Andrea Carriero, Davide Pettenuzzo, Shubhranshu Shekhar
arXiv 27 Jun 2026 · Econometrics
arXiv:2606.28670 · PDF · DOI · OpenAlex · Extracted main text
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.
appendix boundary found by appendix_command · 86% of the source is main text. Read the extracted text to check this.
The works this paper leans on most, across its whole bibliography — not restricted to papers in our corpus. Ranked by composite intensity, which combines how often a work is mentioned, how many sections mention it, and how much of that falls in the main text rather than the appendix.
| Reference | Intensity | Mentions | Sections | Main text | |
|---|---|---|---|---|---|
| 1 | Vladyslav Moroshan and Julien Siems and Arber Zela and Timur Carsten… (2026) TempoPFN: Synthetic Pre-training of Linear RNNs for Zero-shot Time Series Forecasting | 1.000 | 7 | 3 | 100% |
| 2 | Ansari, Abdul Fatir and Stella, Lorenzo and Turkmen, Caner and Zhang… (2024) Chronos: Learning the language of time series | 1.000 | 6 | 3 | 100% |
| 3 | Ansari, Abdul Fatir and Shchur, Oleksandr and Küken, Jaris and Auer,… (2025) Chronos-2: From univariate to universal forecasting | 1.000 | 6 | 3 | 100% |
| 4 | Banbura, Marta and Giannone, Domenico and Reichlin, Lucrezia (2010) Large Bayesian Vector Autoregressions | 0.843 | 4 | 4 | 75% |
| 5 | Michael W. McCracken and Serena Ng (2016) FRED-MD: A Monthly Database for Macroeconomic Research | 0.794 | 10 | 3 | 50% |
| 6 | Stock, James H. and Watson, Mark W (2002) Forecasting Using Principal Components From a Large Number of Predictors | 0.737 | 4 | 3 | 50% |
| 7 | Das, Abhimanyu and Kong, Weihao and Sen, Rajat and Zhou, Yichen (2024) A decoder-only foundation model for time-series forecasting | 0.737 | 3 | 2 | 100% |
| 8 | Carriero, Andrea and Pettenuzzo, Davide and Shekhar, Shubhranshu (2025) Macroeconomic Forecasting with Large Language Models self | 0.644 | 4 | 1 | 100% |
| 9 | Darlow, Luke Nicholas and Joosen, Artjom and Asenov, Martin and Deng… (2023) TSMix: time series data augmentation by mixing sources | 0.644 | 2 | 2 | 100% |
| 10 | Liu, Chenghao and Aksu, Taha and Liu, Juncheng and Liu, Xu and Yan,… (2025) Moirai 2.0: When less is more for time series forecasting | 0.644 | 2 | 2 | 100% |
Showing the top 10 of 58 scored citations.