Paper: arXiv 2412.03668
Authors: Beatrice Foroni, Luca Merlo, Lea Petrella
Abstract
In this paper we develop a novel hidden Markov graphical model to investigate time-varying interconnectedness between different financial markets. To identify conditional correlation structures under varying market conditions and accommodate stylized facts embedded in financial time series, we rely upon the generalized hyperbolic family of distributions with time-dependent parameters evolving according to a latent Markov chain. We exploit its location-scale mixture representation to build a penalized EM algorithm for estimating the state-specific sparse precision matrices by means of an \(L_1\) penalty. The proposed approach leads to regime-specific conditional correlation graphs that allow us to identify different degrees of network connectivity of returns over time. The methodology’s effectiveness is validated through simulation exercises under different scenarios. In the empirical analysis we apply our model to daily returns of a large set of market indexes, cryptocurrencies and commodity futures over the period 2017-2023.
Complexity vs Empirical Score
- Math Complexity: 8.5/10
- Empirical Rigor: 7.0/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper employs advanced statistical techniques including generalized hyperbolic distributions, location-scale mixture representations, and penalized EM algorithms, demonstrating high mathematical complexity. It validates the methodology through simulation studies and applies it to real-world financial data (indices, cryptocurrencies, commodities) over a defined period, showing strong empirical rigor.
Research Flowchart
flowchart TD A["Research Goal<br>Model Time-varying Market Interconnectedness<br>under Dynamic, Heavy-tailed Conditions"] --> B["Methodology: Hidden Markov Graphical Model<br>with State-dependent Generalized Hyperbolic Distributions"] B --> C["Computational Process<br>Penalized EM Algorithm<br>(L1 Penalty for Sparse Precision Matrices)"] C --> D["Simulation Validation<br>Test Model Recovery under Various Scenarios"] C --> E["Empirical Application<br>Daily Returns (2017-2023)<br>Indices, Crypto, Commodities"] D --> F["Key Outcomes"] E --> F F --> G["Regime-specific Conditional Correlation Graphs"] F --> H["Identification of Time-varying Network Connectivity"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes a sparse hidden Markov graphical model with state‑dependent symmetric generalized hyperbolic emission distributions, estimated via a penalized EM/ECME algorithm with an L1 penalty on state‑specific precision matrices.
Source sentence
In this paper, we contribute to the existing literature by introducing a sparse hidden Markov graphical model to investigate time-varying conditional correlation structures in multivariate time series data, without assuming normally distributed returns. To build our network model, we consider multivariate symmetric generalized hyperbolic (GH, McNeil et al. 2015) distributions with time-dependent parameters evolving
Assumes 6
- Emissions follow a symmetric multivariate generalized hyperbolic distribution with time‑dependent parameters. model
Source sentence
To build our network model, we consider multivariate symmetric generalized hyperbolic (GH, McNeil et al. 2015) distributions with time-dependent parameters evolving according to a discrete, homogeneous latent Markov chain.
- The latent Markov chain is discrete, homogeneous and has a fixed number of states K. model
Source sentence
To build our network model, we consider multivariate symmetric generalized hyperbolic (GH, McNeil et al. 2015) distributions with time-dependent parameters evolving according to a discrete, homogeneous latent Markov chain.
- Precision matrices are sparse and can be recovered using an L1 (Lasso) penalty. statistical
Source sentence
We exploit its location-scale mixture representation to build a penalized EM algorithm for estimating the state-specific sparse precision matrices by means of an $L_1$ penalty. Since we are interested in detecting only the most important connections, we also propose a penalized version of the EM by considering the Lasso-type regularization of Tibshirani (1996) for sparse matrix estimation in high-dimensional
- Daily returns are stationary (ADF test at 1% level). data
Source sentence
In concluding, the Augmented Dickey-Fuller (ADF) test (Dickey & Fuller 1979) shows that all daily returns are stationary at the 1% level of significance.
Show 2 more
- Returns are non‑Gaussian and exhibit heavy tails, justifying the use of GH family instead of normality. model
Source sentence
Nevertheless, “non-standard” features of returns cannot be ac-commodated by standard models such as those based on normality assumptions. In this paper, we contribute to the existing literature by introducing a sparse hidden Markov graphical model to investigate time-varying conditional correlation structures in multivariate time series data, without assuming normally distributed returns.
