Real-Time GDP Forecasting in Samoa: A New Model for Data-Driven Policymaking

Samoa has taken a major step in modernizing its economic policy tools through the development of a real-time forecasting model designed to estimate the country’s real GDP ahead of official data releases. This pioneering work is the result of a collaborative effort between the Central Bank of Samoa (CBS) and the International Monetary Fund’s Institute for Capacity Development (ICD), under a technical assistance program supported by the Government of Japan. The initiative directly addresses the persistent challenge faced by many low-income and small island developing states, the long delays in GDP data releases that hinder real-time decision-making. In Samoa, these delays extend up to 90 days after the close of a reference quarter, creating a significant information gap. By building a nowcasting model capable of predicting GDP growth using high-frequency indicators, the CBS is now able to provide timely and data-driven inputs into its macroeconomic assessments, offering a clearer picture of the economy’s current trajectory.

Samoa's Economic Foundations and Vulnerabilities

Samoa’s economic structure is distinctively shaped by its reliance on the services sector, which contributes nearly 80 percent of GDP on the production side. Agriculture and manufacturing each account for about 10 percent. On the expenditure side, household consumption dominates, making up around 80 percent of GDP, with government spending and capital formation also playing significant roles. However, the country’s economic resilience is persistently tested by its exposure to natural and climate-related hazards, including cyclones, droughts, and earthquakes, as well as public health emergencies. The COVID-19 pandemic, following a 2019 measles outbreak, led to three consecutive years of negative growth from 2020 to 2022, with tourism and services sectors particularly hard-hit. A turning point came in August 2022 with the reopening of international borders, which ushered in a phase of recovery supported by increased remittances, stronger tourism activity, and public investment.

Leveraging External Flows to Gauge Economic Activity

External inflows play an outsized role in Samoa’s economy, with remittances contributing about 45 percent of household consumption and tourism accounting for roughly 30 percent of GDP. The main sources of both are New Zealand and Australia, which together accounted for nearly 75 percent of tourism revenue and remittances in 2023. Samoa is also heavily dependent on imports, particularly from Singapore (for mineral fuels) and New Zealand (for food products), while its export destinations are concentrated in American Samoa and New Zealand. These patterns highlight how external economic developments, and changes in visitor arrivals or remittance flows have immediate effects on domestic economic conditions. These indicators thus form a crucial part of the high-frequency data used in the nowcasting model to project near-term GDP movements.

Building the Nowcasting Framework: Indicators and Models

To construct the nowcasting tool, the CBS and IMF teams began by compiling a comprehensive set of high-frequency indicators with potential links to GDP. These included remittances, foreign reserves, commercial bank credit, tourist receipts, market survey data from Samoa’s main produce market, merchandise trade figures, tax revenues, and climate variables such as the El Niño-linked Southern Oscillation Index. After rigorous statistical testing, checking for data quality, availability, seasonality, and correlation with GDP, the team refined the dataset to include eight core variables plus dummies for seasonality and COVID-19.

Three forecasting methodologies were applied: the Bridge model, which uses linear regression to link high- and low-frequency data; the MIDAS (Mixed Data Sampling) model, which handles mixed-frequency data through polynomial weighting; and the U-MIDAS (Unrestricted MIDAS) model, which relaxes structure assumptions and allows for more flexible lag patterns. Each model was evaluated using standard forecast accuracy measures like Root Mean Squared Error (RMSE) and Theil’s U-statistic. While all three performed well, especially compared to naïve forecasts, the U-MIDAS model stood out for its performance during the COVID-19 shock. However, the most consistent and accurate results were obtained by averaging forecasts from all three models, a technique known as forecast combination.

Making Nowcasting a Part of Policy Decision-Making

The CBS has now institutionalized the nowcasting tool as part of its regular policy analysis process. The output is distilled into one-page summaries that provide an updated economic narrative, allowing policymakers to respond more effectively to fast-changing conditions. This approach not only improves the timeliness of macroeconomic monitoring but also strengthens the country’s capacity for proactive fiscal and monetary policy. While the current model already delivers valuable insights, the CBS and IMF teams emphasize that nowcasting is not a one-time exercise. The models must be continuously refined as new data become available, and expert judgment remains essential for interpreting the forecasts within the broader economic context.

Charting the Path Ahead for Forecasting in Small Economies

Looking forward, the potential for expanding Samoa’s nowcasting capacity is substantial. Enhancements could include incorporating additional real-time indicators such as business sentiment surveys, retail sales data, and up-to-date government spending figures. There is also scope for adopting more advanced econometric techniques like dynamic factor models and mixed-frequency vector autoregressions, though these are more computationally intensive. Still, Samoa’s experience shows that even with relatively simple tools and conventional data, significant gains can be made in economic forecasting and policy preparedness. CBS’s success in embedding this model into its analytical toolkit serves as a valuable case study for other Pacific Island countries and low-income economies striving to improve economic management in the face of limited data infrastructure. Through the strategic use of available high-frequency indicators, nowcasting offers a practical and scalable solution for countries looking to make timely and informed policy decisions in today’s fast-moving economic landscape.