pith. sign in

arxiv: 2604.03681 · v1 · submitted 2026-04-04 · 💰 econ.EM

A Dynamic Factor Model for Level and Volatility

Haroon Mumtaz , Sofia Velasco This is my paper

Pith reviewed 2026-05-13 17:21 UTC · model grok-4.3

classification 💰 econ.EM
keywords dynamic factor modelvolatility factorsdensity forecastingtail risksinflation uncertaintystate-dependent riskslarge panel data
0
0 comments X

The pith

A dynamic factor model with jointly evolving level and volatility factors generates endogenous state-dependent and asymmetric tail risks.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper develops a dynamic factor model in which common level and volatility factors evolve jointly within a large panel. The joint dynamics allow conditional means and variances to interact directly so that volatility shifts change both the spread and the central location of predicted outcomes. This structure produces state-dependent and asymmetric tail risks endogenously from the factor system rather than imposing them through reduced-form additions. Empirically the model improves density forecast accuracy with the largest gains appearing in the tails of the distributions and at medium horizons. In an application to international inflation a global level factor dominates in advanced economies while regional and volatility factors play larger roles in emerging economies.

Core claim

The model embeds the joint dynamics of level and volatility factors directly in the factor structure, allowing risk to arise endogenously from the interactions between means and variances across a large panel, which produces predictive distributions with state-dependent asymmetric tail risks and improves density forecast accuracy particularly in the tails at medium horizons.

What carries the argument

The joint evolution of common level and volatility factors that drive co-movement in first and second moments across many series, with heavy-tailed idiosyncratic shocks absorbing transitory outliers to isolate persistent uncertainty.

If this is right

  • Volatility fluctuations affect both dispersion and location of outcomes, generating state-dependent tail risks in the predictive distributions.
  • Density forecast accuracy improves systematically, with the strongest gains in the tails and at medium horizons.
  • International inflation data reveal a dominant global level component in advanced economies and stronger regional plus volatility contributions in emerging economies.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same joint-factor structure could be applied to other macro panels such as output growth to examine endogenous risk transmission.
  • The endogenous generation of asymmetric tails implies that uncertainty shocks may propagate differently across advanced and emerging economies.
  • Comparing the model's implied state-dependent tail risks against external benchmarks such as option-implied volatilities would provide an out-of-sample test of the mechanism.

Load-bearing premise

The joint evolution of level and volatility factors within the factor structure is sufficient to generate the observed state-dependent and asymmetric tail risks without requiring external imposition or leading to overfitting.

What would settle it

A direct comparison in which separate models for levels and volatilities produce equally accurate or superior tail density forecasts at medium horizons would falsify the claimed advantage of the joint evolution.

Figures

Figures reproduced from arXiv: 2604.03681 by Haroon Mumtaz, Sofia Velasco.

Figure 1
Figure 1. Figure 1: Common Volatility Notes: The figure reports estimates from the U.S. specification with two level factors and one volatility factor in the first row and two level and two volatility factors in the second row. The solid blue lines denotes the posterior median of the common volatility factor, and the shaded area denotes the 68% credible intervals. The red and gray lines report the macroeconomic and financial … view at source ↗
Figure 2
Figure 2. Figure 2: Output growth predictive densities and common volatility [PITH_FULL_IMAGE:figures/full_fig_p013_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: One-period-ahead predictive quantiles Notes: The figure reports one-quarter-ahead predictive 5% and 95% quantiles for selected macroeconomic variables. Green lines correspond to the level–volatility DFM, grey lines correspond to the benchmark DFM without a common volatility factor, and blue dashed lines correspond to the quantile regression (QR) benchmark. For GDP growth, the QR specification follows the G… view at source ↗
Figure 4
Figure 4. Figure 4: Estimated level and volatility inflation factors: global and regional drivers [PITH_FULL_IMAGE:figures/full_fig_p024_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Global and regional drivers of inflation: country-level FEVDs [PITH_FULL_IMAGE:figures/full_fig_p027_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Latent Factors: Estimated and True Notes: The figure compares posterior median estimates of the latent factors with their true simulated values. Shaded areas denote 90% credible bands. Factors are shown under the standard normalization used in estimation, which fixes scale and sign for comparability. The top panel reports the level factor (ft), while the bottom panel reports the volatility factor (Ft). 38 … view at source ↗
Figure 7
Figure 7. Figure 7: Estimated versus True: Loadings and Persistence Parameters [PITH_FULL_IMAGE:figures/full_fig_p039_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Estimated versus True: Degrees of Freedom [PITH_FULL_IMAGE:figures/full_fig_p040_8.png] view at source ↗
read the original abstract

