MACROCAST: A Vintage-Consistent Time Series Foundation Model for Real-Time Macroeconomic Forecasting

2026-06-26

AI Summary

This study addresses the pervasive issues of temporal contamination and revision bias in real-time macroeconomic forecasting by proposing a lightweight time series foundation model that, for the first time, completely avoids both forms of data leakage. The approach leverages only synthetic data—generated via Bayesian VAR, dynamic factor models, and ARIMA—for pretraining and fine-tunes exclusively on historical vintage data from ALFRED, thereby ensuring strict alignment between training conditions and actual forecasting environments. Evaluated on out-of-sample real-time data from FRED-MD, the model outperforms the AR(1) benchmark in approximately 80% of series–horizon combinations, achieving performance comparable to or better than the current state-of-the-art TSFM Chronos-2 and substantially surpassing traditional econometric models.

Abstract

We introduce MACROCAST, a lightweight Time Series Foundation Model (TSFM) for real-time macroeconomic forecasting. Existing TSFMs suffer from data leakage in two forms: temporal contamination, as the model may have seen the realized values of the series it forecasts, and revision bias, as training on fully revised data diverges from the preliminary, vintage-specific releases available to real-time forecasters. MACROCAST is, to our knowledge, the first TSFM that rules out both forms of leakage entirely: at no stage of training is the model exposed to information that would not have been available to a forecaster in real time. We train MACROCAST first on purely synthetic time series in approximately one GPU-day and then fine-tune it on synthetic time series drawn from Bayesian VARs, dynamic factor models, and ARIMA specifications estimated on vintage-specific ALFRED data. Because pretraining uses only simulated data and fine-tuning uses only real-time vintages, no observed future or revised value ever enters the model; each fine-tuning run takes nine minutes. Evaluated on the FRED-MD database in a genuine real-time out-of-sample exercise, MACROCAST improves on the AR(1) benchmark for roughly 80% of series-horizon pairs, matches or surpasses Chronos-2 -- the strongest currently available TSFM -- and outperforms the Bayesian VAR and dynamic factor model benchmarks, all in a data-leakage-free manner.

Problem

• data leakage
• temporal contamination
• revision bias
• real-time forecasting
• macroeconomic forecasting

Innovation

• time series foundation model
• real-time forecasting
• vintage-consistent
• data leakage prevention
• synthetic pretraining