How much would it cost to end hunger and malnutrition by 2030? That is the key question addressed in the 2024 State of Food Security and Nutrition in the World (SOFI). It is also a question governments, donors and policy makers have been asking since they made the global commitment in 2015 to achieve Sustainable Development Goal 2 (SDG 2) by 2030.  

Unfortunately, the research community does not provide a simple answer. We reviewed eight reports published between 2015-2024 and found that the cost estimates range from a total of USD 86 billion to USD 4 trillion. This huge difference creates confusion among the intended audiences and serves as an obstacle to effective action.  

So, which one is right? 

The answer largely depends on the question the researchers asked. Three of the eight reports reviewed are numbered in the trillions of dollars because they go beyond SDG 2 to include a range of other linked SDGs or focus on agrifood system transformation. The remaining five reports are numbered in the billions of dollars because they focus on the targets in SDG 2. However, even within the five reports focusing on SDG 2, the cost ranges between a total of USD 86 billion to USD 760 billion. This is due to difference in the researchers’ choice of targets and the quantification of those targets, baselines and data sources used for 2030 projections, and modelling approaches.  

Some of these differences are healthy, like using different modelling approaches to answer a similar question, as a way of increasing the confidence of the model results. Other differences are problematic, like using of different data sources for the baselines, and lead to confusion and inaction.  

To address this, researchers should 1) standardize the use of data sources for modelling; 2) align baseline projections, where they exist and are published; 3) use model approaches that are best design to answer the specific policy question; and, 4) resolve glaring contradictions through stronger peer review. This will create a certain level of consistency between research reports enabling policy makers to more easily compare the studies and make evidence-based decisions. 

The main differences between the reports 

Choice of targets and their quantification  

While the eight reports reviewed all aim to end hunger, their targets vary significantly. The most narrowly targeted report, the Global Nutrition Report (2021), only focuses on specific nutrition interventions, which is sub-target SDG 2.2. Others have much broader targets that encompass multiple SDGs or target agri-food system transformation. This is the case for the Food and Land Use Coalition (2019) report which covers all SDG 2 targets, as well as targets linked to SDG 5, SDG 7, SDG 12, SDG 13, SDG14 and SDG 15. As a result, the cost estimates vary between the trillions of dollars for the reports seeking to achieve multiple SDG targets and the billions of dollars for the reports seeking to only achieve SDG 2 targets (see Figure 1). 

Figure 1. Comparison of SDG 2 modelling reports, SDG sub-targets, and additional costs

The reports also vary in the quantification of the core target of eradicating chronic hunger.  While all reports use the prevalence of undernourishment (PoU) as an indicator, the reduction targets for PoU levels vary from 0 percent to 10 percent. This leads to wide differences in cost estimates since reducing hunger to zero percent is significantly costlier than leaving 3, 5 or 10 percent of the population hungry. While there are important justifications for the different targets, researchers poorly communicate both the percentage selected and the justification, leaving policy makers with the impression that the additional spending estimated will achieve zero hunger or confused about the implications of only reducing hunger to 3 or 5 percent. 

Baselines and data sources used for 2030 projections 

Each report used a different baseline and projection for the number of undernourished people in 2030 in a business-as-usual scenario. This leads to baseline projections ranging from 475 million to 700 million people facing undernourishment in 2030 (see Figure 2).

Figure 2. Projected number of undernourished people by 2030 under a business-as-usual scenario and the number of people lifted out of hunger due to additional investment (in millions) 

It is not unusual for reports to have a different baseline depending on their year of publication. The annual SOFI report updates the number of people suffering from hunger and malnutrition every year along with the baseline projections. However, the situation becomes complicated when reports do not clearly provide the sources for their baseline projections, and yet more so, when the baseline projections differ for reports published during the same year, such as Food and Land Use Coalition (2019) and IFPRI (2019). 

