Working papers

We extend the vector autoregression (VAR) to model binary, censored, count and unrestricted variables jointly, treating missing observations of any type in the same way. Each restricted observation is linked to a latent Gaussian variable, so that the standard machinery for VARs, such as shrinkage priors, stochastic volatility and outlier components, applies directly to the VAR for the latent variables. The latent variables have a truncated normal conditional distribution with a banded precision matrix, from which they are drawn, jointly or equation by equation, using Hamiltonian Monte Carlo, and the sampler scales well to high dimensions. A simulation study shows that treating restricted variables as continuous distorts the levels and scales of the system and worsens forecasts of the restricted outcomes in most designs. In an application to 25 US macro-financial variables, we estimate a business cycle indicator, the intensity of banking distress and shadow rates. We use the model for scenario analysis with restrictions on the recession probability, bank failures and interest rates at the effective lower bound.

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.

We study double descent and benign overfitting in macroeconomic forecasting. We document that double-descent risk curves arise in standard macroeconomic datasets that are driven by a small number of latent factors, and we characterize when the underlying benign-overfitting mechanism holds. The conditions of Bartlett et al. (2020) are satisfied under the exact factor model and can also hold under the more realistic approximate factor model, provided idiosyncratic variances are not too dispersed across series. Because macroeconomic panels have only moderate dimensions, the overparameterization ratio N/T required by the theory is not naturally available. Our solution is to augment the data with synthetic copies from an estimated factor model and we prove that this strategy converges to a kernel ridge regression with a factor-structured kernel. Using monthly (FRED-MD) and quarterly (FRED-QD) US data, the resulting estimator consistently outperforms the Stock-Watson factor model for point forecasting across all series and horizons, with gains that are pervasive, statistically significant, and increasing with the forecast horizon. Our results suggest that benign overfitting, when it works, succeeds because overparameterization implicitly constructs a well-behaved kernel, not because overparameterization is intrinsically desirable.

  • Pettenuzzo, D., Poon, A., Zhu, D., Revealing Growth-at-Risk Through Daily Corporate Activity: New Evidence from a Functional Quantile MIDAS Model (January 2026)
    [Link to SSRN]

This paper introduces a new Growth-at-Risk (GaR) framework that incorporates daily firm-level corporate activity data, an information source previously unused in GaR analysis. Motivated by macro-finance theory emphasizing the role of firm balance sheets in amplifying shocks, we exploit a novel dataset of daily accounting-based indicators capturing firms’ liquidity, leverage, and operating performance. We link these high-frequency measures to quarterly GDP by developing a quantile MIDAS model with a continuous Fourier weighting scheme. We apply this framework to U.S. data and show that both the levels and cross-sectional dispersion of corporate activity provide significant early-warning signals of macroeconomic stress. Our model delivers substantial improvements in predictive accuracy relative to the benchmark GaR approach and MIDAS-based extensions using monthly financial indicators. The results highlight the value of high-frequency corporate information for real-time monitoring of downside macroeconomic risks.

This paper presents a comparative analysis evaluating the accuracy of Large Language Models (LLMs) against traditional macro time series forecasting approaches. In recent times, LLMs have surged in popularity for forecasting due to their ability to capture intricate patterns in data and quickly adapt across very different domains. However, their effectiveness in forecasting macroeconomic time series data compared to conventional methods remains an area of interest. To address this, we conduct a rigorous evaluation of LLMs against traditional macro forecasting methods, using as common ground the FRED-MD database. Our findings provide valuable insights into the strengths and limitations of LLMs in forecasting macroeconomic time series, shedding light on their applicability in real-world scenarios.

What do companies’ 10-Q filings reveal about the state of the macro economy and do specific accounting variables contain particularly relevant information? To address these questions, we analyze the lead-lag patterns of more than twenty accounting variables in relation to aggregate economic activity. We develop new daily corporate account business activity indices that aggregate firm-level accounting information while controlling for shifts in the composition of announcers and reducing firm-specific noise. Our new indices show that firm liquidity becomes significantly lower while corporate debt grows significantly faster several months prior to recessions, and thus can be used as leading indicators. Conversely, operations, earnings, and profitability measures tend to be significantly lower after recessions, suggesting they are mostly lagging, pro-cyclical indicators of economic activity.

