Banco de España's Spain-STING model nowcasts quarterly Spanish GDP growth in real time from early indicators; a 2024 respecification (contemporaneous links, time-varying volatility, revised indicators) produced a documented accuracy gain in the volatile post-pandemic period.
Details
Maturity
Established
Promoter
Banco de España (Directorate General Economics, Statistics and Research)
Period
Model family introduced 2008–2009 (Euro-STING/Spain-STING); respecified and re-evaluated 2024 (Occasional Papers 24/06, 4 March 2024); journal version published 2025
Region (NUTS)
ES30
Keywords
central banking, macroeconomic forecasting, monetary policy, statistics
Context
Banco de España has operated the Spain-STING (Short-Term Indicator of Growth) nowcasting tool since the model family was introduced by economists Máximo Camacho and Gabriel Pérez-Quirós. It estimates a common latent factor from indicators available before official GDP data are released, giving policymakers a real-time read on the Spanish economy weeks ahead of Eurostat's confirmed figures.
Objectives
The 2024 work aimed to fix a known weakness: the model's accuracy had degraded during the high-volatility pandemic period, so the authors sought to respecify it while preserving its established pre-pandemic performance.
Activities
In Occasional Paper 24/06 (published 4 March 2024, later in Revista Economía v.48), Banco de España economists Ana Gómez Loscos, Miguel Ángel González-Simón and Matías Pacce respecified the model with three changes: contemporaneous rather than purely leading indicator relationships, a stochastic (time-varying) process for the common component's volatility, and a revised, post-pandemic-relevant set of input indicators.
Results
The paper reports a notable improvement in nowcasting performance during the high-volatility post-pandemic period while preserving the model's pre-pandemic accuracy, though it states this as a qualitative finding rather than a specific quantified error-reduction figure. The model sits within a wider, independently benchmarked family — Eurostat's 2025 statistical working paper on real-time GDP nowcasting for the euro area evaluates comparable tools across six member states including Spain, and sister models (Euro-STING, Ñ-STING) share the same architecture.
Conclusions
Spain-STING's forecasts and methodology are published routinely by the central bank rather than held as an internal black box, though it remains a statistical/dynamic-factor approach rather than a modern machine-learning system. Banco de España's own paper is candid that accuracy degraded during the pandemic before the 2024 fix, underscoring that even long-standing operational tools require periodic respecification.
Implementation
Indicative cost
Low (< €50k)
Time to results
Short (< 1 year)
Staffing & skills
Named Banco de España research economists (Ana Gómez Loscos, Miguel Ángel González-Simón, Matías Pacce) within the Directorate General Economics, Statistics and Research
Conditions for success
Continuous institutional maintenance and periodic respecification of the model over more than a decade
Publication of methodology and results in the Occasional Papers series and peer-reviewed journals
Existence of a wider benchmarked model family (Euro-STING, Ñ-STING) enabling cross-checking
Commonly funded by
National / regional programmes
Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.
Do you run this practice?
Claim it —
verified implementers get a public contact pathway and can propose corrections.
Data sources
Where this practice's information was retrieved from, and when.
In an April 2025 Staff Memo, Riksbank economists found random-forest and neural-network models measurably outperformed Sweden's traditional time-series forecasts for …
SARB researchers rigorously tested machine-learning models (LASSO, XGBoost, neural networks) against the bank's own published inflation forecasts: ML modestly beat …