Impacts of COVID-19 on People’s Food Security: Foundations for a more Resilient Food System Christophe Béné, Deborah Bakker, Mónica Juliana Chavarro, Brice Even, Jenny Melo, and Anne Sonneveld A report commissioned by CGIAR February 2021 DISCUSSION PAPER The CGIAR COVID-19 Hub provides a coordinated research response to the global pandemic threatening health systems worldwide, along with posing serious risks to food security; local businesses and national economies; and hard-fought progress by stakeholders at all levels towards the Sustainable Development Goals. Convening researchers, funders, and key stakeholders, the Hub focuses on supporting national response and recovery work across CGIAR research themes, harnessing knowledge for emergency response, recovery, and resilience. Learn more at www.a4nh.cgiar.org/covidhub AUTHORS Christophe Béné (c.bene@cgiar.org) is Principal Scientist in Sustainable Food Systems at the Alliance of Bioversity International and CIAT in Cali, Colombia Deborah Bakker (deborah.bakker@wur.nl) is a Research Trainee at Wageningen Economic Research, part of Wageningen University & Research (WUR) in Wageningen, the Netherlands. Mónica Juliana Chavarro (mjchavarror@gmail.com) is a Visiting Researcher at the Alliance of Bioversity International and CIAT in Cali, Colombia. Brice Even (b.even@cgiar.org) is a Sustainable Food System Specialist in the Food Environment & Consumer Behaviour team of the Alliance of Bioversity International and CIAT in Hanoi, Vietnam. Jenny Melo (jenny@labuenaempresa.com) is a Visiting Researcher at the Alliance of Bioversity International and CIAT in Cali, Colombia. Anne Sonneveld (anne.sonneveld@wur.nl) is a Research Trainee at Wageningen Economic Research, part of WUR in Wageningen, the Netherlands. mailto:c.bene@cgiar.org mailto:deborah.bakker@wur.nl mailto:mjchavarror@gmail.com mailto:b.even@cgiar.org mailto:jenny@labuenaempresa.com mailto:anne.sonneveld@wur.nl Contents Foreword iii Acknowledgements iv Acronyms v Executive Summary 1 Preliminary elements of a food system resilience research agenda 3 I. Introduction 6 1.1. Background and justification 6 1.2. General objective of the study and scope of the report 7 II. Analytical framework 8 2.1. Effects of COVID-19 on people life and food security and nutrition 8 2.2. Typology of impacts and affected actors 9 III. Quality of evidence 11 3.1. Knowledge elaboration 12 3.2. Quality of data 13 3.3. Linking the quality of evidence to the review process 13 IV. Key findings 16 4.1. Data analytics 16 4.2. Quality of the evidence and implications for the analysis 18 4.3. Emerging evidence on COVID-19 impacts 19 4.3.1. Loss of income and jobs 19 4.3.2. Clear but difficult-to-assess impact on food security 19 4.3.3. Expected impact on nutrition 21 4.3.4. Effect on different actors of the system 21 4.3.5. Mixed prices effects along the chain 22 4.4. Proposing a (more) holistic and dynamic assessment of COVID-19 22 4.4.1. Relative importance of COVID-19’s disruptions across the food system 23 4.4.2. Impact pathways of COVID-19 on food system actors 26 4.5. Macro-economic considerations 30 4.6. Some (still) open and uneasy questions 31 4.6.1. Changes in food prices 31 4.6.2. Who benefitted from COVID-19? 32 4.6.3. The specific case of the restaurant industry 34 4.6.4. Is COVID-19 really a global crisis? 35 4.6.5. Informal and... invisible 37 V. First steps toward rebuilding a (more) resilient food system 39 5.1. Elements of food system resilience 39 5.1.1. Identifying actors’ and value chains’ sources of vulnerability 39 5.1.2. Understanding actors’ responses to shocks 40 5.1.3. Testing and documenting what works and for whom (and where) 41 5.1.4. Social protection as a way to build people’s resilience 43 5.1.5. Avoiding false debates 44 5.1.6. Learning from the first responses put in place and their outcomes 45 VI. Synthesis and conclusion 51 6.1. Limitations of the assessment 51 6.2. Synthesis: the big picture after the first 12 months of COVID-19… 51 6.2.1. No global collapse of the system but a lot of suffering (for many) and some huge profits (for a few) 51 6.2.2. Not just economic but also physical hurdles 52 6.2.3. From convenience and proximity to ‘constrained choice’ 53 6.2.4. Some unknowns 54 6.3. Conclusions 55 References 57 Appendix 1: Detailed step-by-step methodology and analytical frameworks 71 Appendix 2: Literature on supply chain resilience 80 Tables Table 1. Typology of COVID-19 impacts and affected actors in the context of food systems 15 Table 2. The 3-level score system used to assess the two dimensions of data quality 20 Table 3. Geographical coverage of the review (by region) 25 Table 4. Number of documents referring to the different groups of actors affected by COVID 34 Table 5. The main issues affecting the food system actors as reported in documents 35 Table 6. The different subsectors/commodities discussed in the documents 38 Table 7. Various propositions to strengthen the resilience capacities of actors of food systems 62 Table 8. Examples of short-term responses put in place by different actors during the first few months of the pandemic, their observed or expected outcomes, and some potential recommendations for the near future 68 Figures Fig.1. The different elements to be considered in assessing the impact of COVID-19 on people’s food security in the wider context of food system. 13 Fig.2. The two-dimensional scheme used to assess and account for the quality of evidence of the documents incorporated in the mapping. 18 Fig.3. Decision-tree for the analysis level used in the review 22 Fig.4. PRISMA flowchart summarizing the protocol adopted in the review 23 Fig.5. Monthly number of documents published (among those identified in this review) 24 Fig.6a. Typology of the 337 documents included in the global assessment. 26 Fig.6b. Change in the proportions of types of documents posted or published over time. 26 Fig.7. Impact pathways of COVID-19 on food systems and their different actors based on a Sankey approach. 41 Fig.8. Growth in GDP as recorded in 2019 (left) and projected for 2020 (right) 45 Fig.9. (left) Annual World inflation rate over the period 1990-2020; (right) Inflation rate for 2020 worldwide 48 Fig.10. Groceries daily revenues in the US compared to 2019. 49 Fig.11. Comparison of year-over-year seated diners between 2020 and 2019 52 Maps and Boxes Map 1. Geographical coverage of the Review 25 Box 1. Social protection as a major response to mitigate COVID-19 economic disruptions 64 iii FOREWORD Food systems are the subject of intense interest from multiple perspectives. An important overall narrative is that food systems are failing to effectively include all people and provide healthy and sustainable diets. However, there is also the impressive way diverse food systems have evolved and continue to innovate to supply food to consumers worldwide. The COVID-19 pandemic was the origin of a global shock that affected food, health and socio-economic “systems,” leading to multiple supply- and demand-side disruptions to food systems. In the spirit of using crises to enable fundamental change, research on food and COVID-19 aspires to build food systems back better and achieve improved health, sustainability, inclusion plus resilience. In this context, the CGIAR’s COVID-19 Hub commissioned a systematic review of what relevant lessons were learned in 2020 and how these lessons can inform future food systems research. The first step in this learning process was to assess the actual impact of the COVID-19 pandemic on food security. This report presents the key findings of this assessment. Using the evidence available, Chris Béné and co-authors take a pragmatic approach in assessing the functional capacity of food system actors (producers, supply chain and market agents and consumers) to respond to the COVID-19 crisis. To frame the available evidence, two key concepts are used – food security and nutrition, and food environment. Starting from these concepts, the authors consider the conventional dimensions of food security – availability, access, utilization, and sustainability – which they complement with elements of people’s well-being, including domestic violence, agency and risks. The assessment then focuses on the vulnerability and responses of the different actors of the food system to COVID-19 disruptions, and what actions are considered to have improved – or not improved – the ability of the system to maintain food security. At this stage, the lessons learned are mostly from short-term reactive responses, with more limited medium- to longer-term recovery responses. The evidence for building back better actions and learning from COVID-19 to avoid the next (zoonotic) crisis are, at this early stage of science and research on COVID and its impacts, still somewhat experiential. While predicting longer-term food system transformation is not possible, the systematic framing of these initial lessons will be useful in guiding future research. Looking ahead, the authors conclude that resilience is only one property of food systems that needs to be considered with a longer-term focus on the primary outcomes of health, sustainability and inclusion desired from our food systems. As the CGIAR COVID-19 Hub, we look forward to comments and suggestions on this discussion paper from colleagues as we plan subsequent research into better future food systems. – John McDermott Director, CGIAR Research Program on Agriculture for Nutrition and Health Co-Chair, CGIAR COVID-19 Hub iv ACKNOWLEDGEMENTS The authors are grateful to Vincent Gitz (CIFOR/FTA), Coordinator of Working Group 4 of the CGIAR COVID-19 Research Hub, and to Alexandre Meybeck (CIFOR/FTA) for useful suggestions, as well as to the rest of the working group members who provided overall guidance and oversight: Delia Randolph (ILRI), Inge Brouwer (WUR), John McDermott (IFPRI/A4NH), Jean Balie (IRRI), Ruerd Ruben (WUR), Samuel Benin (IFPRI), Bram Govaerts (CIMMYT), Keith Wiebe (IFPRI), Brian King (Alliance Bioversity-CIAT), Ekaterina Krivonos (CGIAR System Organization), Victor López-Saavedra (CIMMYT), Izabella Koziell (IWMI/WLE), and Guy Hareau (CIP), as well as Frank Place (IFPRI/PIM), Coordinator of the CGIAR COVID-19 Hub’s Working Group 1. The authors thank Sandrine Dury (CIRAD), Alejandro Guarin (IIED), Gordon Prain (CIP), and Ruerd Ruben, who peer-reviewed the report. Janet Hodur (A4NH) helped with editing and layout of the final version. The work was supported by CGIAR. Suggested citation: Béné C., Bakker D., Chavarro Rodriguez M., Even B., Melo J., and Sonneveld A. (2021) Impacts of COVID-19 on people’s food security: foundations for a more resilient food system. Report prepared for the CGIAR COVID-19 Hub Working Group 4, CGIAR, 90p. v ACRONYMS CGE Computable general equilibrium CGIAR Consultative Group for International Agricultural Research ELCSA Latin American and Caribbean Food Security Scale FAO Food and Agriculture Organization of the United Nations FCAFH Food consumed away from home FECAH Food exclusively consumed at home FIES Food Insecurity Experience Scale FLW Food losses and waste GDP Gross Domestic Product HIC High-income country ILO International Labor Organization IMF International Monetary Fund LIC Low-income country LMIC Low- and middle-income country LSMS-ISA Living Standards Measurement Study – Integrated Surveys on Agriculture MSME Mini-, small-, or medium-enterprise NGO Nongovernmental organization OECD Organization for Economic Co-operation and Development PSPR Poverty and Shared Prosperity Report SME Small- or medium-enterprise UN United Nations UNU-WIDER United Nations University World Institute for Development Economics Research WHO World Health Organization vi Closed markets in West Bengal, India, in March 2020: Lockdowns implemented as COVID-19 spread around the world impacted all types of shops and retail outlets. Photo: Soumen Tarafder/Shutterstock 1 EXECUTIVE SUMMARY Background and justification of this report As part of the work implemented by CGIAR on COVID-19, the COVID-19 Research Hub Working Group 