## Forecasting the Present in Developing Economies

## Source details

**Canonical URL:** [Forecasting the Present in Developing Economies](https://www.imf.org/-/media/files/publications/fandd/article/2025/12/kearns.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/fandd/article/2025/12/kearns.pdf.md)
- [Structured JSON version](/-/media/files/publications/fandd/article/2025/12/kearns.pdf.json)

---

### Overview
- Nowcasting uses real-time and high-frequency data to produce timely estimates of current economic activity when official indicators are delayed.
- Originated from a 2008 economics paper by Domenico Giannone, Lucrezia Reichlin, and David Small; developed from work begun in 2002 for a Ben Bernanke request.
- The pandemic highlighted the limits of traditional indicators and accelerated adoption of alternative data and nowcasting in developing economies.

### Data gaps and challenges
- Many developing economies publish official indicators infrequently or with long delays; about a third of countries in the world have only annual GDP.
- First estimates of GDP in many advanced economies come out about a month after the end of the quarter—two months in some major emerging market economies—and in developing economies it can take more than three months.
- Low-income countries often lack staff, funding, and analytical capacity, leaving large populations effectively “flying blind.”
- New data sources can be noisy or nonrepresentative and raise concerns about privacy, transparency, legitimacy, and governance.
- Capacity constraints: central banks and governments may lack budgets and skilled staff (economists, statisticians, data scientists) and advanced computing tools.

### Nowcasting methods and recent advances
- Nowcasting combines traditional high-frequency indicators (trade, money supply, tourism, electricity, wholesale and retail sales) with nontraditional inputs (satellite night-light intensity, Google search trends, electronic billing machine data).
- Models exploit comovement across countries and use global and regional components to interpolate activity for countries with less information.
- Large language models and artificial intelligence enable exploiting text as data and integrating data with metadata.
- Practitioners emphasize starting with high-quality available data or proxies before applying sophisticated techniques: “Start with the data, then you get to the technique.”
- Simple models often perform best for timely signals; nowcasting is more about combining series at different frequencies than about model complexity.

### Practical applications and examples
- Rwanda
  - Central bank nowcasting was used during the pandemic to produce a weekly economic activity index combining exports, imports, and consumer spending from the tax authority’s electronic billing machines.
  - Nowcasting is included in staff briefings before quarterly Monetary Policy Committee meetings; the economics staff is expanding.
- Kenya
  - Official GDP is released about three months after a quarter ends, but the central bank uses nowcasting tools to start gauging the quarter after only a week using private consumer spending data; remittance data arrive after two weeks; trade, money supply, tourism, and electricity data become available in about 40 days.
  - Nowcasting provides a good indication of quarter health in half the usual time.
- South Africa
  - Commercial agricultural data for meat and other crops provided an early read during a foot-and-mouth disease outbreak and helped explain the fastest year-over-year CPI increase in 10 months.
  - Commercial indicators lead CPI by one to three months across components, allowing weekly checkpoints for specific food categories.
- Democratic Republic of the Congo
  - Nowcasting combined copper and cobalt production and prices, money supply, satellite night-light intensity, and Google search trends to feed a quarterly projection model linking short-term assessments with medium-term forecasts and policy scenarios.
  - Mining sector contributes about a third of GDP.
- Guinea-Bissau
  - Deployed a blockchain platform (May 2024) for public wage bill management; results: the wage bill fell to half of tax revenue in 2024.
  - Platform will be expanded to cover all 26,600 public officials and 8,100 pensioners nationwide.
  - Prior condition: public wage bill consumed about four-fifths of tax revenues.
- Madagascar
  - Introduced agentic AI (October) in customs to detect inconsistencies and fraud by cross-analyzing customs declarations, invoices, manifests, and databases; automated many frontline inspection tasks.
- Samoa
  - With fewer than 90 staff members at the central bank and a population of just 220,000 people, Samoa is two years into a formal nowcasting system and plans to expand the economics department from 8 people to 10.

### Benefits observed
- Timely detection of turning points in growth and inflation, enabling faster policy adjustments.
- Improved forecasting, analysis, and communication with the public through forecasting and policy analysis systems.
- Enhanced fiscal integrity and near real-time tracking of wage bills via blockchain, reducing audit burdens and improving data for policy.
- Nowcasting can approximate economic growth ahead of official GDP releases and can be applied where quarterly GDP estimates are unavailable.

### Policy implications and recommendations
- Expand official indicators’ frequency and granularity where possible; nowcasting complements but does not replace rigorous data collection.
- Prioritize capacity development: train and expand teams of economists, statisticians, and data scientists and provide adequate funding and computing infrastructure.
- Start with available high-quality data or appropriate proxies before applying complex models.
- Use a mix of traditional and nontraditional high-frequency indicators tailored to country data environments.
- Address data governance issues: privacy, transparency, legitimacy, and oversight when deploying new data sources and AI tools.
- Leverage regional and global comovement to interpolate for low-information countries, while validating models against available domestic signals.
- Consider targeted technology deployments (blockchain for fiscal transparency, AI for customs fraud detection) as part of broader reform programs.

### Key statistics and specific dates (as cited)
- About a third of countries publish only annual GDP.
- First GDP estimates: about a month after quarter end in many advanced economies; two months in some major emerging markets; more than three months in developing economies.
- Kenya: central bank starts gauging the quarter after only a week using private consumer spending data; remittance data after two weeks; trade, money supply, tourism, and electricity data in about 40 days.
- Commercial CPI indicators lead official CPI by one to three months.
- Democratic Republic of the Congo: mining contributes about a third of GDP.
- Guinea-Bissau: public wage bill consumed about four-fifths of tax revenues prior to reform; wage bill fell to half of tax revenue in 2024; platform will cover 26,600 public officials and 8,100 pensioners.
- Guinea-Bissau blockchain deployment: May 2024.
- Madagascar customs AI introduced: October (year in source context).
- Samoa population: just 220,000 people; central bank staff fewer than 90; economics department expansion from 8 people to 10.
- Domenico Giannone, Lucrezia Reichlin, and David Small paper: 2008; initial model development began in 2002.

*SOURCE: IMF staff.*

---


_Source: https://www.imf.org/-/media/files/publications/fandd/article/2025/12/kearns.pdf_