This paper develops a dynamic factor model in which common level and volatility factors evolve jointly, allowing conditional means and variances to interact endogenously within a large-information setting. The joint evolution of these factors provides a tractable framework for modeling risk, as fluctuations in volatility affect both the dispersion and the location of outcomes, generating state-dependent and asymmetric tail risks in predictive distributions. Volatility is captured by latent common factors that drive co-movement in second moments across a large panel, while heavy-tailed idiosyncratic shocks absorb transitory outliers and isolate persistent uncertainty dynamics. The framework embeds these interactions directly within a factor structure, allowing risk to arise endogenously from the joint dynamics of the system rather than being imposed through reduced-form approaches. Empirically, the model delivers systematic improvements in density forecast accuracy, particularly in the tails of the predictive distribution and at medium horizons. An application to international inflation highlights a dominant global level component in advanced economies and stronger regional and volatility contributions in emerging and developing economies, pointing to substantial heterogeneity in the role of uncertainty across countries.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper develops a dynamic factor model in which common level and volatility factors evolve jointly, allowing conditional means and variances to interact endogenously within a large-information setting. Volatility is captured by latent common factors driving co-movement in second moments, while heavy-tailed idiosyncratic shocks absorb transitory outliers and isolate persistent uncertainty. The joint dynamics generate state-dependent and asymmetric tail risks without external imposition. Empirically, applied to international inflation data, the model claims systematic improvements in density forecast accuracy, particularly in the tails of the predictive distribution and at medium horizons, while highlighting a dominant global level component in advanced economies and stronger regional/volatility roles in emerging economies.

Significance. If the identification and forecast results hold, the framework offers a valuable advance in modeling endogenous risk and tail behavior in high-dimensional macro panels, moving beyond reduced-form approaches. The emphasis on density forecasts and heterogeneity across economies could inform better uncertainty quantification for policy. The endogenous interaction mechanism is a conceptual strength if cleanly separated from added flexibility.

major comments (3)
  1. [§3.2] §3.2, identification discussion: the separation of common volatility factors from heavy-tailed idiosyncratic components is load-bearing for attributing tail-forecast gains to endogenous joint dynamics rather than model flexibility; the current description does not provide explicit orthogonality conditions or loading restrictions to ensure separate identification.
  2. [Forecast evaluation section] Forecast evaluation section, Table 4 (medium-horizon rows): the reported density-forecast improvements (especially 5%/95% quantiles) must be tested against a nested benchmark that shuts off the level-volatility interaction while retaining the same number of factors and heavy tails; without this, the headline claim that gains arise from endogenous interactions is not isolated.
  3. [§5.3] §5.3, heterogeneity results: the conclusion that volatility contributions are stronger in emerging economies depends on the estimated factor shares; the manuscript should report robustness to alternative numbers of volatility factors (e.g., 1 vs. 2) to confirm the cross-country pattern is not an artifact of factor selection.
minor comments (2)
  1. [Abstract] Abstract: state the exact number of countries, sample period, and number of level/volatility factors used in the inflation application to give readers immediate context.
  2. [Notation] Notation: ensure subscripts distinguishing level factors (e.g., f_t) from volatility factors (e.g., h_t) are used consistently in all equations and text.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments. We address each major point below and will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [§3.2] §3.2, identification discussion: the separation of common volatility factors from heavy-tailed idiosyncratic components is load-bearing for attributing tail-forecast gains to endogenous joint dynamics rather than model flexibility; the current description does not provide explicit orthogonality conditions or loading restrictions to ensure separate identification.