4 “Address food systems’ fragility and build back better” was tasked with implementing a global assessment of the impacts of COVID-19 on food systems and their actors, focusing specifically on the consequences that the pandemic had brought on the food security and nutrition of those who have been affected by the crisis. This includes formal and informal actors of the food supply chains (from producers to street vendors) as well as consumers, in both rural and urban environments. Building on this assessment, the task was then to draw on key principles of resilience in the context of humanitarian and food security crisis, to identify preliminary elements of a food system resilience research agenda. General approach and framework The assessment was based on a methodical mapping of the information available worldwide, collected with electronic search engines in four different languages (English, French, Spanish and Portuguese). Analytically, two main concepts were used to unpack and analyse the effects of COVID-19 on people’s food security and nutrition: the concept of food security per se and the concept of food environment. Several dimensions were then included in the analysis: food availability (supply); food access (affordability and physical accessibility); food utilization (quality and safety); stability; proximity; convenience; food waste and losses; and diversity of food items. In addition, elements of people’s wellbeing were considered, including agency and self-efficacy, prevalence of domestic violence, and increased risk of exposure to the virus. The quality of the evidence was assessed using two standard criteria: knowledge elaboration1 and quality of data, and the level of analytics applied to the data was adjusted to the quality of the information. Coverage and limits of the evaluation In total, more than 9,630 documents discussing the impact of COVID-19 on the food security of the different actors of food systems between January and December 2020 were identified, using a combination of keywords specifically chosen to address the objectives of the study. After removal of documents with low representativity and/or low reliability (mainly news media and personal social media reports), we were left with 337 documents covering 62 countries from Africa, Asia, Europe, Oceania and the Americas. Several limitations of the analysis should be mentioned. First, although great attention was paid to ensure the comprehensiveness of the identification process, it is difficult to achieve a perfectly exhaustive review. Consequently, some documents that would have been useful for the analysis might have been missed. Second, the majority of the 337 documents reviewed were material that was posted or published during the phase of the pandemic when it was difficult for researchers to operate in the field and to obtain direct primary data. As a consequence, the information made available through those documents is for a large part anecdotal or based on experiential knowledge. Even when more reliable and representative protocols were applied, the nature of the surveys used to generate data (telephone interviews) has led to a bias toward tangible, easily or quickly ‘measurable’ or quantifiable data/indicators to the detriment of 1 See definition in section 3.1. 2 more nuanced or qualitative data. Third, the analytical framework used for this study focuses essentially on food system actors and their direct (food) environment – a methodological choice induced by the primary objective of assessing the impact of COVID-19 on these actors’ food security and nutrition. As a consequence, the main entry point for the analysis is the individual level (actor, enterprise). This means that elements and processes important to consider in relation to the dynamics and/or the resilience of food systems but taking place at higher levels (e.g. drivers of food systems, institutional actors’ political agendas and priorities, local and national policies, etc.) have not been thoroughly explored. Initial key findings The review confirms what other analyses have also highlighted, namely, the magnitude and the severity of an unprecedented crisis that has spread worldwide and has spared only a few. But the review also reveals some other important elements. First it highlights that despite the attention that this global crisis received so far from the scientific community, we still have a relatively poor understanding (both quantitatively and qualitatively) of the actual impact of the pandemic on people’s food security and nutrition. This state of incomplete knowledge can be explained by the relatively short period of time since the pandemic began (meaning that only a small number of peer-reviewed, rigorous, research articles had been published by the time this review was conducted), and by the fact that research on the ground was severely constrained by the successions of lockdowns and mobility restrictions that have been imposed worldwide. Using the information available, the analysis reveals that the dimension of food security that has been most affected is accessibility, with reasonably solid evidence suggesting that both financial and physical access to food have been disrupted, in particular in urban areas and in low and middle income countries (LMICs). In contrast, there is no clear evidence that the availability of food has been affected beyond some initial disruptions due to panic buying; and there is not enough information to provide robust conclusions about the effects of the pandemic on the utilization of food (safety or quality). We note that those various disruptions in access (or even temporarily in availability) can be re-interpreted as disturbances in the stability dimension of the concept of food security, justifying the use of the concept of resilience in the second part of this report. Finally, the impact of COVID-19 on the nutritional status of people (so far conceptualized essentially as a consequence of the disruption in the economic accessibility to food on children), is still poorly documented but expected to be substantial in the long run. Beyond these direct effects, anecdotal accounts of degradation in people’s wellbeing were also found (especially in relation to domestic violence as well as voluntary or involuntary exposure to the virus), but the absence of detailed analyses in the documents available at the time of completing this review prevents more robust conclusions. COVID-19 impact pathways The impact pathway analysis that was built on these initial findings provides additional important insights. Of particular importance is the observation that contrary to what had been concluded in several other documents, the disruption in access to food due to people’s loss of employment or reduction in income/revenues is not limited to its financial component (affordability). Another important pathway that contributed to this outcome relates to the disruption in physical access to food outlets in urban context, especially during the time of complete lockdowns. This disruption in physical access was then shown to affect proximity and convenience, which, combined to the reduction in affordability induced by the decrease in people’s revenues, eventually led to a degradation in food choice and diversity. 3 Major conclusions Serious concerns had been initially expressed about the severe disruptions that the successive waves of lockdowns have induced on the food system actors and, more generally, on people’s livelihoods and local and global economies. The fears were that these disruptions may lead to local –or even global– food shortages. The evidence suggests that those fears –albeit justified– did not materialized. Overall, food systems ‘resisted’ the shock and no major episodes of severe food shortage were observed. This resilience of the food systems came, however, at great costs, with the majority of the systems’ actors having to cope with severe disruptions in their activities. At the same time, a group of actors was able to take advantage of the crisis; those are the grocery stores and supermarkets which made billions of dollars in profits in 2020, thus raising questions about the best way part of these profits could be redistributed or used to cover some the costs that the crisis inflicted. Overall, although the (short-term) capacity of food system actors to resist, adapt and innovate in the face of the economic challenges imposed by the lockdowns led some experts to emphasise the intrinsic resilience of the system, it should also be kept in mind that a large part of that resilience resulted simply from the special status of the larger actors as “essential services,” which allowed them to continue operating while many other economic sectors had to shut down. This apparent resilience was also built at the cost of hundreds of thousands of smaller or informal food system actors who disappeared during the crisis. The longer-term implications of the COVID-19 crisis for the dynamics and performances of the local and global food systems are difficult to predict. PRELIMINARY ELEMENTS OF A FOOD SYSTEM RESILIENCE RESEARCH AGENDA The various findings synthesized above have implications for both policy and research. Several lessons and propositions are distilled throughout the report and are synthesised below. First, the review reveals important gaps in our knowledge about resilience in relation to food systems. Several factors explain this situation, including the recognition that the concept of resilience is still very often used in a rhetorical manner in food system policies and very theoretically in the academic communities where it is discussed essentially in the context of high income countries. As such, these academic pieces are of limited use to guide research on the resilience of food systems and their actors in LMICs and very little is currently known about the different elements that would be necessary to strengthen the resilience of both the actors and the systems in the context of those LMICs. This report lays out some initial elements of a research agenda in that direction. Identifying actors’ and value chains’ vulnerabilities An initial task in building policy-relevant science on food system resilience in LMICs will be to improve our knowledge and understanding of the actors that operate in those systems. At the present time, very little is known (especially among CGIAR researchers) about the “missing (or hidden) middle” – that part of the food system located between production (the farmers) and consumption (nutrition), the two areas where CGIAR has directed most of its research effort to date. It is critical that more attention is paid to 4 the formal and informal actors that make up the rest of the system, and to the factors that make these actors more (or less) vulnerable to disruptions and shocks. Mapping the different sources of vulnerability that affect particular actors (e.g. processors, retailers or street vendors), commodities (e.g. fruits, vegetables), markets (open, closed) or value chains (e.g. small livestock) in low-income countries should be a priority. For this, comparative analyses built on common frameworks should be conducted in which criteria such as seasonality, supply spikes, perishability, or exposure to extreme weather events could be used to identify, assess, and compare the level of vulnerability of actors operating in different commodities and value chains. It is informative to notice that no systematic comparative analysis has been proposed across the CGIAR system to compare different value chains in relation to their respective exposure and vulnerability to COVID-19. Instead most of the documents reviewed here were single- commodity-focused (often in direct line with the institutional interest of the Center to which the authors were affiliated). Even those that discussed several commodities presented them separately. These comparative frameworks should not stop however at the technical (shelf life, perishability, storage, food-borne disease risks, etc.) aspects of the commodity