As the world works to achieve the United Nations Sustainable Development Goal 1 (SDG1): End poverty in all its forms everywhere, by 2030, it is critical to understand its strong interrelationship with two other SDGs: End hunger, achieve food security and improved nutrition, and promote sustainable agriculture (SDG2) and Reduce inequality within and among countries (SDG10). Progress in one area depends on efforts in the others, and all three goals are part of the global public health agenda. Famines, which can be understood as “acute episodes of extreme hunger that result in excess mortality due to starvation and hunger-induced diseases” [1, 2], represent one of the most serious consequences of poverty and inequality, but they also contribute to the further immiseration and marginalization of affected populations. Recognizing these interrelationships has helped transform our approach to famine in the past and could provide a starting point for exploring new advances for the future.

In his landmark 1981 book, Poverty and Famines: An Essay on Entitlement and Deprivation [3], future Nobel Laureate Amartya Sen proposed a revolutionary shift in our understanding of these crises. He argued that famines often occur not from a lack of availability of food, but from the inability of certain populations to access it. While this insight suggested that poverty was a principal reason that people experience famines, Sen pressed for a more nuanced understanding of why certain groups are more at risk of starvation than others during a crisis. He suggested that it was an inability to ‘command’ adequate food because of a failure of entitlements that led to mortality. His entitlements approach laid the foundation for decades of research that have extended, contested, and moved beyond his views by, for example, emphasizing the importance of the processes that lead to famines [4, 5], highlighting the critical role of politics and power in their causation and differential impacts [6], and addressing the global dimensions of the crises [7].

Although in the early 2000s trends suggested that these crises were diminishing in number and scale [2, 8], famines  have killed hundreds of thousands [9] of people in the first two decades of the twenty-first century and left a legacy of livelihood  damage, emotional trauma, physical impairment, and social disruption. In 2021 and 2022, there have been concerns about the risk of famine in Afghanistan, Ethiopia, Nigeria, Somalia, South Sudan, and Yemen [10]. As of November 2021, it was estimated that more than 45 million people in 43 countries were experiencing emergency food security conditions [11]. With the looming threats of climate change, continued conflict, and emerging pandemics, the world is likely to continue to face the risk of famine in the years to come, making it critical to understand with greater certainty when and where famines will occur, and which populations will be at greatest risk, and thereby help to trigger appropriate early action [12].

Just as the risk of famine appears to be increasing again globally, three trends—related to famine theory, measuring and modeling, and humanitarian practice—are converging to offer an opportunity for a step-change in our ability to understand and forecast these crises. While the focus in the discussion will be on ‘famine,’ the trends (and the proposed initiative) apply to food and nutrition security crises more generally.

First, drawing on previous literature, academics have recently suggested that famines can be understood as complex systems and have identified conceptual models that describe their evolution from formation to collapse [13]. This systems approach to famine offers new possibilities for understanding the dynamics of these crises and could help in defining driving forces, characteristic milestones, and well-tailored metrics, which would facilitate the translation of these concepts into quantitative and analytical models and produce famine forecasts.

Second, there have been rapid developments in the fields of primary data collection and computational, mathematical, and statistical modeling. Real-time data collection capabilities, through electronic devices and crowdsourcing, have accelerated and changed the possibilities for gathering information. Innovations in predictive analytics make it possible to handle the large volume of complex data required to model and forecast famines [14]. At the same time, a suite of different types of models—systems dynamics, agent-based models, stochastic models, and regression time series models—offer a wide range of approaches to challenging problems and have gained in sophistication, accuracy, and applications. These developments have enabled progress on critical problems as complex as climate change and the COVID-19 global pandemic. They offer hope for meaningful advances to deepen our understanding of famines as systems and develop robust conceptual and forecasting models.