    Authors: We agree that the identification discussion in §3.2 requires greater explicitness. In the revision we will add the precise orthogonality conditions between the common volatility factors and the idiosyncratic shocks, together with the loading restrictions that enforce their separation. revision: yes

  2. Referee: [Forecast evaluation section] Forecast evaluation section, Table 4 (medium-horizon rows): the reported density-forecast improvements (especially 5%/95% quantiles) must be tested against a nested benchmark that shuts off the level-volatility interaction while retaining the same number of factors and heavy tails; without this, the headline claim that gains arise from endogenous interactions is not isolated.

    Authors: We accept the need for a nested benchmark that disables the level-volatility interaction while preserving the factor count and heavy-tailed errors. The revised manuscript will report the corresponding density-forecast results in Table 4 for the medium-horizon rows, allowing direct comparison with the full model. revision: yes

  3. Referee: [§5.3] §5.3, heterogeneity results: the conclusion that volatility contributions are stronger in emerging economies depends on the estimated factor shares; the manuscript should report robustness to alternative numbers of volatility factors (e.g., 1 vs. 2) to confirm the cross-country pattern is not an artifact of factor selection.

    Authors: We will supplement §5.3 with robustness checks that re-estimate the model using one and two volatility factors and verify that the cross-country pattern—stronger volatility contributions in emerging economies—remains stable. revision: yes

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper introduces a new dynamic factor model structure in which common level and volatility factors evolve jointly, with heavy-tailed idiosyncratics isolating persistent uncertainty. Density forecast improvements are reported as outcomes of an empirical application to inflation data rather than any quantity that reduces by construction to fitted parameters or self-citations. No self-definitional steps, fitted-input predictions, load-bearing self-citations, uniqueness theorems imported from prior work, or ansatzes smuggled via citation appear in the model description or claims. The derivation remains self-contained.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

Abstract-only review limits visibility; the model relies on standard factor assumptions plus new latent volatility factors and heavy-tailed shocks.

free parameters (2)
  • Number of common level and volatility factors
    Not specified in abstract but required to fit the panel data.
  • Parameters governing joint factor dynamics
    Evolution parameters for level-volatility interactions fitted to data.
axioms (1)
  • domain assumption Common factors drive co-movement in both first and second moments across large panels
    Core premise of the factor structure invoked throughout the abstract.

pith-pipeline@v0.9.0 · 5470 in / 1057 out tokens · 46290 ms · 2026-05-13T17:21:08.853937+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

40 extracted references · 40 canonical work pages

  1. [1]

    Boyarchenko, and D

    Adrian, T., N. Boyarchenko, and D. Giannone (2019): Vulnerable growth, American Economic Review, 109, 1263--1289

    work page 2019

  2. [2]

    Aguilar, O. and M. West (2000): Bayesian dynamic factor models and portfolio allocation, Journal of Business & Economic Statistics, 18, 338--357

    work page 2000

  3. [3]

    Alessandri, P. and H. Mumtaz (2019): Financial regimes and uncertainty shocks, Journal of Monetary Economics, 101, 31--46

    work page 2019

  4. [4]

    (2024): Weighted scoringrules: emphasizing particular outcomes when evaluating probabilistic forecasts, Journal of Statistical Software, 110, 1--26

    Allen, S. (2024): Weighted scoringrules: emphasizing particular outcomes when evaluating probabilistic forecasts, Journal of Statistical Software, 110, 1--26

    work page 2024

  5. [5]

    Giannone, and L

    Ba \'n bura, M., D. Giannone, and L. Reichlin (2010): Large Bayesian vector auto regressions, Journal of applied Econometrics, 25, 71--92

    work page 2010

  6. [6]

