land use driven by expansion (or contraction) of crop harvested area and pastureland. We used simulations that linked IMPACT and the Landshift model to derive the relationship between changes in crop area and forest area (see especially Rosegrant et al. 2017 and Schaldach et al. 2011). The estimated changes in land use driven by changes in area and livestock production were then combined with the Tier 1 GHG emissions coefficients for the relevant land use types to compute the estimated GHG emissions changes. All emissions are then converted to CO2 equivalents by multiplying the amount of GHG by the respective global warming potential. 22 Box: Modeling Productivity Scenarios in Aquaculture The aquaculture sector is a crucial and fast-growing component of future food systems across the globe. Farmed fish supply substantial amounts for income, employment, and nutrition for millions in many countries. The modeling of fish futures for this analysis was accomplished via the IMPACT-FISH model which is maintained and updated through a collaboration between the WorldFish Center and IFPRI (Chan et al. 2021; Kobayashi et al 2015; World Bank 2013). While managed apart from the main IMPACT model used in the rest of this report, IMPACT-FISH performs in much the same fashion as the main model though is limited in that it is not integrated with a global general equilibrium model for feedback from the full economy and only a subset of the metrics reported in the main model are available at this time. IMPACT-FISH model results are presented separately. The scenario of investments in fish sector productivity improvement was set up in the same spirit as the other commodity group improvement simulations detailed above (i.e., zero growth in all other sectors while investments in yield improvements for aquaculture fisheries achieve yield increases of 25% by 2030). This modeling serves as the source for any comments on the fisheries sector in this report. 23 Results of the Parity Model Value of agricultural commodities in all target countries and by region Table 3 summarizes the value of crop, livestock, and aquaculture commodities in the target countries and provides their value shares as well as values and value shares weighted by the prevalence of poverty (at the $1.90/capita/day poverty line) and stunting among children under six in the general population. The total value of agricultural production (excluding plantation crops) in the target countries was $1,347 billion, two-thirds of which was crop production. The poverty-weighted value of agricultural production was $224 billion, 74% of which was from crops. The poverty-weighted values can be interpreted as the part of agricultural output that is produced or consumed by individuals living in that condition (i.e., about 17% of the population in these countries live in extreme poverty and thus the poverty-weighted value of agricultural output is a similar share of total agricultural output). The increase in the crop value share when output is poverty-weighted (from 66% to 74%) reflects the increased importance of crops for populations in poverty since they are less able to afford more diverse diets. The stunting-weighted value of agricultural output is $40 billion. Since the population of stunted children under six constitutes about 2-3% of the total population of these countries, the stunting-weighted value of output is only about 2-3% of total agricultural value. The parity model suggests that the commodity value shares provide a good basis for an efficient allocation of research funding across commodity programs. Allocating research funding according to total value share would be consistent with an objective of maximizing agricultural growth so long as the technological opportunities for advancing productivity through research were roughly the same across commodities. Allocating research funding according to poverty-weighted or stunting-weighted value shares would be consistent with an objective of maximizing growth in the commodities most important to these target populations (or more precisely, to the commodities of greatest importance to the countries where poverty and stunting are most prevalent). A straightforward application of the total value shares for research resource allocation would suggest that 66% of commodity research funding would be allocated to crops research and 34% to livestock, forages, and fish research. Cereal grains would