Third, in terms of global reach and innovation, humanitarian practice has evolved in ways that could both drive these efforts forward and translate them into significant, real-world impact. Early warning analysts have continued to improve systems for predicting food insecurity crises and famines through widespread monitoring that combines sophisticated data analysis with on-the-ground insight. The Integrated Phase Classification (IPC) platform, developed in 2004, offers comparable analyses of food insecurity and malnutrition situations globally and provides a widely accepted process for determining whether a famine has occurred based on an internationally agreed definition. Humanitarian agencies are transforming responses through greater use of cash, integration into social protection systems, and emphasis on early action. United Nations Security Council Resolution 2417, which requests regular updates on crises and strongly condemns the use of hunger as a weapon of war, has reinforced accountability for famine prevention. In all these ways, it is clear that humanitarian agencies have the capability to create, adopt, and apply innovations to address crises on a global scale.

The convergence of these trends could permit two important changes in our approach to famine. The first is to help researchers create models of famine as complex systems and describe their evolution from formation to collapse, as has been done for hurricanes and infectious outbreaks. Such models could help us gain new insights into questions such as: What are the components of a famine system? What are its spatial and temporal dimensions? How do the various parts of the system interact to produce critical outcomes such as malnutrition and mortality and why are certain populations especially vulnerable? Based on these insights, the second change would be for experts to better identify the signals of famine formation and therefore improve forecasts. If successful, these forecasts could help (as part of a wider set of tools) reduce uncertainty about the likely occurrence of food and nutrition security crises, contribute to timelier, more targeted, and life-saving action, help deter famine creation, and perhaps spark new fields of inquiry. The efforts to model and forecast hurricanes and infectious outbreaks suggest these kinds of benefits are possible if there is an iterative process that continually promotes learning from experience.

In moving toward a ‘Poverty and Famine 2.0 approach,’ we identify measuring, modeling, and forecasting as three essential and interrelated processes for gathering insight based on the goals and primary activities of each. Measuring aims to collect data to create information and knowledge. Modeling aims to process data, information, and knowledge to form casual paths, rules, and systems thinking, along with their uncertainties. Forecasting aims to infer future unknown situations based on data, information, knowledge, and system thinking. To describe a process of estimating future unknown situations, the terms ‘forecasting,’ ‘prediction,’ ‘projection,’ and ‘prognosis’ are often used by researchers and practitioners interchangeably. We are also making a distinction between forecasting and predictions suggesting that forecasting is an extrapolation of the past into the future, while predictions and projections are typically subjective and judgmental in nature. While both approaches are useful in considering changes that may take place in the future, ideally, forecasting is free from intuition and personal forecasters’ biases, whereas prediction is based on judgment. In short, all forecasts are predictions but not all predictions are forecasts.

Recognizing the potential of these three converging trends, especially the power of accurate relevant data and sound analytical solutions, several actors have engaged in pioneering efforts to use predictive analytics to improve food security and nutrition forecasting based on data sources, machine learning, and novel algorithms [14, 15]. For example, the Famine Early Warnings System Network (FEWSNET) has partnered with scientists to incorporate climate models into their scenario-building [16]; the World Bank has developed sophisticated algorithms to forecast food crises globally [17, 18]; Consortiums have used models to understand the risk of malnutrition; and the United States Agency for International Development (USAID) has used economic models to analyze the benefits of investments in resilience [19]. To a large extent, however, these efforts have been fragmented. The learnings from one are not always widely shared to inform the efforts of others. They are also partial in that most do not try to model famine itselfe, focusing instead on forecasting IPC phases, food insecurity indicators, or other outcomes. Some incorporate climate science models but have not yet found ways to bring in political and social dimensions to algorithmic forecasts. Recognizing that we are in the early stages of an emerging field, the IPC and the Global Network Against Food Crises have made famine forecasting a key component of their long-term strategies.

However, we are cognizant of the daunting challenges and genuine risks associated with pursuing famine modeling and forecasting. Modeling famine is complicated by the complexities of interacting economic, social, environmental, and political systems and the challenges of characterizing human action and decision-making. The track record on forecasting social phenomena and conflict has sometimes been discouraging [20, 21]. These efforts also require large amounts of data from some of the most challenging contexts in the world. As a result, there is a danger that the international community will expend substantial resources and time on a venture with highly uncertain results. Such an initiative may also inadvertently reinforce the notion that there is a technical solution to famine and divert attention from pressing political issues central to its prevention. Moreover, because of the complexity of the models, they may hide assumptions and biases, reduce transparency, and perpetuate inequalities, creating ethical concerns [14, 15]. Relatedly, the emphasis on data gathering activities and modeling exercises could lead to data-driven products that are increasingly divorced from local realities, and the experiences, views, and inputs of affected populations. Finally, while these efforts may contribute to improved forecasts, they do not directly address the critical challenge of translating early warning into early action [12].

