Nowcasting Real GDP in Samoa

IMF Working Papers, May 16, 2025

Source details

Authors: Geoffrey M Heenan, Karras Lui, Ian J Nield, Eucharist Muaulu, Viiaonaoperesi Reupena, Ivy Sabuga, Aiulu Tolovaa
Published: May 16, 2025
Series: IMF Working Papers
DOI: https://doi.org/10.5089/9798229009782.001

Purpose: Strengthen nowcasting capacity at the Central Bank of Samoa (CBS) to estimate the present and near-term forecast of real GDP in Samoa.
Motivation: Real GDP data release experiences up to a 90-day lag.
Authors: Geoffrey M Heenan, Karras Lui, Ian J Nield, Eucharist Muaulu, Viiaonaoperesi Reupena, Ivy Sabuga, Aiulu Tolovaa
Date: May 16, 2025

High-frequency datasets compiled: tourism receipts, agriculture market survey, remittances, among others.
Nowcasting methods employed:
- Bridge
- Mixed Data Sampling (MIDAS)
- Unrestricted MIDAS (U-MIDAS)
Evaluation metric: root mean squared error (RMSE)
Samples analyzed: full sample and pre-COVID-19 sample
Crisis performance assessment: analysis includes the COVID-19 pandemic period to evaluate turning-point detection.

All three nowcasting methods (Bridge, MIDAS, U-MIDAS) significantly outperform naive forecasts.
Forecast combination of the three methods minimizes the RMSE for both full and pre-COVID-19 samples.
U-MIDAS performs better during crises, particularly in identifying turning points during the COVID-19 pandemic.
Strengthening nowcasting capacity is important given the up to 90-day lag in official real GDP releases.

Continue developing and operationalizing nowcasting tools at CBS to reduce informational lags for policy decisions.
Use forecast combinations (Bridge, MIDAS, U-MIDAS) to minimize RMSE in routine nowcasting.
Deploy U-MIDAS or place greater weight on it during crisis periods to improve early detection of turning points.
Maintain and expand high-frequency data collection (tourism receipts, agriculture market surveys, remittances, etc.) to feed nowcasting models.