Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Opening the Black Box: Nowcasting Singapore's GDP Growth and its Explainability

Luca Attolico

TL;DR

The study develops a real-time nowcasting framework for Singapore GDP growth using a high-dimensional indicator panel and a diverse set of modeling approaches (penalized regressions, dimensionality reduction, ensemble methods, and neural networks). It couples expanding/rolling window validation with Bayesian hyperparameter optimization, and it quantifies predictive uncertainty via block bootstrap intervals while delivering model explainability through model-specific and integrated explainable AI tools. A Model Confidence Set (MCS) identifies superior models, which are then combined through Simple, Weighted, and Exponentially Weighted Averages, yielding time-varying interpretation of model contributions. Empirically, penalized regressions, PCR/PLSR, and GRU networks outperform benchmarks with RMSFE reductions around 40–60%, and ensemble aggregation (especially EWA) provides additional gains, with feature-attribution pointing to industrial production, external trade, and labor-market indicators as key drivers of Singapore’s short-run growth dynamics. The framework offers policy-relevant, transparent nowcasts with quantified uncertainty, demonstrating substantial improvements for decision-making in volatile, globally linked economies and enabling robust, period-aware interpretation of macroeconomic shocks.

Abstract

Timely assessment of current conditions is essential especially for small, open economies such as Singapore, where external shocks transmit rapidly to domestic activity. We develop a real-time nowcasting framework for quarterly GDP growth using a high-dimensional panel of approximately 70 indicators, encompassing economic and financial indicators over 1990Q1-2023Q2. The analysis covers penalized regressions, dimensionality-reduction methods, ensemble learning algorithms, and neural architectures, benchmarked against a Random Walk, an AR(3), and a Dynamic Factor Model. The pipeline preserves temporal ordering through an expanding-window walk-forward design with Bayesian hyperparameter optimization, and uses moving block-bootstrap procedures both to construct prediction intervals and to obtain confidence bands for feature-importance measures. It adopts model-specific and XAI-based explainability tools. A Model Confidence Set procedure identifies statistically superior learners, which are then combined through simple, weighted, and exponentially weighted schemes; the resulting time-varying weights provide an interpretable representation of model contributions. Predictive ability is assessed via Giacomini-White tests. Empirical results show that penalized regressions, dimensionality-reduction models, and GRU networks consistently outperform all benchmarks, with RMSFE reductions of roughly 40-60%; aggregation delivers further gains. Feature-attribution methods highlight industrial production, external trade, and labor-market indicators as dominant drivers of Singapore's short-run growth dynamics.

Opening the Black Box: Nowcasting Singapore's GDP Growth and its Explainability

TL;DR

The study develops a real-time nowcasting framework for Singapore GDP growth using a high-dimensional indicator panel and a diverse set of modeling approaches (penalized regressions, dimensionality reduction, ensemble methods, and neural networks). It couples expanding/rolling window validation with Bayesian hyperparameter optimization, and it quantifies predictive uncertainty via block bootstrap intervals while delivering model explainability through model-specific and integrated explainable AI tools. A Model Confidence Set (MCS) identifies superior models, which are then combined through Simple, Weighted, and Exponentially Weighted Averages, yielding time-varying interpretation of model contributions. Empirically, penalized regressions, PCR/PLSR, and GRU networks outperform benchmarks with RMSFE reductions around 40–60%, and ensemble aggregation (especially EWA) provides additional gains, with feature-attribution pointing to industrial production, external trade, and labor-market indicators as key drivers of Singapore’s short-run growth dynamics. The framework offers policy-relevant, transparent nowcasts with quantified uncertainty, demonstrating substantial improvements for decision-making in volatile, globally linked economies and enabling robust, period-aware interpretation of macroeconomic shocks.

Abstract

Timely assessment of current conditions is essential especially for small, open economies such as Singapore, where external shocks transmit rapidly to domestic activity. We develop a real-time nowcasting framework for quarterly GDP growth using a high-dimensional panel of approximately 70 indicators, encompassing economic and financial indicators over 1990Q1-2023Q2. The analysis covers penalized regressions, dimensionality-reduction methods, ensemble learning algorithms, and neural architectures, benchmarked against a Random Walk, an AR(3), and a Dynamic Factor Model. The pipeline preserves temporal ordering through an expanding-window walk-forward design with Bayesian hyperparameter optimization, and uses moving block-bootstrap procedures both to construct prediction intervals and to obtain confidence bands for feature-importance measures. It adopts model-specific and XAI-based explainability tools. A Model Confidence Set procedure identifies statistically superior learners, which are then combined through simple, weighted, and exponentially weighted schemes; the resulting time-varying weights provide an interpretable representation of model contributions. Predictive ability is assessed via Giacomini-White tests. Empirical results show that penalized regressions, dimensionality-reduction models, and GRU networks consistently outperform all benchmarks, with RMSFE reductions of roughly 40-60%; aggregation delivers further gains. Feature-attribution methods highlight industrial production, external trade, and labor-market indicators as dominant drivers of Singapore's short-run growth dynamics.

Paper Structure

This paper contains 84 sections, 60 equations, 13 figures, 10 tables.

Table of Contents

  1. Introduction and Literature Review
  2. Bridging the Information Gap in Macroeconomics: The Importance of Nowcasting
  3. Machine Learning as a Catalyst for Nowcasting
  4. Why Singapore? An Ideal Testbed
  5. Existing Approaches and Related Literature
  6. Limitations of Traditional and Emerging Approaches
  7. Objectives and Contributions of the Research
  8. Methodology and Pipeline
  9. Tools and Pipeline Organization
  10. Tools
  11. Research workflow
  12. Data and Preprocessing
  13. Data Sources: Collection, Integration, and Indexing
  14. Transformations and Feature Engineering
  15. Target Variable
  16. ...and 69 more sections

Figures (13)

  • Figure 1: Research pipeline workflow
  • Figure 2: Walk-forward nowcast with expanding+rolling window procedure
  • Figure 6: Actual and predicted deflated GDP growth rate for Penalized Linear models (LASSO, Ridge, and Elastic Net). Shaded areas represent the 95% prediction intervals obtained through the block bootstrap method.
  • Figure 7: Actual and predicted deflated GDP growth rate for Dimensionality Reduction-based models (Principal Component Regression and Partial Least Squares Regression). Shaded areas represent the 95% prediction intervals obtained through the block bootstrap method.
  • Figure 8: Actual and predicted deflated GDP growth rate for Ensemble Learning models (Random Forest and eXtreme Gradient Boosting). Shaded areas represent the 95% prediction intervals obtained through the block bootstrap method.
  • ...and 8 more figures