Estimating the cost of SDG 2

We need more public spending to achieve the United Nations’ Sustainable Development Goal 2—Zero Hunger—by 2030. But how much more? And how do we distribute the spending?

 

Photo: Adapted from image by Tomás Guardia Bencomo, via iStock.

Ceres2030 is building a cost model—a mathematical and economic model of the world over time—to answer these questions. Prior work on the cost model found that we are not on track to end hunger by 2030, but we can meet that goal if we have additional resources, prioritize countries with the highest need, and use the right balance of our best interventions. 

The Ceres2030 cost model will also include the costs for three critical dimensions of SDG 2: ending hunger, ensuring sustainability and doubling small-scale producer productivity.

The cost model estimates the total additional spending needed to reach SDG 2, including from private investors, domestic governments, and international donors. The goal of the Ceres2030 cost model is to inform international donors how much additional spending is needed to reach SDG 2, which countries are highest priority for additional commitments, and what mix of investments will yield the best results.

Country profiles suggested by the cost model are intended to be a first iteration. The cost model will produce profiles for those countries where there is sufficient data.

What can we estimate?

Modeling can help us to see how different parts of an economy are related, and how spending on agriculture affects a wide range of economic and social outcomes. For example, when a farmer’s net disposable income increases because a storage bin is put at her disposal, reducing post-harvest losses and allowing her some control over when to sell the crop, the model can track the effects of the intervention and the resulting income gains on the wider economy, from the household to the national level.

Any estimate for additional spending needed for SDG 2 can only be as good as the data available. For this reason, we cannot provide a cost estimate with exact coverage of all the aspects of SDG 2. 

We also cannot simulate the political contexts that can affect hunger. We do our best with available data, and we will focus our work on the three major SDG 2 targets: SDG 2.1, 2.3, and 2.4.

The world cannot wait for perfect information to act. SDG 2 donors need a baseline now to build an action plan for additional spending; we do our best to provide estimates.

How is the cost model built?

Determining how much money is needed for SDG 2, where it should be spent, and how it should be spent means we will need to account for economic relationships from the micro level to the macro level from now until 2030. To account for these complex relationships, the team uses a dynamic multi-country, multi-sector Computable General Equilibrium model built on household data, adapting from the MIRAGRODEP economic model.

The model is a set of equations linking inputs to outcomes, where targets are achieved using numerical optimization. It processes information for each country and for each year, infusing public spending and allocating it among countries and interventions so that the three goals are reached by 2030 at minimum cost.

Micro Level: People are the base of the model. The Ceres2030 model is built on household survey data, a bottom-up approach that allows the model to identify and target hungry, poor, and small-scale producer households directly. This household targeting is key to spending efficiency. 

Meso Level: Our model also incorporates “meso level” data. These involve micro variables, like people’s food consumption and incomes, which affect and are affected by regional and sectoral circumstances, such as food prices, fertilizer prices, and wages.

.

Macro Level: Our model also includes macro level linkages to account for how changes in markets or government policies can affect a small-scale producer’s income or the food a hungry household can afford. For example, an increase in maize yields in Tanzania could lower prices for maize in Ethiopian markets through trade, helping hungry households in Ethiopia meet their caloric needs. At the same time, Ethiopian farmers who grow maize could suffer income losses from the competition.

Interventions influence other variables through a set of parameters. Interventions are represented by parameters in the model’s equations. They can directly or indirectly affect variables at any level. The numerical optimization process allows the model to choose the best bundle of interventions to reach the three goals at minimal cost.

After the model is run, the team can extract its dependent variables to answer our three questions. These variables will tell us:

|1| The additional amount of public spending the model used to accomplish the costed SDG 2 targets.

|2| How the model distributed the additional spending among countries.

|3| How the model allocated the money among types of interventions in each country.

Data Sources: The model uniquely leverages harmonized data from a variety of sources. Household data relies on the World Bank’s Living Standards Measurement Study and other sources, while meso- and macro-level data comes primarily from the GTAP database and FAOSTAT.

How will the results be used?

Ceres2030 aims to improve people’s lives in a sustainable way through effective public spending on SDG 2. The model reflects this through the targets it sets and the relationships it includes. The model’s results therefore provide an initial strategy for additional spending by SDG 2 donors, and we can then refine the results through including new data as available, and following further consultation with national governments and other stakeholders.

 

The results from Ceres2030 can help donor governments determine where to direct their investments, and how this spending will affect these economies. It will then be up to donor governments to consult with recipient government experts and other stakeholders to refine how much money to allocate for each type of intervention.