世界银行-利用SWIFT实现发展中国家的高频实时贫困监测：通过即时和频繁跟踪进行的福祉调查（英）-2024.4-55页_11mb

SWIFT Framework for Poverty Monitoring Summary

Introduction

The Survey of Wellbeing via Instant and Frequent Tracking (SWIFT) is an innovative methodology for producing timely and cost-effective poverty statistics. Official poverty data in low-income countries is often infrequent, available only every seven years on average, missing yearly fluctuations. Traditional methods are slow and expensive, requiring detailed interviews and months for analysis. SWIFT addresses this gap by using machine learning models trained on household surveys, enabling estimates from 10-15 poverty correlates with minimal data collection.

Framework Components

1. Model Development: Uses stepwise regression on a harmonized training dataset to predict household expenditure.
2. Data Harmonization: Ensures target datasets align with training data through questionnaires or variable adjustments.
3. Imputation: Applies trained models to target datasets to estimate poverty rates, with multiple imputations for robustness.
4. SWIFT Plus & SWIFT 2.0: Addresses model instability from economic shocks by incorporating fast-changing variables or conducting mini-Household Budget Surveys (mini-HBS).

Key Advantages

• Efficiency: Reduces data collection time to weeks and costs to under $25,000 per round.
• Flexibility: Works with various data sources, including traditional surveys, phone surveys, and community-based data collection.
• Accuracy: Imputation errors typically within ±2 percentage points of actual poverty rates.
• Applications: Increases data frequency, produces estimates without reliable training data, monitors rapid changes, and restores comparability for time-series analysis.

Caveats

• Data Quality: Estimates rely on high-quality training and target datasets to avoid reporting bias.
• Model Validation: Requires cross-validation and out-of-sample testing to ensure reliability and mitigate overfitting.
• Cost Variations: Dependent on data availability, with higher costs for mini-HBS (e.g., up to $200,000).
• Socio-Economic Context: Local experts are needed for validation to ensure contextual accuracy.

Country Applications

• Frequency Increases: Paraguay, Botswana, DRC use SWIFT with existing surveys to generate quarterly or annual poverty data.
• Disaster Response: Zimbabwe and Uganda implement SWIFT 2.0 during crises for timely estimates.
• Alternative Data Collection: Malawi and Uganda leverage community-based and phone surveys for real-time monitoring.
• Comparability Restoration: Mongolia and Zambia harmonize datasets to track poverty trends over time.
• Historical Trends: Nigeria uses SWIFT to fill data gaps and analyze past poverty changes.

Conclusion

SWIFT provides a flexible, cost-efficient framework for enhancing poverty monitoring, enabling frequent and reliable data in developing countries. Its integration with digital and innovative data sources supports evidence-based policymaking, crisis response, and poverty reduction efforts, though continuous validation and investment in data infrastructure are essential for accuracy.

# SWIFT Framework Summary

- **Definition**: SWIFT (Survey of Wellbeing via Instant and Frequent Tracking) is a machine-learning-based method for rapid, frequent poverty estimation.
- **Core Benefits**: Reduces time and cost compared to traditional surveys; enables timely insights during crises and policy changes.
- **Implementation**: Requires careful data harmonization and model validation for reliable estimates.
- **Global Adoption**: Used in over 70 countries, supporting various data scenarios from frequent surveys to community-based systems.