itself. Ineffective rule of law, economic or political marginalization of particular groups, gender inequity, price changes, “invisibility” of the informal sector, etc. are all existing sources of vulnerability that will need to be better understood if we want to be in a position to strengthen the resilience of the food systems’ actors in LMICs. Understanding actors’ responses to shocks One of the key principles in resilience analysis is that the final outcome of a situation where an individual, household, enterprise, sector, or the whole system is hit by a shock, does not depend merely on the direct impact of the shock, but on the combination of that shock with the responses that the different actors (as individuals or as groups) put in place to mitigate or counteract its initial effects. The distressing experience of the impacts of COVID-19 on food systems perfectly illustrates this point: the current threat to the food security and wellbeing of millions of people worldwide does not derive from the effect of the virus itself (the initial shock), but from the disruptions in food deliveries, market linkages, economic activities and household incomes and revenues induced by the successive waves of mobility restrictions and lockdowns that have been put in place by national or local governments as an attempt to mitigate the initial health impact of the pandemic. Beyond its direct informative value, this observation has important implications from a resilience research perspective. It means that documenting and understanding more thoroughly the types of responses put in place by different actors in the wake of an adverse event (flood, political collapse, zoonotic epidemic, etc.) is a second essential step (after understanding their vulnerability) toward building more resilient food systems in the future: without a good understanding of actors’ motives and behaviour and the way they respond to shocks, it is impossible to anticipate their reactions and put in place interventions and policies that can mitigate the negative effects of some of the detrimental responses. Understand better resilience capacity It is now well established that a useful way to conceptualise resilience is to conceive it as an emerging property resulting from a combination of capacities. These capacities are themselves built on social, human, financial, natural, physical or mental capitals which households accumulate or develop during non-crisis periods and can then draw on in anticipation of, or in response to, a sudden or predicted shock. While our understanding of what resources are important for farmers to build their resilience capacities is 5 improving rapidly, in contrast, our understanding of the situation for midstream actors, for whom very little data is collected, is still extremely limited. Yet until we have a better sense of what constitute the elements of each actor’s resilience capacity in a given food system, it will be difficult to design appropriate interventions to help those actors build their own capacity to respond more positively to future shocks. Beyond rhetoric, and beyond resilience As mentioned earlier, statements about resilience are often rhetorical. For instance, it is often claimed that local food systems are more resilient than global ones. No empirical evidence is available, however, to back-up those statements. One obvious implication would be to develop research to test this hypothesis empirically. The underlying mental model, however, is one that assumes there is an ‘optimal scale’ at which resilience operates. Our view is that, instead of trying to determine the optimal scale which allegedly makes a food system (be it local or regional) more resilient, research should be designed to explore and identify the conditions (type of shocks, characteristics of the food system, behaviour of the actors, etc.) that make a given food system more (or less) resilient. This type of information would be very useful for policy makers who are increasingly interested in investing in food system resilience at different scales (local but also regional). However, ultimately, the choice of the ‘right’ investment or policy should be driven, not by resilience considerations, but by the more important objective of making those food systems more sustainable, that is, socially more equitable, nutritionally healthier, inclusive, and environmentally sounder. In this agenda, resilience is the mean, not the end. 6 I. INTRODUCTION “Our food systems are failing, and the Covid-19 pandemic is making things worse. Unless immediate action is taken, it is increasingly clear that there is an impending global food emergency that could have long-term impacts on hundreds of millions of children and adults” António Guterres, Secretary-General of the United Nations, 9 June 2020 1.1. Background and justification As of December 24 2020, just 11 months after the first cases were reported in China’s Hubei province, COVID-19 has taken officially 1.7 million lives around the world, infected more than 76 million persons, and upended the livelihood of billions of people, severely damaging both local and global economies. No country has been spared. No socio-economic group remains unscathed. No one appears immune to its impacts. The pandemic threatens to reverse years of progress on poverty, hunger, health care and education. The world is facing the worst economic recession since the Great Depression. Real gross domestic product (GDP) per capita is expected to decline by 4.2% in 2020, world trade to plunge by 13 to 32% and foreign direct investment by up to 40%. Remittances to LMICs are projected to fall by 20%. So far, COVID-19 is estimated to have caused the equivalent of 400 million job losses globally (UN/DESA, 2020; UNODC and World Bank, 2020). While the virus has affected everyone, it is impacting the world’s poorest and most vulnerable people the most. By the end of 2020, the pandemic was projected to have pushed an additional 88 to 115 million people into extreme poverty (World Bank, 2020a). Although the agriculture sector (host of large number of these poor) and specific actors along the food supply chains have been purposely protected by governments to reduce the risk of national or global food supply crises, the pandemic through its direct effects on individuals’ health and indirectly through the disruptions that the local and national authorities’ responses have created- has had tremendous effects on the ability of food systems to operate effectively. For several months, the documentation of these impacts has often been anecdotal and based on restricted scopes. In late 2020, an increasing number of peer-reviewed articles have been published, substantially raising the quality of information available. For the most part, however, these scientific articles remain based on limited samples, focused on geographically specific areas, or on case studies. Although some global assessments are available (e.g. HLPE, 2020), those are not always exhaustive or systematic in nature. This restricts the ability of decision-makers, at both national and international levels, to get the ‘full picture’ of the situation, potentially identify patterns across countries or regions, and subsequently identify effective recovery policies and interventions that can lead to more resilient national food systems, those that reduce the likelihood of future shocks to occur and increases society’ ability to handle these shocks when they do occur. There is a need, therefore, to conduct a comprehensive and systematic review of the impact of COVID-19 on food systems with the main objective of identifying and mapping out the ‘key fragility points’ of these food systems and documenting the nature and scope of the disruptions that the pandemic and subsequent government implemented control measures have foisted on the different actors of those food systems, from producers all the way to consumers. 7 1.2. General objective of the study and scope of the report As part of the work implemented by CGIAR on COVID-19, the CGIAR’s COVID-19 Hub Working Group 4 “Address food systems’ fragility and build back better” was tasked with conducting a global assessment of COVID-19’s impacts on food systems and their actors, focusing specifically on the consequences the pandemic brought on the food security and nutritional status of those affected. The assessment, therefore, includes formal and informal actors of food supply chains (from producers to street vendors) as well as consumers, in both rural and urban settings, and the changes induced by the COVID- 19 crisis on their food environments. The scope of the assessment is global in scale. The intention was to conduct a rigorous mapping of the information available at national and international levels, and in doing so, to produce the first comprehensive assessment of this type at the global level. For this purpose, a total of 337 documents published or made available in four different languages (English, French, Spanish and Portuguese) between January and December 2020 were scanned and systematically reviewed2. Particular attention was paid to the situation in LMICs, where most of the poor and food insecure households currently live, but information from higher-income countries was also included in the assessment. Building on the empirical information and evidence collated and synthesized through the review, we then propose to revisit some of these pieces of evidence from a resilience perspective, assessing the potential usefulness of that concept in the specific context of the COVID-19 crisis. The last section offers some preliminary reflections for policy makers and researchers, identifying in particular areas of policy, interventions and research aiming to ‘building back and better’ our food systems. 2 The detailed step-by-step used to scan, organize and review the documents included in the assessment is presented in Appendix A.1, while the full list of documents that were reviewed are available at https://a4nh.cgiar.org/impacts-of-covid-19-on-peoples- food-security-documents-reviewed/ https://a4nh.cgiar.org/impacts-of-covid-19-on-peoples-food-security-documents-reviewed/ https://a4nh.cgiar.org/impacts-of-covid-19-on-peoples-food-security-documents-reviewed/ 8 II. ANALYTICAL FRAMEWORK Several elements need to be considered in order to provide a comprehensive framework for this analysis: First what the effects of the COVID-19 crisis on people’s life and food security are; second how these actors are affected (the causal pathways); and third, who those actors3 are. We detail those different elements of the framework in the remaining part of this section. 2.1. Effects of COVID-19 on people life and food security and nutrition In relation to the wider conceptualization of ‘food system’ as now widely adopted in the academic community (e.g. HLPE 2017 p.26), two primary concepts were used to unpack the effects of COVID-19 on people’s food security and nutrition: the concept of Food Security (as historically defined by FAO - see, e.g., FAO 1996, 2008) and the concept of Food Environment (as proposed recently by several authors, e.g. Herforth and Ahmed (2015) or Downs et al. (2020)). Together these two concepts are useful as they emphasize complementary ‘dimensions’ which are important in view of the main objective of this assessment. These complementary dimensions are captured by the four components of the concept of food security: availability, access, utilization (quality and safety) and stability; and five elements that are recognized to be critical in determining food environment: proximity, convenience, availability, affordability, and quality of food items (Downs et al. 2020). Note that several of those dimensions are common to both concepts. Building on recent conceptualizations (e.g., Devereux et al. 2020; Savary et al. 2020), several additional elements need to be considered when one intends to conduct a comprehensive evaluation of COVID-19 impact on people food security while at the same time embracing a wider food system approach (HLPE, 2017, Brouwer et al., 2020). These include: the diversity of food items (at the interface between food security, food environment and health) (Downs et al., 2020); the quantity of food waste and losses (in relation to the disruptive effects of COVID-19 on food systems’ efficiency) (Aldaco et al., 2020); and a series of criteria related to the potential impacts of COVID-19 on the health and wellbeing of actors within the food system, including their agency and sense of self-efficacy (e.g. Yildirim and Guler, 2020). Finally, we propose to include two additional elements which are not generally considered in the food system literature, but have been mentioned in relation to the outbreak of COVID-19: the occurrence of domestic violence and social unrests at household and community levels (e.g. Hamadani et al., 2020; Gumede, 2020); and the increased risk of exposure to COVID-19 due to the adoption of ‘risky’ coping strategies by those actors (Chan et al., 2020). Together these different elements are presented in Fig.1. They constitute the different dimensions that will be more systematically explored through this exercise. In addition, macro-level estimations of change in GDP and (income) poverty will be considered. 