Effective policymaking depends on high quality and timely development data. However, collecting poverty data remains an expensive and complex process. Household surveys are the main way to measure household income and expenses, but they can cost millions of dollars. Furthermore, poverty estimates often take years to be published (Lanjouw and Noda, 2022). These delays hinder our ability to understand welfare trends and develop targeted and effective public policies.
The World Bank has developed Survey of Well-Being with Instant and Frequent Tracking (SWIFT) In 2014, it aims to accelerate and streamline poverty and inequality monitoring. We apply machine learning techniques to the latest available household survey data to produce high-frequency poverty estimates that can be compared with official statistics. Rather than directly collecting comprehensive income and consumption data, SWIFT collects data on correlates of poverty, such as household size, asset ownership, education level, and food consumption, and uses estimation models to into poverty statistics. Data on these correlations are collected through short interviews that take approximately 2 to 5 minutes to conduct.
One of SWIFT's distinguishing features is the reliability of its poverty estimates after major climate change and economic shocks. Traditional machine learning models often break down in such situations, but continuous enhancements to the methodology since SWIFT's inception have enabled reliable poverty analysis even after the massive crisis in Afghanistan in 2016. were able to generate estimates (see Yoshida et al., 2022).
Beyond poverty estimates
Although originally designed to produce indicators of poverty and inequality, SWIFT's scope has expanded significantly. It now plays a key role in restoring the comparability of official poverty statistics, facilitating real-time monitoring in times of crisis, tracking welfare in vulnerable states, and identifying households in urgent need of social assistance. is playing. SWIFT has more than 200 implementations in 75 countries, spanning a diverse range of socio-economic contexts (see “Closing the Gap: Using SWIFT to Rapidly Monitor Poverty and Welfare in Times of Crisis”) reference).
In our new report, Enabling High-Frequency, Real-Time Poverty Monitoring in Developing Countries, SWIFT leverages machine learning and statistical techniques to provide real-time, reliable coverage of poverty dynamics across a variety of use cases. Learn more about how to generate insights.
Monitoring poverty after climate change
In southern Malawi, SWIFT has been integrated with the Rapid and Frequent Surveillance System, a region-based data collection system, to continuously generate poverty estimates for rural areas in the southern part of the country since 2020. In February 2022, he said, when two cyclones hit the region, and in March 2023, SWIFT reported that although the effects of the storms were not immediate, they caused damage to crops, which could lead to delays in the next harvest season. It has been shown that household income has been affected. As a result, this delayed effect manifested itself in the form of a gradual increase in poverty during the harvest season.