Published and forthcoming articles

  1. Chan, J., Pettenuzzo, D., Poon, A., Zhu, D. (2025), Conditional Forecasts in Large Bayesian VARs with Multiple Equality and Inequality Constraints , Journal of Economic Dynamics and Control, 173
    [Published version]     [Link to SSRN]

  2. Pettenuzzo, D., Timmermann, A., Sabbatucci, R. (2023), Payout suspensions during the Covid-19 pandemic, Economics Letters, 224:111024
    [Published version]    

  3. Pettenuzzo, D., Timmermann, A., Sabbatucci, R. (2023), Dividend Suspensions and Cash Flows During the Covid-19 Pandemic: A Dynamic Econometric Model, Journal of Econometrics, 235:1522–1541
    [Published version]     [Working paper]

  4. Pettenuzzo, D., Timmermann, A., Yong, S. (2022), Corrigendum to “Predictability of stock returns and asset allocation under structural breaks” [J. Econometrics 164 (2011) 60–78], Journal of Econometrics, 227: 513-517
    [Published version]    

  5. Pettenuzzo, D., Timmermann, A., Sabbatucci, R. (2021), Outlasting the pandemic: Corporate payout and financing decisions during Covid-19, COVID Economics, 78
    [Published version]     [Working paper]    

  6. Korobilis, D., Pettenuzzo, D. (2020) Machine Learning Econometrics: Bayesian Algorithms and Methods, Oxford Research Encyclopedia: Economics and Finance
    [Published version]     [Working paper]    

  7. Pettenuzzo, D., Timmermann, A., Sabbatucci, R. (2020) Cash Flow News and Stock Price Dynamics, Journal of Finance, 75: 2221-2270
    [Published version]     [Working paper]     [Online Appendix]

  8. Carvalho, C., Fisher, J., Pettenuzzo, D. (2020) Optimal Asset Allocation with Multivariate Bayesian Dynamic Linear Models, Annals of Applied Statistics, 14: 299-338
    [Published version]     [Working paper]    

  9. Pan, Z., Pettenuzzo, D., Wang, Y. (2020) Forecasting Stock Returns: A Predictor-constrained Approach Journal of Empirical Finance, 55: 200-217
    [Published version]     [Working paper]    

  10. Korobilis, D., Pettenuzzo, D. (2019) Adaptive Hierarchical Priors for High-Dimensional Vector Autoregressions, Journal of Econometrics, 212: 241-271
    [Published version]     [Working paper]    

  11. Koop, G., Korobilis, D. , Pettenuzzo, D. (2019) Bayesian Compressed Vector Autoregressions, Journal of Econometrics, 210: 135-154
    [Published version]     [Working paper]     [Online Appendix]

  12. Gargano, A., Pettenuzzo, D., Timmermann, A. (2019) Bond Return Predictability: Economic Value and Links to the Macroeconomy, Management Science, 65: 508-540
    [Published version]     [Working paper]     [Online Appendix]

  13. Metaxoglou, K., Pettenuzzo, D., Smith, A. (2019) Option-Implied Equity Premium Predictions via Entropic Tilting, Journal of Financial Econometrics, 17: 559-586
    [Published version]     [Working paper]     [Online Appendix]

  14. Pettenuzzo, D., Timmermann, A. (2017) Forecasting Macroeconomic Variables Under Model Instability, Journal of Business and Economic Statistics, 35: 183-201
    [Published version]     [Working paper]     [Online Appendix]

  15. Pettenuzzo, D., Timmermann, A., Valkanov, R. (2016) A MIDAS Approach to Modeling First and Second Moment Dynamics, Journal of Econometrics, 193: 315-334
    [Published version]     [Working paper]

  16. Pettenuzzo, D., Ravazzolo, F. (2016) Optimal Potfolio Choice under Decision-Based Model Combinations, Journal of Applied Econometrics, 31: 1312-1332
    [Published version]     [Working paper]     [Online Appendix]

  17. Pettenuzzo, D., Timmermann, A., Valkanov, R. (2014) Forecasting Stock Returns under Economic Constraints, Journal of Financial Economics, 114: 517-553
    [Published version]     [Working paper]

  18. Pettenuzzo, D., White, H. (2014) Granger Causality, Exogeneity, Cointegration, and Economic Policy Analysis, Journal of Econometrics, 178: 316-330
    [Published version]     [Working paper]

  19. Pettenuzzo, D., Timmermann, A. (2011) Predictability of Stock Returns and Asset Allocation under Structural Breaks, Journal of Econometrics, 164: 60-78
    [Published version]     [Working paper]     [Matlab codes]

  20. Pesaran, H., Pettenuzzo, D., Timmermann, A. (2007) Learning, Structural Instability, and Present Value Calculations, Econometric Reviews, 26: 253-288
    [Published version]     [Working paper]

  21. Pesaran, H., Pettenuzzo, D., Timmermann, A. (2006) Forecasting Time Series subject to Structural Breaks, Review of Economic Studies, 73: 1057-1084
    [Published version]     [Working paper]     [Matlab codes]


Work in progress and older working papers