    Boivin, and P

    Bernanke, B., J. Boivin, and P. S. Eliasz (2005): Measuring the Effects of Monetary Policy: A Factor-augmented Vector Autoregressive (FAVAR) Approach, The Quarterly Journal of Economics, 120, 387--422

    work page 2005

  7. [7]

    Mumtaz, and M

    Caldara, D., H. Mumtaz, and M. Zhong (2024): Risk in a Data-Rich Model, Unpublished manuscript

    work page 2024

  8. [8]

    Scotti, and M

    Caldara, D., C. Scotti, and M. Zhong (2021): Macroeconomic and Financial Risks: A Tale of Mean and Volatility , International Finance Discussion Papers 1326, Board of Governors of the Federal Reserve System (U.S.)

    work page 2021

  9. [9]

    Large Bayesian vector autoregressions with stochastic volatility and non-conjugate priors

    Carriero, A., J. Chan, T. E. Clark, and M. Marcellino (2022): Corrigendum to “Large Bayesian vector autoregressions with stochastic volatility and non-conjugate priors” [J. Econometrics 212 (1) (2019) 137–154], Journal of Econometrics, 227, 506--512

    work page 2022

  10. [10]

    Carriero, A., T. E. Clark, and M. Marcellino (2016): Common drifting volatility in large Bayesian VARs, Journal of Business & Economic Statistics, 34, 375--390

    work page 2016

  11. [11]

    --- -.1pt --- -.1pt --- (2018): Measuring Uncertainty and Its Impact on the Economy , The Review of Economics and Statistics, 100, 799--815

    work page 2018

  12. [12]

    --- -.1pt --- -.1pt --- (2024): Capturing macro-economic tail risks with Bayesian vector autoregressions, Journal of Money, Credit and Banking, 56, 1099--1127

    work page 2024

  13. [13]

    Tuzcuoglu, and L

    Castelnuovo, E., K. Tuzcuoglu, and L. Uzeda (2022): Sectoral Uncertainty , Tech. rep

    work page 2022

  14. [14]

    --- -.1pt --- -.1pt --- (2025): Sectoral Uncertainty: A Hierarchical-Volatility Approach, Journal of Business & Economic Statistics, 1--13

    work page 2025

  15. [15]

    Chan, J. C. (2023): Comparing stochastic volatility specifications for large Bayesian VARs, Journal of Econometrics, 235, 1419--1446

    work page 2023

  16. [16]

    Hauzenberger, H

    Chernis, T., N. Hauzenberger, H. Mumtaz, and M. Pfarrhofer (2025): A Bayesian Gaussian Process Dynamic Factor Model, arXiv preprint arXiv:2509.04928

    work page arXiv 2025

  17. [17]

    Chiu, C.-W. J., H. Mumtaz, and G. Pintér (2017): Forecasting with VAR Models: Fat Tails and Stochastic Volatility, International Journal of Forecasting, 33, 1124--1143

    work page 2017

  18. [18]

    Litterman, and C

    Doan, T., R. Litterman, and C. Sims (1984): Forecasting and conditional projection using realistic prior distributions, Econometric Reviews, 3, 1--100

    work page 1984

  19. [19]

    (1993): Bayesian Treatment of the Independent Student-t Linear Model, Journal of Applied Econometrics, 8, S19--40

    Geweke, J. (1993): Bayesian Treatment of the Independent Student-t Linear Model, Journal of Applied Econometrics, 8, S19--40

    work page 1993

  20. [20]

    Geweke, J. and G. Amisano (2010): Comparing and evaluating Bayesian predictive distributions of asset returns, International Journal of Forecasting, 26, 216--230

    work page 2010

  21. [21]

    Reichlin, and D

    Giannone, D., L. Reichlin, and D. Small (2008): Nowcasting: The real-time informational content of macroeconomic data, Journal of monetary economics, 55, 665--676

    work page 2008

  22. [22]