3 In this document the term ‘actor’ is used to refer interchangeably to either the persons (women, men, youths) who are actively engaged in economic activities in the food system, or the micro, small, medium or larger-scale enterprises that make the food system (an individual enterprise is also an actor of the food system). This amalgam may however be conceptually confusing in specific circumstances, for instance when we refer to “the food security of food systems’ actors” (in that case obviously we refer to the food security of the individual persons only, not the enterprises…) or when we claim that “eventually all actors in the food systems are consumers” –again this refers to individual persons, not enterprises. When possible, we substituted ‘actors’ with ‘people’ to reduce this confusion. 9 Fig.1. The different elements to be considered in assessing the impact of COVID-19 on people’s food security in the wider context of food system. 2.2. Typology of impacts and affected actors The two other elements of the framework (which actors are affected, and how they are affected) needed to be considered together, mainly because causal pathways are usually actor-specific. Building on and expending some recent reviews of the impacts of COVID-19 on value chains (e.g. OECD, 2020a) and on people’s food security (Béné, 2020; Savary et al., 2020), a series of 25 related but distinct potential effects of COVID-19 on food system actors were identified from the literature. Those are listed in Table 1, along with the groups of actors which they are expected to affect, and organized along four generic steps: Direct effects and responses → Immediate consequences → Subsequent repercussions → Final impacts. For sake of clarity, these different actors have been grouped into three ‘meta-groups’: producers (including wage workers), mid-stream actors and consumers. The ‘mid-steam’ meta-group, however, includes several distinct sub-groups, that will be differentiated subsequently at the analysis stage using four generic sub-groups: processors, transporters, wholesalers/retailers, and food vendors, in line with the main types of activities usually recognized as present in food systems (e.g. HLPE, 2017). Table 1 also reflects the fact that all the actors within the food systems are eventually consumers and, as such, may be affected through two main impact pathways: (i) as actors in the food systems, and (ii) as consumers. Conceptually, this approach allowed us to propose a simple typology of COVID-19’s impact pathways and associated groups of affected actors. Those effects are not exclusive or isolated, in the sense that many are expected to create ripple effects (Béné, 2020) that will affect one or several groups of actors beyond their original impact. Finally, note that while it is conceptually possible to distinguish short-term/immediate effects (0-4 weeks) from medium-term (1-6 months) and longer-term impacts (6+ months) -as some others proposed (e.g. Savary et al., 2020; HLPE, 2020), the occurrence of successive waves of the disease and (more importantly) the series of subsequent enforcements and relaxations of lockdowns and related mobility Food Environment Food Security Health & Wellbeing Foot Print ▪ Stability ▪ Availability (supply) ▪ Access (physical accessibility) ▪ Affordability (economic accessibility) ▪ Proximity ▪ Convenience ▪ Waste and Losses▪ Diversity of food items ▪ Agency and self-efficacy ▪ Domestic violence and unrest ▪ Risk of contagion ▪ Quality and safety (utilization) • change in GDP • change in poverty Macro-economyFood system 10 restrictions make it extremely difficult to differentiate with certainty the medium or long-term impacts of the first wave from the immediate effects of subsequent waves. Additionally, given that waves and periods of lockdowns occurred at different times in different countries, the emerging pattern observed at the global level results from a combination of unsynchronized effects at countries’ individual levels, again making the empirical distinction short versus long-term effect empirically impossible. Table 1. Typology of COVID-19 impacts and affected actors in the context of food systems Typology of impacts induced by COVID-19 Actor affected by the event Direct effects of COVID or directly-related responses by authorities a. COVID related illness or death All actors b. Mobility restriction and lockdown All actors c. Safety or sanitary decrees/regulations Primarily mid-stream actors Immediate consequences on food system actors 1. Disruption in upstream supply chain (e.g. fertilizer) and/or subsequent effects on prices or quantity/accessibility/quality of inputs Producers, workers and/or mid-stream actors 2. Disruptions in actors’ own activities due to mobility restriction and lockdown Producers, workers and/or mid-stream actors 3. Loss of or reduced connectivity with established downstream actors (direct consumers, contracted business partners, e.g. processors, retailers, etc.) Producers, workers and/or mid-stream actors 4. Reduction in labour/workers availability (due to mobility restriction, increase in public transport costs, or fear of exposure to virus) Producers, workers and/or mid-stream actors 8. Forced closure of business due to safety or sanitary decrees/regulations Producers, workers and/or mid-stream actors 9. Degradation in Rules of Law (e.g. contractual issues, enforcement issues, information access issues, etc.) Producers, workers and/or mid-stream actors 13. Disruption in food supply due to hoarding behaviour Producers, workers, mid-stream actors and/or consumers Subsequent repercussions on food system actors and/or other (non-food system) actors 5. Drop in (agri)food business profitability Producers, workers and/or mid-stream actors 6. Reduction in downstream demand Producers, workers and/or mid-stream actors 7. Increased wasted food/post-harvest loses due to disruption in supply chain (upstream or downstream) Producers, workers and/or mid-stream actors 10. Increased gender discrimination against women in particular subsectors (processing, retailing, selling) Producers, workers and/or mid-stream actors 11. Increased abuse against marginalized individual or groups in particular subsectors (processing, retailing, selling) Producers, workers and/or mid-stream actors 14. Loss of job and/or reduction in income/revenues (due to mobility restriction, forced closure of business, etc.) Producers, workers, mid-stream actors and/or consumers 15. Voluntary or involuntary increased risk of exposure to COVID health impact (contagion) due to the adoption of particular copying strategies Producers, workers, mid-stream actors and/or consumers 17. Disruption in access to (usual) food outlets Consumers(a) 18. Increased price of food – lower purchasing power Consumers(a) Final impacts on consumers’ food security dimensions and food system actors’ health & well-being 12. Drop in perceived self-efficacy or agency among individuals or particular groups Producers, workers and/or mid-stream actors 16. Domestic violence and/or increased tension in households Producers, workers, mid-stream actors and/or consumers 19. Degradation in food choice and diversity (e.g. shift to cheaper, fewer or less nutritious food items) Consumers(a) 20. Reduction in proximity and/or convenience – due to mobility restriction, increase in public transport costs, or fear of exposure to virus Consumers(a) 21. Increased risk of consumption of unsafe food due to reduced access to usual/ traditional food suppliers/outlets Consumers(a) 22. Forced shift to more expensive food outlets due to closure of those outlets or due to mobility restriction Consumers(a) Notes: (a) ‘Consumers’ includes producers, workers and/or mid-stream actors as consumers 11 III. QUALITY OF EVIDENCE In the first few months following the outbreak of COVID-19 and until mid-2020, experts and the science were not able to provide all the answers that society needed about the disease. At the same time, uncertainty and the need for information were high, creating an important information gap in which other sources of knowledge came into play. In particular, what is called ‘experiential knowledge’ – that is, knowledge acquired as a consequence of experience (either personal or other people’s experience) (Blume 2017) has been a major source of information and data, available mainly from web-based material (e.g. blogs), grey literature, news and social media accounts and first hand observations. In a period where the concept of fake news is a reality and the COVID-19 situation was (and still is) evolving on a daily basis, there is a need to acknowledge and account for the risk of inaccurate, incorrect, unverified, and (intentionally or unintentionally) mis-leading or fabricated information4. The World Health Organization (WHO) referred to epidemics of rumours or ‘infodemics’ in reference to ‘the rapid spread of information of all kinds, including rumours, gossip and unreliable information’ (WHO 2018, p.26). Fig.2. The two-dimensional scheme used to assess and account for the quality of evidence of the documents incorporated in the mapping. In this context, we propose to adjust the level of analysis to the quality of the information. Three levels of analysis were therefore distinguished, based on three versions of the analytical framework: (i) an abridged 4 Between March 1, 2020 and April 8, 2020 for instance Facebook AI Research (FAIR) put warning labels on about 50 million pieces of content related to COVID-19 on Facebook and removed more than 2.5 million pieces of content for the sale of masks, hand sanitizers, surface disinfecting wipes and Covid-19 test kits. Source of information Blogs, new/social media (experiential knowledge) Expert re-interpretation (cognitive authority) Blog, webinar, report/review (experiential knowledge) Type 1 Type 2 Absence of rigorous and replicable data-generating protocol Primary data (field collected) Expert analysis (cognitive authority) Peer-review (cognitive authority) Peer-reviewed articles (scientific knowledge) Type 3 Presence of rigorous and replicabledata-generating protocol Kn o w le d ge e la b o ra ti o n Quality of data Representativity Reliability Validity* Presence / absence of a sound scientific sampling protocol/design Criterion, content and construct validities satisfied/not satisfied Presence / absence of a rigorous survey method for unit data analysis * Wi l l not be assessed in this review Type 4 Technical report Level 1: not considered further in the review Level 2: expected low representativity and reliability Level 3: expected medium representativity and reliability Level 4: expected medium to high representativity and reliability * Wi l l not be assessed in this review 12 version, (ii) a simplified version, and (ii) a full-fledged version5. The level of ‘thoroughness’ of the analysis was then decided, depending on the quality of the evidence being reviewed. To assess this quality of evidence and incorporate it into the overall analysis, a two-dimensional assessment scheme was used. The two dimensions considered were: (i) knowledge elaboration and (ii) quality of data. The two- dimensional assessment scheme combining those two dimensions is represented graphically in Fig.2 and discussed in greater detail below. 