Another variation is found in the use of different climate change scenarios. The Intergovernmental Panel on Climate Change’s (IPCC) Shared Socioeconomic Pathways (SSP) offer five pathways of climate change scenarios based on projected global socioeconomic changes up to 2100. The reports that used these climate scenarios chose different pathways because of their different assumptions about future global development impact the estimated costs of the models. The use of a pessimistic scenario with high population growth, economic stagnation, and significant income inequality, will result in higher costs due to the need to address more severe social and economic challenges to end hunger. 

These choices were not clearly communicated in the reports. While there are strong arguments for using different climate change scenarios in hunger-related modelling exercises, researchers need to be much clearer about the scenarios selected and their rationale, to help policy makers make sense of the results. 

Modelling approaches 

The eight reports reviewed estimate use three different modelling approaches: computable general equilibrium (CGE), partial equilibrium, and the marginal abatement cost curves (MACC). The use of different models approaches to answer similar questions is positive. As different models have different strengths, using multiple different model approaches enables policy makers to have a more complete answer to their questions. For example, CGE models are typically better at simulating macro-economic scenarios, such as trade shocks, while partial equilibrium models are preferable for simulating productivity and yield scenarios for crops or forests at the sub-national level. Policy makers often do not know which model approach is best to answer a question and modellers should educate policy makers on the usefulness of the different modelling approaches rather than trying to promote their own model approach as superior to others .   

What can be done? 

Four simple steps can help narrow the gap. This does not mean that there will not be differences, but it will mean less confusion, more alignment, and clearer communication to the policymakers asking these questions. This will allow policy makers to more easily understand and compare the studies and assess the relevant actions to take.  

1. Standardize the use of data sources  

The research community needs to agree on the data sources for the key indicators where robust data exists, such as for hunger, poverty, population, and gross domestic product (GDP). For the prevalence of undernourishment, the most recently published data from SOFI should be used. Similarly, the latest data from the World Bank should be used for the prevalence of extreme poverty.   

Unfortunately, robust data does not exist for all the SDG indicators and all the questions that policy makers need answers to. This should not be an excuse for inaction. Sufficient data exists to support evidence-based decision making and where less robust data sources exist, they can be used with clear communication and justification.  

2. Align baseline projections where they exist 

The research community needs to align the baseline projections under a ‘business as usual’ scenario, where they exist. This includes, for example, projections for population growth, GDP growth or undernourishment. These figures should be taken from official and agreed-upon sources.  

Since 2022, SOFI has been publishing the business-as-usual projections to 2030 using the MIRAGRODEP CGE model. The research community should agree to use this baseline for future modelling work on hunger and SDG 2.  

3. Apply the most relevant models 

Each of the models used in the eight reports has its benefits. However, no single model can address all research questions. For this reason, the research community should use the modelling approach most suitable for the question being asked, rather than promoting their own model approach. Modellers should also educate policy makers on the usefulness of the different modelling approaches and help them choose the right one. 

4. Resolve glaring contradictions 

Before publishing their work, researchers should share their results within the research community to resolve glaring contradictions. This will not only help bolster the quality of the research but help understand contradictions and differences with other reports. 

Conclusion 

The research community has an important role in guiding governments, donors and policy makers towards achieving global goals. Policy makers rely on researchers and accurate cost estimates to better understand the scale of the problem, what needs to be done, and how much it will cost. Cost estimates allow policymakers to allocate limited resources efficiently, set realistic goals, and avoid expensive mistakes. 

However, currently, cost estimates vary between USD 86 billion to USD 4 trillion. While some of these differences can be easily accounted for, some are more difficult to parse through and leave policy makers confused. This undermines the credibility of the research community.  

Policy makers will have more confidence in results when multiple researchers reach similar conclusions using different models and approaches. To succeed, the research community needs to agree to certain guidelines to ensure consistency between reports and allow for easier comparisons. 

We may not be able to know exactly how much it will cost to end hunger. But we need to get as close as possible to effectively guide action.