Are SDGs 1 and 2 diverging? New data innovations to better monitor poverty and food security

Tracking progress towards the Sustainable Development goals is a priority for the development community, and the subject of the recent SDG Atlas. Two of these goals, SDG1 and SDG2, pertain to eradicating poverty and achieving zero hunger. Recent official data on the indicators used to track SDG progress reveal a concerning trend: while reported global SDG1 data improved consistently up to the onset of the pandemic, SDG2 data veered off track. This raises the question: Why did food security deteriorate when extreme poverty improved?

The answers are not straightforward and require more granular data, better data coverage, improved data standards, and improved transparency.

Understanding SDG1 and SDG2: Eradicating Poverty and Achieving Zero Hunger

SDG1 strives to end poverty in all its forms everywhere, while SDG2 focuses on eradicating hunger, attaining food security, enhancing nutrition, and encouraging sustainable agriculture. Each goal encompasses sub-goals, matched with specific development indicators. To gauge progress towards SDG1.1—eradicating extreme poverty by 2030—the UN uses official World Bank estimates of the share of the global population falling below the International Poverty Line, currently set at $2.15 in 2017 PPP terms. For SDG2.1 — Zero Hunger — the UN relies on FAO’s Prevalence of Undernourishment (PoU) to measure progress towards SDG 2.1.1 and Prevalence of Moderate or Severe food insecurity for SDG 2.1.2.

While the divergence is more pronounced in the countries with lower data coverage, missing data alone is insufficient to explain the divergence. There may be economic factors at play that are more pronounced in data-poor regions, which can only be captured by introducing new data and analysis tools.

For instance, the poverty rate is 4.5% in the high-data coverage sample in 2019 versus 8.5% globally, reflecting the fact that data-poor countries also tend to be poorer economically.

New analysis and tools to monitor SDG1 and SDG2

The divergence in primary SDG data underscores the importance of strengthening statistical capabilities for measuring progress towards the goals, notably in low and middle-income nations that have weaker data systems while significantly contributing to global poverty and food insecurity. It also points to the risks of relying solely on a few aggregate indicators for gauging complex changes in living standards, and the importance of better understanding different methodologies for generating global estimates.

Challenges in tracking data include lack of data collection infrastructure, including access to reliable survey methods and adequate data storage. Researchers at the World Bank have been at the forefront of efforts to enhance monitoring and data collection in data-poor regions through alternative strategies. Recognizing the critical role of comprehensive data in promoting development, the World Bank has initiated several innovative projects that leverage machine learning methods and alternative data sources to produce more granular and more timely statistics.

One such initiative involves the utilization of satellite imagery and remote sensing data to supplement survey data through the use of machine learning methods that predict poverty and other indicators. Existing evidence indicates that this approach significantly increases the accuracy and precision of survey estimates in the cross-section, can enable poverty estimates to be reported at more disaggregated levels as in Tanzania and The Gambia. While geospatial estimates yield similar patterns to census-based estimates, there are marked discrepancies between the two approaches in particular areas, and the use of geospatial estimates are a second best option when current census data are not available. It is worth exploring further which remote sensing indicators can help nowcast changes in well-being and the impact of weather shocks. Another initiative has involved filling gaps in household surveys for countries which are home to many of the world poor using survey-to-survey imputation. This has been applied to countries such as India and Nigeria.

In an effort to enhance innovation in food security analytics, the use of machine learning methods to monitor food insecurity risks has been explored. This initiative involved the development of two models: one focused on predicting outbreaks of food crises at the sub-national level for 21 countries, and another employing a stochastic model for 15 countries to simulate longer-term risks, country-level risks. Both models utilized a dataset containing food prices, conflict events, and remote sensing variables to capture critical weather and climate-related factors. Additional efforts have been directed towards expanding the scope to include a wider range of countries. For example, novel modeling techniques incorporated machine learning models within a simulation of historical global food crises, enabling the projection of future food insecurity risks in alignment with economic outlooks provided by the IMF. This model, which encompasses all countries found in both World Bank and IMF databases, has been used to assess financing needs in IDA (2020) and IDA (2021), and has been applied in the Middle East and North Africa (MENA) Regional Economic Update. The projections are now updated three times a year and made available to complement official statistics.

Ongoing efforts also focus on tracking food prices at a local level. Methods have been developed to impute price data to estimate food price inflation. These approaches can be carried out with high frequency data in real time. This allows estimating inflation of important goods such as staple food items, even in areas affected by conflict and violence and where official data are not reported. For instance, in Yemen where these data are used to support the Joint Monitoring Report in collaboration with the Global Alliance for Food Security (GAFS):

The data are available and updated on a weekly basis for more countries.

These data-driven approaches not only bridge the data gaps but also enable more accurate and timely decision-making, ultimately contributing to more effective development strategies and improved outcomes in data-poor regions.

As we track global SDG data closely, it is important to remember that progress towards these goals is inherently complex and requires action on common themes. Reaching the SDGs mandates international cooperation, data-driven policy formation, transparency, and a shared determination to realize a world free from poverty and hunger.