3.1. Knowledge elaboration During the first few weeks after the pandemic outbreak, the initial stage of knowledge elaboration – the act of adding more information to existing information to create a more complex, emergent understanding of a process (Kalyuga, 2009) – has been principally based on accounts, stories, and anecdotes shared by individuals mainly through web-based material (e.g. blogs), grey literature, news medias, and personal social media accounts, as well as first-hand observations. We propose to refer to this type of data / information as Type 1 documents whereby the process of knowledge elaboration relied on subjective “observations” (experiential knowledge –Berg 2008) made by individuals without necessarily any form of institutional or other type of endorsement or scientific validation. The second level of knowledge elaboration (Type 2 documents) corresponds to situations where experts with cognitive authority6 started to use Type 1 information and reinterpret or synthesize it in the form of grey literature – e.g., blogs – made available on their own institutions’ websites or those of recognized or well-established institutions. As such, their professional position and title, their expertise, and the institutions which owned the website or hosted the blogs offered a form of implicit recognition. The origin of the data that was used for these documents is however difficult to verify, and although some of these documents may be based on genuine information/analysis, the absence of rigorous and replicable sampling protocol means that it is difficult to rely uncritically on them for a rigorous assessment. Type 3 documents in the knowledge elaboration refer to situations where the creation of knowledge is based on validated data. This validation will usually have started with the generation of primary data collected and analysed through a precise and/or clear protocol based on some form of analytical framework, sometimes reinforced by explicit sampling and/or survey procedures. Type 3 documents refer for instance to technical reports, information/facts sheets or similar policy briefs issued by United Nations (UN) or other international experts groups and academic institutions (e.g. university, research centres, Organization for Economic Co-operation and Development, European Union, World Bank, etc.) incorporating (sometimes implicitly) these sampling and structured data collection procedures. Finally, when the resulting interpretation of data by the author(s) of the analysis (generally a scholar/expert with cognitive authority) is further confirmed and validated through a peer-reviewed process, the document would be categorized as Type 4 document. This Type 4 includes therefore primarily peer-reviewed academic papers published in international journals. 5 The abridged version considers only two (meta)groups of actors (food system actors, and consumers) and seven generic types of impacts (those are detailed in Appendix A.1); the simplified version distinguished three categories of actors: producers, mid- stream food system actors, and consumers, and details 14 different categories of impacts; and the full-fledged version of the framework considers the six groups of actors (producers, transporters, processors, wholesaler/retailers, food vendors and consumers), and use the full 22+3 (25) types of impacts identified in Table 1. 6 ‘Cognitive authority’ of an information source is conceived when a certain community is allowed to negotiate what counts as an authorized source of information (Neal and McKenzie, 2011) https://en.wikipedia.org/wiki/Emergent_properties 13 We propose using these four incrementing levels (Types 1-4) to categorize the knowledge elaboration and use those to define the first dimension of the quality of evidence scheme. 3.2. Quality of data The second dimension considered in the quality of evidence scheme is the ‘quality of data.’ Irrespective of whether or not the information shared in the document has been generated by an observer with cognitive authority and validated through a peer-reviewed process, it is possible to distinguish different levels of data quality. Three criteria are generally proposed in the academic literature to assess the (scientific) quality of evidence (West et al. 2002): (i) representativeness (measured by the presence of a sound scientific sampling protocol/design that ensures that (a) insights on the characteristics of the whole population is available, and (b) individual units interviewed/sampled are selected through some type of random fashion; (ii) reliability measured by the presence of a clear rigorous principle or method for individual unit data collection (survey methodology); and (iii) validity which should include criterion- related, content and construct validities –see details in Cronbach & Meehl (1955). In the case of this review, the three elements of the validity dimension (criterion-related, content and construct validities) would be difficult to assess. It was decided therefore that only the first two criteria, representativeness and reliability, would be included in the scheme, and to use a three-level score (from zero to two) system to categorize the quality of these two criteria, as detailed in Table 2. Table 2. The 3-level score system used to assess the two dimensions of data quality score Representativeness Reliability 0 absence of sound scientific sampling protocol/design absence of rigorous survey method for unit data analysis 1 presence of some element of sampling design but not one that ensures that the whole population is represented, and that a random selection is used presence of some form of survey method for unit data analysis 2 presence of a sound scientific sampling protocol/design presence of a rigorous survey method for unit data analysis 3.3. Linking the quality of evidence to the review process As will appear later in this report, a large proportion of the data available/published so far on COVID-19 displays a relatively low quality of evidence (low representativeness and/or limited reliability). We therefore structured the analysis so that more attention was paid and more information could be extracted/collected from the documents that were considered more reliable (Type 3 and Type 4 documents), while a ‘lighter’ analysis was applied to the other, more numerous but less reliable, documents (Type 2 documents). Type1 documents were not considered further in the analysis. Type 2 documents were entered in the database, as they may offer some of the earliest and unique descriptive information about the short-term immediate impacts of COVID-19 from the first few weeks of the outbreak of the pandemic. They may also offer some unique accounts of the short-lasting effects of the pandemic (e.g. hoarding behaviour). They were, however, expected to be characterized by a low level of representativeness and reliability. The information was treated accordingly, using an abridged version of the analytical framework. 14 Type 3 documents were entered in the database and analysed as follows: those with medium scores for both representativeness and reliability were analysed using a simplified version of the framework, while those with low scores in either one or both of representativeness and reliability dimensions were analysed using the abridged version. Type 4 documents : those with high scores for both representativeness and reliability were analysed using the full-fledged analytical framework, while those with medium scores in either or both representativeness and reliability were analysed using the simplified version of the analytical framework and finally those with low scores analysed with the abridged version. The decision-tree presented in Fig.3 summarizes the process. Fig.3. Decision-tree for the analysis level used in the review –based on the quality of evidence of the documents. Rep = representativeness; Rel = reliability. See text for details. Type 1 documents Type 2 documents Type 3 documents Type 4 documents Discard Abridged analytical framework if Rep or Rel scores = 0 if Rep and Rel scores = 1 Simplified analytical framework if Rep or Rel scores = 0 if Rep or Rel scores = 1 Full-fledged analytical framework Abridged analytical framework if Rep and Rel scores = 2 Simplified analytical framework 15 A woman trader in the Kranggan market, Temanggung, Central Java, Indonesia, in September 2020: Many women operate as informal actors in food systems, and faced significant impacts from lockdowns and other measures taken in response to the COVID-19 pandemic. Photo: Ma Andyanto/Shutterstock 16 IV. KEY FINDINGS 4.1. Data analytics In total, more than 9,630 documents were identified between October 26 and December 15, 2020, using a combination of keywords that had been specifically designed to address the objectives of the study (see Appendix A.1 for the detailed list of these keywords). The vast majority of those documents publicly available appear to be Type 1 documents (i.e. news medias, and personal/non academic blogs)7. In line with the proposed methodology these Type 1 documents were discarded, leaving us with 363 documents. After removal of duplicate documents (documents published simultaneously in several languages) 337 Type 2 or higher documents were available, which were then entered in the database and reviewed. A PRISMA-like flowchart is proposed in Fig.4 that summarizes the different steps of the protocol adopted for the review. Fig.4. PRISMA flowchart summarizing the protocol adopted in the review Those 337 documents cover the period January to December 2020, with the highest number of documents published between April and July (Fig.5). One hundred sixty-seven of these documents were published in English (50% of total), 90 in French (27%), 64 in Spanish (19%) and 13 in Portuguese (4%). 7 “Posts” on Facebook, Twitter etc. were not taken into account. Between March 20, 2020, and Dec 15, 2020 more than 800 million tweets (in English only) were posted on COVID-19. https://ieee-dataport.org/open-access/coronavirus-covid-19-tweets- dataset 17 Fig.5. Monthly number of documents published (amongst those identified in this review) Those 337 documents describe the effects of COVID-19 in 62 different countries covering the 5 major regions of the inhabited world: Africa, Asia, Europe, Oceania and the Americas (see Map 1 and Table 3). Some countries were referred to multiple times (e.g. India 15 times; Myanmar five times), while others were not covered at all (e.g. Norway, Pakistan). We recognize that part of this situation reflects the fact that only four languages were used to conduct the review and that certain parts of the world (e.g. the Middle East, East Asia, Central and Southern Europe, Scandinavia) may not have been appropriately represented by documents in these languages. Map 1. Geographical coverage of the Review (i.e. countries discussed or mentioned in the documents included in this review). Number of documents 18 Table 3. Geographical coverage of the review (by region) Region Africa Asia Europe Oceania Americas Total Number of documents 66 44 11 11 71 203(a) Percentage of the total 33% 22% 5% 5% 35% 100% Note: (a) total differs from the overall number of documents reviewed as not all documents consider specific regions/countries Conversely, some countries with a high number of associated documents may be countries with active ‘bloggers’, including researchers or experts from international research organizations and development agencies working/living in these countries. In addition a substantial number of documents (106, or 31%) were ‘international’ in scope and discussed aggregated information at the global level. At the other end of the spectrum, only seven documents (2%) present subnational data, focusing either on states (such as, e.g., Odisha State in India) or cities (such as Addis Ababa in Ethiopia). 