- The mixture representation of GH allows use of Gaussian graphical tools (glasso) within the M‑step. execution
Source sentence
To induce sparsity and recover only the most prominent relations, we develop a Penalized ECME algorithm exploiting the Gaussian location-scale mixture representation of the GH distribution with a Lasso L1 penalty on the elements of the state-specific precision matrices. Our procedure is computationally efficient and can be implemented by making use of the glasso of Friedman et al. (2008) within the conditional M-step
Data
- Daily returns of 29 financial assets (stock indices, commodity futures, cryptocurrencies) market-prices public
Source sentence
In this paper, we are interested in investigating how the degree of network connectivity across different asset classes changes in severe market turbulence, which can possibly threaten the integrity of the financial system and represent a potential source of financial instability. For this purpose, we apply the proposed methodology to daily returns of d = 29 financial assets comprising stock market indices, commodity
- Synthetic data for simulation study synthetic unknown
Source sentence
4 Simulation study We conduct a simulation study to validate the performance of our model under different scenarios in terms of: 1. recovering the true values of the parameters for a varying number of hidden states, 2. clustering performance, 3. the ability to correctly retrieve the edge sets associated to each hidden state. Scenario 1: In the first scenario we consider a 2-dimensional vector Y t with sample size of T
Costs
Not applicable
Finds 5
- Simulation results show that the estimated location, scale, and GH parameters are very close to the true values across all six data‑generating models. statistical
Source sentence
For every model, the estimated location and scale parameters are very close to the true values. This is also true for the parameters (λk, χk, ψk) in every state.
- When dealing with non‑Gaussian heavy‑tailed data, the proposed dynamic graphical model matches the performance of the graphical lasso, offering a robust alternative for static and dynamic settings. performance
Source sentence
Overall, these results suggest that when we deal with non-Gaussian heavy tailed data, our static and dynamic graphical models have the same performance of the graphical lasso, representing a robust alternative to existing static approaches and a novelty for dynamic graphical models in non-Gaussian settings.
- Cryptocurrencies exhibit the highest volatility among the three asset classes, with the largest standard deviations and strong deviations from normality. statistical
Source sentence
As one can see, each asset displays the typical stylized facts that characterize the financial markets, and cryptocurrencies are generally much more volatile than commodity futures and stock market indices, having the highest standard deviations. Further, the Jarque-Bera test strongly rejects the normality assumption of daily returns for all series considered.
- The model successfully captures regime‑specific conditional correlation structures, revealing varying degrees of network connectivity over time. statistical
Source sentence
The proposed approach leads to regime-specific conditional correlation graphs that allow us to identify different degrees of network connectivity of returns over time. The proposed methodology allows us to capture regime-specific conditional correlation structures for probability distributions within the GH family.
- Clustering performance depends on the underlying data‑generating distribution, with slightly worse results for Cauchy‑generated data and better performance for fewer hidden states. performance
Source sentence
Firstly, we observe that the data generating model plays a fundamental role in estimating the true states partition, as we obtain slightly worst clustering performance for the model with the Cauchy distribution. Secondly, the goodness of the clustering obtained depends, as expected, on the number of states of the Markov chain considered, being the values slightly higher with K = 2 than with K = 3.
Checked by others
Supported by
- Complexity of Financial Time Series: Multifractal and Multiscale Entropy Analyses
Compare the findings
This paper: Cryptocurrencies exhibit the highest volatility among the three asset classes, with the largest standard deviations and strong deviations from normality.