get 23.2%, roots, tubers, and bananas (RTB) 9.3%, oilseeds and pulses 5.2%, etc. An allocation based on poverty-weighted value shares would increase the crop share of commodity research to 74% and decrease the livestock, forages, and fish share to 26%. The share to cereal grains would decline somewhat to 21.8%, the share to RTB would substantially increase to 16.1%, and the share to oilseeds & pulses would increase to 7.6%. The shift in commodity R&D allocation away from livestock, forages & fish, and cereal grains to provide more emphasis to RTB, oilseeds & pulses reflects the greater importance of RTB and oilseeds & pulses to countries where poverty rates are higher, namely, in Sub-Saharan Africa. The share of commodity research allocated to cassava, yam, banana, millet, sorghum, groundnuts, beans, and pulses would increase using poverty- weighted value shares. These are all crops especially important to African smallholders. Using stunting- weighted value shares would result in a research portfolio in between the growth-maximizing and poverty-minimizing allocations. This is because the prevalence of stunting remains more widespread across these developing countries, while the prevalence of extreme poverty is more concentrated in Sub-Saharan Africa (see Figure 2). 24 Appendix 2 contains tables which provide the information in Table 3 for each of the six regions (ESA, WCA, SASIA, SEASIA, LAC, and CWANA). These tables provide useful guides for research funding allocation within these regions. Table 3. Total and weighted values of commodities & commodity groups for all target countries Total Poverty Stunting Total Poverty Stunting Cereal Grains 312,421 48,280 4,375 23.3 21.8 12.1 Rice 193,956 26,895 2,038 14.4 12.1 5.6 Wheat 51,236 7,414 519 3.8 3.3 1.4 Maize 45,184 8,325 1,153 3.4 3.8 3.2 Millet 9,191 2,999 351 0.7 1.4 1.0 Sorghum 7,923 2,371 291 0.6 1.1 0.8 Barley 4,930 275 22 0.4 0.1 0.1 Roots, Tubers & Bananas 124,611 35,717 8,259 9.3 16.1 22.9 Potato 31,953 5,379 500 2.4 2.4 1.4 Cassava 39,044 13,171 3,223 2.9 5.9 8.9 Yam 19,775 7,022 914 1.5 3.2 2.5 Sweet Potato 7,321 3,081 664 0.5 1.4 1.8 Taro (cocoyam) 956 144 1,695 0.1 0.1 4.7 Banana & Plantain (excluding plantation AAA) 25,563 6,920 1,262 1.9 3.1 3.5 Oilseeds & Pulses 69,474 16,825 1,998 5.2 7.6 5.5 Groundnut 19,812 5,604 747 1.5 2.5 2.1 Soybean 8,029 1,589 127 0.6 0.7 0.4 Sesame 6,280 1,623 205 0.5 0.7 0.6 Beans (phaseolus ) 17,061 3,645 531 1.3 1.6 1.5 Chickpea (cicer ) 9,044 1,767 89 0.7 0.8 0.2 Cowpea (vigna unguiculata ) 2,891 1,148 156 0.2 0.5 0.4 Pigeonpea (cajanus ) 3,996 1,037 118 0.3 0.5 0.3 Lentil (lens ) 1,766 289 14 0.1 0.1 0.0 Fababean (Vicia faba ) 593 123 11 0.0 0.1 0.0 Smallholder Cash Crops 66,453 12,670 1,439 4.9 5.7 4.0 Cotton 31,621 7,098 804 2.4 3.2 2.2 Coffee 14,075 1,800 212 1.0 0.8 0.6 Coconut 9,551 1,002 61 0.7 0.5 0.2 Cocoa 7,703 1,620 160 0.6 0.7 0.4 Cashew 3,504 1,149 202 0.3 0.5 0.6 Vegetables & Melons 158,177 26,212 4,326 11.8 11.8 12.0 Solanum (tomato, eggplant) 44,328 6,487 1,058 3.3 2.9 2.9 Allum (onion, shallot, garlic, leek) 25,624 4,359 760 1.9 2.0 2.1 Cucurbit (cucumber, pumpkin, melon) 12,445 1,305 248 0.9 0.6 0.7 Brassica (cabbage, cauliflower, broccoli) 7,093 1,358 198 0.5 0.6 0.5 Okra 9,024 2,322 344 0.7 1.0 1.0 Legume vegetables (green beans, peas) 3,691 518 93 0.3 0.2 0.3 Leafy vegetables (lettuce, spinach) 1,718 155 31 0.1 0.1 0.1 Vegetable, other 54,253 9,707 1,594 4.0 4.4 4.4 Tree Fruits 160,357 23,682 4,283 11.9 10.7 11.9 Citrus (oranges, tangerines, lemons, other) 72,402 10,287 1,942 5.4 4.6 5.4 Mango 30,937 6,121 969 2.3 2.8 2.7 Pineapple 9,394 1,332 246 0.7 0.6 0.7 Papaya 4,211 760 118 0.3 0.3 0.3 Other tropical fruit 11,581 1,031 253 0.9 0.5 0.7 Other temperate fruit 31,832 4,151 756 2.4 1.9 2.1 Cultivated Forage Crops 18,718 1,660 333 1.4 0.7 0.9 Livestock Products 379,483 51,413 9,737 28.2 23.2 27.0 Dairy (milk) 104,270 16,560 2,619 7.8 7.5 7.3 Cattle (meat) 80,945 14,033 2,261 6.0 6.3 6.3 Small ruminants (meat, milk, wool) 56,669 9,604 1,880 4.2 4.3 5.2 Poultry (meat & eggs) 110,764 8,877 2,347 8.2 4.0 6.5 Pigs 26,835 2,339 630 2.0 1.1 1.7 Aquaculture - Freshwater fish 53,897 4,973 1,338 4.0 2.2 3.7 Carp 10,972 1,789 318 0.8 0.8 0.9 Tilapia 4,812 325 131 0.4 0.1 0.4 Misc freshwater fish 7,471 863 208 0.6 0.4 0.6 Crustaceans 30,642 1,995 681 2.3 0.9 1.9 ALL CROPS 891,493 163,385 24,679 66.4 73.8 68.4 ALL LIVESTOCK, FORAGES & FISH 452,098 58,046 11,408 33.6 26.2 31.6 