    Balabdaoui, and A

    Gneiting, T., F. Balabdaoui, and A. E. Raftery (2007): Probabilistic forecasts, calibration and sharpness, Journal of the Royal Statistical Society Series B: Statistical Methodology, 69, 243--268

    work page 2007

  23. [23]

    Gneiting, T. and R. Ranjan (2011): Comparing density forecasts using threshold-and quantile-weighted scoring rules, Journal of Business & Economic Statistics, 29, 411--422

    work page 2011

  24. [24]

    Ha, J., M. A. Kose, and F. Ohnsorge (2023): One-stop source: A global database of inflation, Journal of International Money and Finance, 137, 102896

    work page 2023

  25. [25]

    (2006): Asset allocation with a high dimensional latent factor stochastic volatility model, The Review of Financial Studies, 19, 237--271

    Han, Y. (2006): Asset allocation with a high dimensional latent factor stochastic volatility model, The Review of Financial Studies, 19, 237--271

    work page 2006

  26. [26]

    Jo, S. and R. Sekkel (2019): Macroeconomic uncertainty through the lens of professional forecasters, Journal of Business & Economic Statistics, 37, 436--446

    work page 2019

  27. [27]

    Jurado, K., S. C. Ludvigson, and S. Ng (2015): Measuring uncertainty, American Economic Review, 105, 1177--1216

    work page 2015

  28. [28]

    Kim, C. J. and C. R. Nelson (1998): State-Space Models with Regime-Switching: Classical and Gibbs-Sampling Approaches with Applications, MIT Press

    work page 1998

  29. [29]

    (2003): Bayesian econometrics, Chichester, England : Wiley & Sons

    Koop, G. (2003): Bayesian econometrics, Chichester, England : Wiley & Sons

    work page 2003

  30. [30]

    Lindsten, F., M. I. Jordan, and T. B. Sch \"o n (2014): Particle Gibbs with Ancestor Sampling, Journal of Machine Learning Research, 15, 2145--2184

    work page 2014

  31. [31]

    Lopes, H. F. and C. M. Carvalho (2007): Factor stochastic volatility with time varying loadings and Markov switching regimes, Journal of Statistical Planning and Inference, 137, 3082--3091

    work page 2007

  32. [32]

    Ludvigson, S. C., S. Ma, and S. Ng (2021): Uncertainty and business cycles: exogenous impulse or endogenous response? American Economic Journal: Macroeconomics, 13, 369--410

    work page 2021

  33. [33]

    McCracken, M. W. and S. Ng (2016): FRED-MD: A monthly database for macroeconomic research, Journal of Business & Economic Statistics, 34, 574--589

    work page 2016

  34. [34]

    (2018): A generalised stochastic volatility in mean VAR, Economics Letters, 173, 10--14

    Mumtaz, H. (2018): A generalised stochastic volatility in mean VAR, Economics Letters, 173, 10--14

    work page 2018

  35. [35]

    Mumtaz, H. and A. Musso (2021): The evolving impact of global, region-specific, and country-specific uncertainty, Journal of Business & Economic Statistics, 39, 466--481

    work page 2021

  36. [36]

    Mumtaz, H. and P. Surico (2012): Evolving international inflation dynamics: world and country-specific factors, Journal of the European Economic Association, 10, 716--734

    work page 2012

  37. [37]

    Shin, M. and M. Zhong (2020): A new approach to identifying the real effects of uncertainty shocks, Journal of Business & Economic Statistics, 38, 367--379

    work page 2020

  38. [38]

    Sims, C. A. and T. Zha (1998): Bayesian methods for dynamic multivariate models, International Economic Review, 39, 949--968

    work page 1998

  39. [39]

    Stock, J. H. and M. W. Watson (2002): Macroeconomic Forecasting Using Diffusion Indexes, Journal of Business & Economic Statistics, 20, 147--62

    work page 2002

  40. [40]

    --- -.1pt --- -.1pt --- (2012): Disentangling the Channels of the 2007-2009 Recession, NBER Working Papers 18094, National Bureau of Economic Research, Inc

    work page 2012