4.2. Quality of the evidence and implications for the analysis Using the two criteria of quality of evidence (the knowledge elaboration process and the quality of data) proposed earlier, the following key-findings emerge. As far as the knowledge elaboration is concerned (Fig.6a), at the time of closing the analysis (December, 15, 2020) the largest number of documents included in the review were Type 3 documents (48%), that is, technical reports, information/facts sheets or similar policy briefs issued by international experts groups and academic institutions. The second most frequent type of documents was Type 2 documents (34%), blogs and similar grey literature posted by authors with cognitive authority but which did not always rely on transparent or rigorous data collection protocol. Fig.6a. Typology of the 337 documents included in the global assessment. Fig.6b. Change in the proportions of types of documents posted or published over time. Type 2 documents: blogs and similar grey literature; Type 3: technical reports, information/facts sheets or similar policy briefs; Type 4: peer-reviewed articles. Finally the data indicate that, as per December 2020, still relatively few peer-reviewed articles were available (18% of the total documents reviewed) even though special efforts had been made by several journals’ editorial teams to facilitate and expedite the review process for COVID-related manuscripts. It is expected that the number of these scientific articles will progressively increase in 2021, and their proportion is indeed increasing progressively over time (Fig.6b). N=116 (34%) N=160 (48%) N=61 (18%) type 2 type 3 type 4 0 0.2 0.4 0.6 0.8 1 1.2 1 2 3 4 5 6 7 8 9 10 11 12 type 4 type 3 type 2 19 The second component of the assessment relates to the data quality. Under this criteria, we assessed the level of representativeness and reliability of the documents following the protocol described in Sections 3.2 and 3.3 above. We were initially expecting that while Type 2 and Type 3 documents would offer minimal to medium quality of evidence, Type 4 (peer-reviewed articles) would be characterized by high level of representativeness and reliability. The data indicates however that this assumption is only partially verified. Amongst the 61 Type 4 documents (scientific articles) reviewed, 26 are effectively characterized by a high levels of representativeness and reliability (scored 2 for both criteria as per Table 2); the rest of these peer-reviewed articles showed lower-than-anticipated quality in their data collection (scored 1 or even 0 for one or more of the two criteria). Overall 180 documents were eventually analysed using the abridged version of the analytical framework, 131 were analysed with the simplified version and only 26 with the full-fledged framework – thus breaking down the 337 documents into the three levels of analysis presented in Fig.4 above. 4.3. Emerging evidence on COVID-19 impacts Data confirm that the COVID-19 pandemic has had major impacts on health across the globe. As of December 24, 2020, more than 1.7 million people had died from the virus and 76 million were infected in the course of several waves that spread around the world (WHO 2020). In response, governments have imposed a range of measures, including social distancing, restrictions on mobility, curfews, and temporary closure of workplaces, generally known as a ‘lockdown’, as an attempt to contain the spread of the virus (Swinnen and McDermott, 2020; FSIN and GNAFC, 2020; World Bank, 2020). 4.3.1. Loss of income and jobs There is a large consensus among the literature that with the notable exception of those who lost members of their family to the virus, the major direct effect of COVID-19 has been, and continue to be, through its impact on the employment, income and associated purchasing power of all those whose jobs and livelihoods have been affected by the measures put in place by the local and national authorities at local and/or national levels (FSIN and GNAFC, 2020; Robins et al., 2020; FAO, 2020a; CARICOM et al., 2020; GIEA, 2020; Arévalo et al., 2020; UN/MEPD, 2020). In Ethiopia for instance, about 60% of the households interviewed in Addis Ababa between May and July reported a loss of income (Hirvonen et al. 2020a); in Nepal 31% (WFP, 2020); in Myanmar 80% (Headey et al., 2020a), in Nigeria around 75% (Amare et al., 2020). In the Caribbean about 45% of households surveyed mentioned a loss of job or a reduction in income/salaries (CARICOM et al., 2020). In Bangladesh, 96% of the more than 2400 women surveyed by Hamadani et al. (2020) through phone interviews reported a reduction in paid work for their family, with the median monthly family income falling from US$212 at baseline to $59 during lockdown. Several of those reports also highlight that the figures are usually higher for urban households than for rural ones (e.g. Headey et al., 2020a) and for women than for men (CARICOM et al., 2020). 4.3.2. Clear but difficult-to-assess impact on food security Although not always measured with the same methods or techniques, all the documents reported that this sudden reduction of income has had repercussions on different aspects of households’ food security and nutrition. In Nigeria the comparison of pre-COVID LSMS-ISA data (collected in 2018) with the 2020 LSMS-ISA data shows significant difference for all four indexes used: skip meal, run out of food, went 20 without eating for a whole day, and food insecurity (Amare et al., 2020). Using the Food Insecurity Experience Scale (FIES) Headey et al. (2020a) show that in Myanmar it is mainly access to healthy food that was reported to be affected. Likewise, in India, 62% of the farm households interviewed by Harris and colleagues reported disruptions to their diets. In particular, while around 80% of these households reported an ability to protect their consumption of staple food, the largest declines in consumption were in fruit and animal source foods other than dairy, in around half the households (Harris et al., 2020). In Mexico, using the Latin American and Caribbean Food Security Scale (ELCSA) included in three waves of a phone survey, Gaitán-Rossi et al. (2020) show that the COVID-19 lockdown was associated with an important decline in food security, affecting 25% in households with children (compared to 39% in 2018). An online cross-sectional survey conducted in two favelas in Sao Paulo (Brazil) between March and June 2020 shows that 47% of respondents experienced moderate or severe food insecurity; 89% of them reported uncertainty to access food, 64% eating less than they should, and 39% skipping a meal (Manfrinato et al., 2020). Data from Nigeria also suggest that households living in remote and conflict- affected areas are more likely to experience deterioration in food security (Amare et al., 2020). Those declines in different aspects of food security, however, do not affect only those populations in low- income countries (LICs). In Vermont (USA), using the six-item validated food security module Niles et al (2020) showed that there was nearly a one-third increase (32.3%) in household food insecurity since COVID-19, with 35.5% of food insecure households classified as newly food insecure. Reduction in incomes/revenues is one of the main reasons for higher food insecurity. However, many other reasons were identified. In Nepal for instance, among the households who reported food insufficiency, 21% identified a shortage of food in markets and food outlets (WFP, 2020a). In Odisha (India) travel restrictions were reported by households as the main reason for insufficient quantities of food (IAG and WFP, 2020). In Vermont food access challenges included not finding as much or the kinds of food that someone wanted, going to more places than usual to find food, and not being able to afford the food a household wanted (Niles et al., 2020). The situation may however be multifaceted and sometimes difficult to interpret clearly. In Addis Ababa for instance, 60% of the households surveyed by Hirvonen and his colleagues had reported income losses, suggesting that the impact was mainly on urban household food security (Hirvonen et al., 2020a). Yet another survey suggests that only 5% of households consider that shortage of food had the greatest impact on their households; even social distancing or being sick (or fear of being sick) were perceived as having more impact (Abate et al. 2020). While the consumption of legumes and vegetables was reported to have decreased significantly compared to September 2019, the consumption of staples appears to have increased notably (Hirvonen et al., 2020a). Likewise, in other cases such as in India where a very large majority of households seem to be able to protect their staple consumption, the same does not apply to other food items, with the largest consumption declines in fruit and animal source foods other than dairy (Harris et al., 2020). In parallel vegetable consumption was reported to have fallen in 30% of the households, but increased in another 15% (Harris et al., 2020). Finally, to further complicate the assessments, in some cases, although specific figures on the level of food insecurity were reported, no baseline or control/reference values were offered that would allow to compare the situation prior to COVID-19 (e.g., Headey et al., 2020a). In sum, while the overall detrimental effect of COVID-19 on different aspects of people’s food security is clear and unquestionable, the intensity and forms that this food insecurity takes is more difficult to establish precisely. Many reasons can be identified for this: first, the very fluid and rapidly evolving 21 situation and the fact that the impact on people appears to be time- and geography-specific, but also depends on the food item/value chain considered and the socio-economic group interviewed; and finally, the fact that multiple and heterogeneous sets of various, mixed and sometimes modified indicators and approaches have been used by the researchers. 4.3.3. Expected impact on nutrition For nutrition (for which the measurement toolbox available is well-established and the protocols considered quite rigorous) the current situation may not, however, be clearer. While there is a large consensus in the nutrition community that the COVID-19 pandemic is likely to increase the risk of all forms of malnutrition (FSIN and GNAFC, 2020), primary data are not yet available to confirm these predictions. As a consequence, current discussions around the effects of COVID-19 on nutrition are primarily based on macro or micro-level simulations (e.g. Akseer et al., 2020; Headey and Ruel, 2020; Roberton et al., 2020). These predict a potential substantial increase in the prevalence of moderate or severe wasting among children younger than five years of age due to projected losses in gross domestic income per capita (Headey and Ruel 2020). If these projections are correct this would translate into an additional estimated 6.7 million children with wasting in 2020 compared with projections for 2020 without COVID-19 (Headey et al., 2020b). In parallel, the disruption of health services during lockdowns is expected to further compromise maternal and child health and mortality (Roberson et al., 2020) as well as other forms of malnutrition with the deepening of economic and food systems crises, including child stunting, micronutrient malnutrition, and maternal malnutrition (Akseer et al., 2020). With the exception of Werneck et al. (2020) who look at the incidence of elevated consumption of ultra-processed food consumption and lower consumption of fruits and vegetables during the COVID-19 pandemic, there has not been any attempt yet to assess the effects of COVID-19 on over-weight and obesity, even if change in consumers’ behaviour and general degradation in food choice and diversity have been widely reported (Villaseñor Lopez et al., 2021, Casco, 2020; Harris et al., 2020; Zidouemba et al., 2020; Hamadani et al., 2020). 