ALL AGRICULTURE 1,343,590 221,431 36,087 100.0 100.0 100.0 Commodity Production Value (million 2015 I$) Value Share (%) 25 Using evidence of impact of past R&D to inform research priorities The parity model is designed primarily as an indicator of the potential value of research investment if the research is successful at raising productivity (Ruttan’s Q2 – see above). What about the likelihood of research success? Ruttan suggested that scientists working at the leading edge of their disciplines are best situated to make informed judgements about the possibilities for research to advance the knowledge and technologies needed to address productivity constraints and other problems facing agriculture. Another important source of information that can help inform these judgements is evidence of impact from past investments in research. Socio-economists have conducted hundreds of ex post impact assessments of agricultural research projects (Alston et al. 2000) and a recent review by Fuglie and Echeverria (2021) aggregated the evidence from nearly 150 studies on adoption and impact of crop technologies originating from CGIAR centers since 1960. In particular, Fuglie and Echeverria (2021) quantify the cumulative impacts of CGIAR-related technology adoption by commodity. These estimates show wide variation in the impacts from past investments in CGIAR crop improvement programs, ranging from $638 billion for rice to $0.3 billion for banana and plantain. Part of these differences, of course, is due to different levels of R&D investment in these crops, but a large part is also likely due to differences in the potential of research to successfully develop technologies that can be widely and quickly disseminated to farmers. Table 4 summarizes evidence on impact of past agricultural R&D in developing countries. It also lists other relevant considerations for funding prioritization in international agricultural research, such as availability of alternative suppliers of technologies such as the private sector or other international institutions. For past CGIAR research, very high impacts (at least $100 billion since 1960) have been documented for rice, wheat, cassava and forage crops. High impacts ($50 -$100 billion) have been documented for maize and other dryland cereals, and medium-level impacts ($10-$50 billion) from research on potato, groundnuts, beans, and pulses. Outside of the CGIAR system, research by the private sector has achieved significant productivity impacts in developing countries on hybrid maize, cotton, and poultry. For other commodities, evidence on impact of past research is limited, either due to low research investment (which may itself reflect scientists’ judgements about poor prospects for success), failure of research to make significant technological gains, or lack of field studies to provide documentation of adoption and impact. Among crop commodities, there is limited evidence of research impact for yam, sweet potato, banana/plantain, and tree crops. For CGIAR research on livestock, natural resource management and policy, reviews commissioned by the CGIAR Standing Panel on Impact Assessment (SPIA) have noted that there are relatively few ex post impact studies in these areas documenting successful research outcomes (CGIAR SPIA 2006; Jutzi and Rich 2016; Renkow 2018). This reflects, in part, limitations in available methods for conducting credible impact studies. For example, it is often difficult to establish clear attribution of policy research to decisions taken by policy makers (Renkow 2018). Evidence on impact of previous research should be weighed together with expert judgement to assess prospects for advancing technology and knowledge through new research (Ruttan’s Q1). Analyses from the parity and simulation models provide quantitative analysis of impacts if the research is successful and technologies widely adopted (Ruttan’s Q2). Knowledge of research systems helps to establish whether there is a unique and unfulfilled role for donor and CGIAR investments in agricultural research for developing countries (Q3 of our essential questions for efficient research prioritization). Significant variation in answers to Q1 and Q3 can serve as justification for adjusting the parity or weighted-parity research funding allocation rules that inform Q2. 26 Table 4. Evidence on impact from past agricultural R&D in developing countries Commodity/Research Area Evidence on imp