We see potential responses to these concerns. For example, the cost of investing in these efforts is substantially smaller than the resources that are required to address current crises and that could be saved through the insights accurate forecasts could provide. Moreover, modeling and forecasting should not be seen as a substitute for social and political efforts (or existing early warning systems), but rather a complementary tool. Deliberate, joined-up approaches could also help address ethical concerns, engage affected populations, make the link to early action, and deal with data issues such as the need for reliable data repositories, tools to abstract and examine data, intra-agency and intergovernmental agreements, and data-curation standards and model-sharing protocols. Considering these challenges and potential responses, we offer six principles that could guide a famine modeling and forecasting initiative in terms of both content and process.

  1. (1)

    Focus on both modeling famine itself and forecasting its occurrence. It is difficult to forecast what is not well understood. It is also difficult to act when the forecast is poorly focused and not well explained to end users. If we have better conceptual and analytical models of the formation, evolution, and collapse of famines as complex systems, it should help our attempts to better predict their occurrence. Likewise, progress on forecasting will point to new areas to explore in understanding the dynamics of these crises.

  2. (2)

    Integrate multiple dimensions. While we may be more advanced in the use of climate and economic data in models, it will be important to think creatively about how political and social dimensions can be better integrated. Human behavior at any stages of decision-making could alter forecasts.

  3. (3)

    Take an inclusive approach with global scope. The initiative requires a joined-up effort that brings together the insights of affected communities, humanitarian practitioners, famine theorists, public health professionals, data holders, and modelers across the globe. Similarly, a wide range of tools employed in science and practice—from machine learning and predictive analytics to intervention strategies to participatory approaches to simulation exercises—should be utilized.

  4. (4)

    Be realistic, learn, and invest for the long-term. The history of using models in other fields—whether climate change or hurricanes or pandemics—suggests that they can provide early warning and deeper understanding. But the process of developing and deploying sophisticated, accurate models is challenging and the immediate payoffs uncertain. Results from conceptual developments, data analysis, and modeling need to be widely disseminated and built upon in an iterative process by a broader community. Decades may be needed to achieve the initiative’s full potential.

  5. (5)

    Address ethical and other concerns. These models and forecasts entail a number of potential ethical risks that could undermine their usefulness and inadvertently perpetuate biases and inequalities. These concerns should be articulated and addressed upfront by the wider community involved [12, 14, 15], perhaps through development of protocols and procedures to guide the process.

  6. (6)

    Look beyond models and forecasts. Models and forecasts are a potential tool for better understanding famines. But they should only be viewed as a part of a wider famine and public health agenda involving theoretical insights, enhanced early warning and action approaches, improved practice, and political efforts to prevent these crises more effectively.

Forty years after the publication of Sen’s seminal work, famine studies have identified its shortcomings and evolved in new directions, but his concern about the relationship between poverty and famines remains and challenges the global community to take creative approaches to more systematically address these crises that continue to threaten the lives and well-being of humans. Given recent trends, we believe it is an opportune moment to make a step-change in our efforts by investing in a 2.0 approach to crisis modeling and forecasting––thereby also supporting the achievement of the interrelated SDGs on poverty, hunger, and inequality.

The Journal of Public Health Policy is joining Springer in seeking submissions to a new Collection on Reducing Poverty and Its Consequences, in support of the International Day for the Eradication of Poverty. This multi-journal Collection aims to synthesize and integrate social, behavioral, and public health perspectives on systemic structures bolstering poverty and inequality, poverty-reduction interventions, as well as gaps in our knowledge and future research directions. In promoting the UN Sustainable Development Goals, we see the need for proactive dialogs across multiple stakeholders, forward-looking intervention study designs, understanding of long-term consequences of hunger and poverty, as well as the need for modeling and forecasting incorporating human behaviors beyond the technical solutions. We invite readers and contributors to share your thoughts, findings, and experiences through this new multi-journal Collection.