4.3.4. Effect on different actors of the system In parallel to the reported impact of COVID-19 on consumers, a large number of documents have highlighted the disruptive effect of the pandemic on the livelihood and economic activities of the other food system actors, starting with the primary producers (e.g. Termeer et al., 2020; Rosen, 2020; Reis- Filho and Quinto, 2020; Urioste Daza et al., 2020; Quiroga Mendiola et al., 2020; Tounkara 2020). These disruptions include the loss or reduction of access to farming input supply or the sharp increase in their prices. Burkart et al. (2020), for instance, report that urea fertilizer prices have on average increased by 9.1% between March and April 2020 in Colombia, severely affecting the livestock sector. Input suppliers as well as other actors along the chain have been affected. Cattle slaughtering had decreased by 30–40% during the first weeks of April (Burkart et al. 2020). In Andra Pradesh (India), Nedumaran et al. (2020) reported that due to transport and contact restrictions, agriculture input suppliers lost up to 75% of their business. Three quarter of these input dealers reported an average 44% decrease in the number of farmers visiting their shops to buy farm inputs. Still in India, Harris et al. (2020) also reported that 87% of the vegetable producers they interviewed had their production interrupted. In some areas (e.g. Jharkhand State) the figure was 94%. Aggarwal et al. (2020) found large reductions in profits among farmers in 22 Liberia, declining to almost zero by May 2020, and smaller but still substantial losses in Malawi of about 40% in April and 20% in June. In parallel the same study estimated that 98% of market vendors closed or reduced business hours, relative to 25% in Malawi. In Ethiopia Hirvonen et al. (2020b) observed changes and disruptions in business practices of traders, including increased costs of transport (reported by 93% of the wholesalers interviewed), decrease in downstream demand (reported by 83% of wholesalers and 82% of retailers), and subsequent losses in business (76% of the wholesalers and 62% of the retailers). In China, using a multiplier model built on China’s most recent social accounting matrix (SAM) for 2017 with 149 economic sectors, Zang et al. (2020) estimated that more than 46 million agri-food system workers (about 27% of total employment) temperately lost their jobs to COVID-19 during the initial lockdown phase. While many of these jobs resumed afterward, the level of agri-food system employment continues to be lower than prior to the COVID-19 outbreak. Overall agri-food system employment in China is estimated to have dropped by 8.6 million, which accounts for about 33% of the total jobs lost (Zang et al., 2020). 4.3.5. Mixed prices effects along the chain Overall, the effects of COVID-19 on local farming products is difficult to assess precisely. In particular when quantitative data is available, it does not necessarily support the view that COVID-19 induced a systematic increase in food prices. A distinction needs also to be made between production/ farm- gate/rural prices and retail/consumers/urban prices. While the former often decreased due to the ‘collapse’ of the demand following the disruption in value chain and the lockdown of (informal) traders, prices at retail and selling points may have increased in many urban centres. Overall, this creates a relatively complex and kaleidoscopic picture. For instance, in India price reductions were reported by more than 80% of the farmers interviewed by Harris et al. (2020), with reductions by more than half for 50% of them. Likewise, in Myanmar more than half the traders interviewed reported that the price of oilseed and pulses have decreased by at least 10% while another 34% estimated that prices of maize, oilseed and pulses had not changed compared to 2019 (Goeb et al. 2020). In contrast, in Liberia, Aggarwal et al. (2020) show that traditional crops’ prices increased by 3-9% during the COVID-19 period (relative to the month before), but in Malawi they had decreased by about 20-24%. When restricted to staple crops (rice, cassava, sweet potatoes, maize, beans), prices increased by 18-20% in Liberia but declined by even more (29-36%) in Malawi. In Ethiopia, Hirvonen et al. (2020b) observe that “retail price trends were quite heterogeneous during the pandemic” (p.7). While tomato and onion prices increased by 33 and 20%, respectively, green pepper and cabbage prices went down by 13 and 12%, respectively. In sum no clear trend seems to emerge at the global level. 4.4. Proposing a (more) holistic and dynamic assessment of COVID-19 The review presented above provides a good initial overview of the different impacts of COVID-19 on local food systems and their actors, based on some of the most reliable quantitative evidence available in the current literature. As such the review is useful in offering detailed accounts of the situation. But it does so in a way that may suffer two potential limitations. First, the different documents included in the review explore the effects of COVID-19 with a ‘lens’ dictated by the nature or the data that was possible to collect at the time of the surveys. Since operating directly in the field was not possible (due to lockdowns, social distancing, and mobility/travel 23 restrictions), most surveys were conducted via telephone interviews. This means that the majority of these studies have put a strong emphasis on tangible, easily or quickly ‘measurable’ or quantifiable data/indicators such as self-reported changes in incomes or profit, level of activity or (volume of) production, etc. –often recorded using pre-coded/structured questionnaires-, rather than on more intricate, nuanced or contextual qualitative types of data or processes which would have required more time- intensive methods (e.g. ethnography, grounded theory, etc.) to be collected. Consequently, changes in behaviour and/or shifts in preferences are hardly considered. Beyond the nature of the data per se, the second potential limitation relates to the fact that those quantitative analyses generally focused on specific aspects/activities or particular groups of actors of the food systems but did not necessarily adopt an approach that allowed them to capture the systemic, interactive nature of the processes they were observing. What is proposed in these documents is therefore a detailed, yet fragmented/partial account of the situation. To palliate these issues and, in particular, to reduce the potential bias introduced by the fragmented/piece- meal nature of the collected evidence, we complemented the review proposed above with an analysis based on a more holistic approach, where emphasis was put on the relative importance of each different type of disruption observed (as opposed to their individual reported severity). By adopting such a framework structured around a system-based comprehensive approach of the processes at work, we were able to reconstruct a more balanced and nuanced, but also more holistic inventory of the different aspects of the COVID-19 impacts on different actors of the food systems. In a second step, we revisited the data, focusing our attention to the interactions observed between the different types of disruptions reported in these documents, with the ambition to build the first complete impact pathway of COVID-19 on food systems. 4.4.1. Relative importance of COVID-19’s disruptions across the food system Among the 337 documents reviewed, 250 (74%) discuss the impact of COVID-19 on consumers and 278 (82%) discuss the impact of the pandemic on the rest of the food system actors (Table 4 top). Table 4. Number of documents referring to the different groups of actors affected by COVID Groups of actors Number of documents (%) Consumers 250 (74%) Food system actors including producers 278 (82%) Food system actors(a) Food vendors 8 (31%) Wholesalers 8 (31%) Processors 10 (38%) Transporters 8 (31%) Producers 15 (57%) Consumers 18 (69%) Note: (a) analysis based on 26 documents reviewed with the full-fledged framework The more detailed analysis undertaken with the full-fledged version of the analytical framework (Table 4 bottom) suggests that within the groups of actors operating in food systems, primary producers (mainly family-based farming/dairy enterprises, but also fishers, pastoralists, fish-farmers) have received proportionally more attention than any other actors in the systems (e.g. Harris et al., 2020; Aggarwal et al., 2020; Nedumaranet al., 2020; CIMMYT, 2020). Consumers, however, are the group on which the 24 majority of the peer-reviewed articles (69%) have focused their work (e.g. WFP 2020; Headey et al., 2020a; CARICOM et al., 2020; Hamadani et al., 2020; Gaitán-Rossi et al., 2020). Table 5 synthesizes the main issues faced by the different groups of actors during COVID-19 as reported in these peer-reviewed articles, per category. For consumers, the main issues reported were (in decreasing order of importance): the degradation in the choice and diversity of food items available to households (due to the lockdown, reduction of mobility and closure of some of their usual food suppliers) (e.g. Villaseñor Lopez et al., 2020; Casco, 2020; Niles et al., 2020; Ebata et al., 2020; Hamadani et al., 2020); the increase in relative food prices (partially due to the closure of the usual [informal] food suppliers/outlets and/or the increase in prices in the remaining open food outlets) (e.g. Basilico and Figueroa, 2020; Casco, 2020; Hamadani et al., 2020); the disruption in accessing food supply due to the lockdown and restriction in mobility (e.g. Robins et al., 2020; Power et al., 2020; Gaitán-Rossi et al., 2020; Tesfaye et al., 2020); and the loss or reduction of consumers’ income and associated purchasing power due to the closure or reduction in their own business or that of their employers (DNPGCA, 2020; Hirvonen et al., 2020c; FAO-WFP, 2020; Arteaga Garavito et al., 2020). Table 5. The main issues affecting the food system actors as reported in documents Note: (a) as reported in the 26 documents reviewed with the full-fledged framework; (b) percentage of document reporting these issues. Only issues reported by 30% or more documents are listed. While being cautious not to over-interpret these results, it is interesting to notice that those four issues are all related to the access dimension of food security as understood in the Food and Agriculture Organization of the United Nation’s (FAO) original definition; two to the economic sub-component (or affordability) of food: (i) the (relative) increase in food prices/ lower affordability of consumers, and (ii) the loss or reduction in income of consumers and the subsequent decline in their purchasing power; and two to the physical accessibility of food: (i) the degradation in the choice and diversity of food items available, and (ii) the disruption in accessing food supply. In contrast less evidence about the impact of COVID-19 on the other three conventional dimensions of food security (availability, utilization [quality and safety] and stability) was revealed by the analysis, even though access to food supply could also be re-interpreted as a stability issue. Group of actors affected and main issues reported (a) Documents(b) Consumers ▪ degradation in choice and/or diversity of food items available 56% ▪ increase in (relative) food prices/lower affordability 50% ▪ disruption in accessing food supply 44% ▪ loss or reduction of income and associated purchasing power 44% Primary producers ▪ disruption in upstream input supply chains 67% ▪ decline in business profitability / revenues, incomes 60% ▪ reduction in laborer/workers availability 40% ▪ reduction in demand for farm products 40% ▪ loss of or reduced connectivity with established business partners 33% Mid-stream food system actors ▪ disruption of business practices 39% ▪ forced closure of business 37% ▪ loss of connectivity with their established business partners 31% ▪ disruption in upstream input supply chains 31% ▪ reduction in downstream demand for products 30% 25 For primary producers, the main issues reported in the documents were: disruption in the upstream input supply chains (fertilizers, seed supply, spare machinery parts, etc.) (e.g. Claudino, 2020; Termeer et al., 2020; Robins et al., 2020; MSSRF, 2020); decline in business profitability and associated revenues (e.g. Macías-Chóez et al., 2020; Harris et al., 2020; Quiroga Mendiola et al., 2020; Niang and Faye, 2020); reduction in labourer/workers availability due to mobility restrictions, increase in public transport costs, or fear of exposure to virus (e.g. IFAD et al., 2020; Macías-Chóez et al., 2020; DNPGCA, 2020); reduction in demand for their products (e.g. Varshney et al. 2020; Harris et al. 2020; FAO 2020b), and loss of or reduced connectivity with their established business partners or consumers (e.g. Ebata et al., 2020; Nedumaran et al., 2020). For mid-stream food system actors (including processors, transporters, wholesales retailers, and vendors), the documents reported issues occurring primarily around the disruption of business practices due to lockdown and mobility restriction affecting their own activities and the activities of their upstream and downstream partners/clients (e.g. Varshney et al. 2020; Dai et al. 2020; Termeer et al. 2020); the partial or complete closure of their business (imposed by safety or sanitary decrees/regulations8) (e.g. Fang et al. 2020; Burkart et al. 2020); the loss of connectivity with their business partners or customers (e.g. Fang et al., 2020; Tesfaye et al., 2020); the disruption in upstream input supply and subsequent effects on prices or quantity/accessibility/quality of inputs (e.g. World Bank, 2002b; Rosen, 2020; Mogues, 2020); and the reduction in demand for their products leading to decline in business and profit/revenues (e.g. Tounkara, 2020). Again, while avoiding over-interpreting the data, we can notice several issues identified as being common to both primary producers and the other actors of the food system. These include disruption in upstream input supply and subsequent effects on prices or quantity/accessibility/quality of inputs; reduction in demand for their products leading to a decline in business and profit/revenues; and loss of connectivity with their business partners or customers. Overall, Table 5 provides an overview of the specific types of impacts different food system actors have been facing in different parts of the world since the outbreak of COVID-19. However, it does not allow for comparison across these groups (at least not in a rigorous way) and therefore does not permit determining with certainty whether one group of actors has been more at risk than the others. This means, it is not possible to confirm some of the statements made by experts that “Direct impacts on farm populations and farm production will be much smaller than on the food supply chains downstream and midstream” (e.g. Reardon et al., 2020, p.79). A certain number of the documents also discuss in greater detail some specific commodities and/or value chains. The data (summarized in Table 6) indicate that livestock and rice were the focus of a larger number of analyses (e.g. FAO, 2020c; Burkart et al., 2020; Balié and Valera, 2020; Arouna et al., 2020) compared to other value chains. This situation does not necessarily mean that these two subsectors have been more exposed or more vulnerable to COVID-19 than others, but rather that they received more attention. One probable reason for this greater attention (at least for livestock) is the zoonotic origin of the pandemic and the subsequent pressing need to better understand some of the possible root causes of the outbreak of such zoonoses9. 8 Many local food markets have been forced to close for instance because of perceived high risks of COVID-19 due to the density of people and animal products and/or low abilities to enforce hygiene and social distancing measures. 9 Other possible reasons include the fact that livestock need to be cared for and fed every day and may as such be more sensitive to mobility disruptions of people and feed, than, say cereals (with the exception of harvest season). 26 Table 6. The different subsectors/commodities discussed in the documents Commodity / value chain Frequency Commodity / value chain Frequency Aquaculture 4 Livestock 12 Cocoa 1 Mango 1 Dairy and Milk 3 Potato / sweetpotato 1 Family home garden 1 Poultry 2 Fisheries 2 Rice 6 Fruits and vegetables; Meliponiculture 3 Wheat flour, pork and Chinese cabbage 1 The review of the documents also reveals that most of these sub-sectors/commodities-focused analyses were descriptive in nature. Their main objective was simply to depict the impacts of COVID-19 on the actors of these specific commodity value chains. They did not attempt, on the other hand, to provide any specific framework that would allow comparison between different commodities and establish for instance whether particular sub-sectors or value chains are more (or less) vulnerable than others to the disruptions induced by COVID-19 and why. We argue in Section 5.1.1 below that this question – what makes a specific commodity/subsectors more vulnerable to COVID-19? – is however a critical question to address if we want to be in the position to reduce the risk of occurrence and the impacts of future crises similar to that triggered by COVID-19. 4.4.2. Impact pathways of COVID-19 on food system actors The information and data provided in the sections above offer some first elements toward a more comprehensive account of the impacts of COVID-19 on food systems and their actors. So far, however, the analysis has been primarily static, in the sense that although the assessment identifies and maps out the different types of disruptions observed in various components of the food systems, no specific attempt was made to link together these different disruptions or to determine whether some degree or forms of interactions between them could be established. Yet the literature on food systems is clear about one point: one of the main characteristics of food systems is the interconnectivity (forward and backward linkages) and associated feedback loops that exist between the different actors and components of the systems (Ericksen, 2008; HLPE, 2017). In his recent review on food system resilience, Béné (2020) argues that this interconnectivity should be a central element of any food system resilience analysis as it is at the origin of what he refers to as the ‘ripple effects’- the fact that when one group of actors is affected by a shock, the effects of that shock rarely remain confined to that group. Instead the effects and the subsequent responses they trigger from different actors are likely to ripple upward and downward and affect other actors along the supply chain –see his figure 2, p.816. We attempted to explore this dynamic aspect of food systems in this section. Using the information included in the 26 documents analysed with the full-fledged framework, we were able to construct the impact pathway of COVID-19 on food systems and their actors. To build and represent this impact pathway, we developed a Sankey diagramme (Schmidt, 2008) where the relative importance of each connection between the different potential impacts is used to identify directionality and intensity between the elements of the pathways. To structure the analysis, we started with the set of 25 potential effects as listed in Table 1 above, organized along the four proposed steps of causal pathway: Direct effects and responses → Immediate consequences → Subsequent repercussions → Final impacts. The levels of relative importance/contribution of each connection were then estimated by computing the number of 27 times two consecutive events were observed in the 26 documents. Fig.7 presents the result of this Sankey analysis. The diagramme reveals a series of important points. First, the analysis confirms the importance of adopting a systemic and dynamic approach that does not just describe the different types of impacts affecting the components/actors of the food systems, but also identifies the interactions (links, feedbacks) that exist between these impacts. We were able to identify 56 forward and backward links observed across the whole system. As such, the analysis confirms the relevance of the concept of ripple effects (Béné, 2020). Second, those different interactions form not just one single impact pathway but a whole combination of intermingled, non-linear paths characterized by multiple ‘branches’ and loops. Those interact with, and reinforce each other, intensifying and combining their individual effects into a relatively complex and dynamic intricacy of causal effects. The existence of these multiple impact pathways demonstrates why an individual mitigation intervention focusing on one single issue is unlikely to be effective. Instead, mitigation interventions that acknowledge the existence of these multiple impact pathways and embrace a system perspective would have higher chances of being effective. Third, having uncovered the existence of multiple, non-linear impact pathways, the analysis also reveals that not all these pathways are equal in importance/contribution. Some of them indicate very strong causal links while others are associated to more minor contributions. This observation is not in itself surprising - one should not expect all causal mechanisms to be of the same intensity- but it provides us with very useful information regarding the overall dynamic of these ripple effects. In particular, the fact that one of the most prominent pathways (between ‘loss of job/reduction in income/revenues’ and ‘degradation in food choice and diversity’ –see Fig.7) is not only very short but also involves all actors (primary producers, midstream actors and consumers10) is worth noticing as it confirms that the impacts of COVID-19 have been general, affecting everyone and not just some particular subsectors or specific groups of actors in the food system. In fact, we would argue that although originated from data directly derived from a food system focused analysis, this particular pathway is appropriate to describe the impact of the pandemic beyond the food system per se. In essence, although there is now a consensus in the international literature that one of the major consequences of COVID-19 is likely to be an increase in food insecurity (HLPE, 2020; FSIN and GNAFC, 2020; FAO-WFP, 2020), the main causal mechanisms involved (‘loss of job/reduction in income/revenues’) operates outside – or beyond – the food system itself. In sum, the impact pathway analysis suggests that while the final outcome is a substantial deterioration of food security, the cause is the disruption of the global economy rather than the collapse of the food system. 10 as indicated by its color code violet in the diagramme. 28 Fig.7. Impact pathways of COVID-19 on food systems and their different actors based on a Sankey approach. The thickness of the connecting lines is proportional to the number of times a connection between two elements was mentioned across the different documents. Numbers in the diagramme refer to the numbering system used in Table 1. 1. Disruption in upstream supply chain 2. Disruption in actors’ own activities 3. Loss of or reduced connectivity 4. Reduction in labour/ workers availability 5. Drop in profitability 7. Increased wasted food 8. Forced closure of business due to safety or sanitary decrees c. Safety or sanitary decrees/regulations 9. Degradation in Rules of Law 11. Increased abuses against marginalized individual or groups 12. Drop in perceived self-efficacy or agency 16. Domestic violence 15. Increased exposure 21. Increased risk of consumption of unsafe food Affecting consumers (including producers, workers and mid-stream food system actors) Affecting producers, workers and food system mid-stream actors Affecting producers,