That paper: Bitcoin and natural gas show wider probability density functions, reflecting higher volatility than GBP/USD and gold.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (148)
- CAV ia R : conditional autoregressive value at risk by regression quantiles (2004)
- A comparison of generalized hyperbolic distribution models for equity returns (2014)
- Estimating the tails of loss severity via conditional risk measures for the family of symmetric generalised hyperbolic distributions (2015)
- Modeling heavy-tailed stock index returns using the generalized hyperbolic distribution (2009)
- Nonparametric risk management with generalized hyperbolic distributions (2008)
- Time-varying beta risk of P an- E uropean industry portfolios: A comparison of alternative modeling techniques (2008)
- Expectile hidden M arkov regression models for analyzing cryptocurrency returns — → 2301.09722
- Quantile and expectile copula-based hidden M arkov regression models for the analysis of the cryptocurrency market — → 2307.06400
and 140 more
- The generalised hyperbolic distribution and its subclass in the analysis of a new era of cryptocurrencies: Ethereum and its financial risk (2019)
- A mixture of generalized hyperbolic distributions (2015)
- A dynamic analysis of stock markets using a hidden M arkov model (2013)
- Long memory of financial time series and hidden M arkov models with time-varying parameters (2017)
- Hidden M arkov and semi- M arkov models with multivariate leptokurtic-normal components for robust modeling of daily returns series (2019)
- Exploring the dependencies among main cryptocurrency log-returns: A hidden M arkov model (2022)
- Hidden M arkov and other models for discrete-valued time series (1997)
- Comparing partitions (1985)
- Penalized estimation in high-dimensional hidden M arkov models with state-specific graphical models (2013)
- On fitting mixture models (1999)
- Estimating the dimension of a model (1978)
- Distribution of the estimators for autoregressive time series with a unit root (1979)
- Confidence intervals for hidden M arkov model parameters (2000)
- The EM algorithm and extensions (2007)
- The ECME algorithm: a simple extension of EM and ECM with faster monotone convergence (1994)
- R: A Language and Environment for Statistical Computing (2023) — link
- Hidden M arkov models with nonelliptically contoured state densities (2010)
- Herding behaviour in cryptocurrencies (2019)
- A maximization technique occurring in the statistical analysis of probabilistic functions of M arkov chains (1970)
- The generalized hyperbolic family and automatic model selection through the multiple-choice LASSO (2023) — link
- Quantitative risk management: concepts, techniques and tools-revised edition (2015)
- Using time-series and sentiment analysis to detect the determinants of B itcoin prices (2015)
- What are the main drivers of the Bitcoin p rice? E vidence from wavelet coherence analysis (2015)
- A model of financialization of commodities (2016)
- Bitcoin, gold and the dollar–A GARCH volatility analysis (2016)
- Quantile Connectedness: modelling tail behaviour in the topology of financial networks (2018)
- Bitcoin: Medium of exchange or speculative assets? (2018)
- Co-explosivity in the cryptocurrency market (2019)
- Cryptocurrencies as hedges and safe-havens for US equity sectors (2020)
- Bitcoin, gold, and commodities as safe havens for stocks: New insight through wavelet analysis (2020)
- A rational expectations model of financial contagion (2002)
- Coronavirus (COVID-19)-An epidemic or pandemic for financial markets (2020)
- Financial markets under the global pandemic of COVID-19 (2020)
- On the hedge and safe haven properties of B itcoin: Is it really more than a diversifier? (2017)
- Network causality structures among B itcoin and other financial assets: A directed acyclic graph approach (2018)
- Crypto price discovery through correlation networks (2021)
- Lead behaviour in b itcoin markets (2020)
- Is Bitcoin a real currency? An economic appraisal (2015)
- Virtual currency schemes-a further analysis (2015)
- An analysis of the literature on systemic financial risk: A survey (2017)
- Global Financial Stability Report: Responding to the Financial Crisis and Measuring Systemic Risks (2009)
- Economic networks: The new challenges (2009)
- CoRisk: credit risk contagion with correlation network models (2018)
- Bank lending and systemic risk: A financial-real sector network approach with feedback (2018)
- Interconnectedness in the interbank market (2019)
- Axioms for centrality (2014)
- Dynamic asset trees and Black Monday (2003)
- Dynamics of market correlations: Taxonomy and portfolio analysis (2003)
- High-dimensional semiparametric G aussian copula graphical models (2012)
- Regularized rank-based estimation of high-dimensional nonparanormal graphical models (2012)
- The nonparanormal: Semiparametric estimation of high dimensional undirected graphs. (2009)
- Tiger: A tuning-insensitive approach for optimally estimating gaussian graphical models (2017)
- A SINful approach to G aussian graphical model selection (2008)
- Model selection for G aussian concentration graphs (2004)
- Covariance selection (1972)
- A constrained $\ell_1$ minimization approach to sparse precision matrix estimation (2011)
- Quantile graphical models: prediction and conditional independence with applications to systemic risk (2016) — link
- The cross-quantilogram: Measuring quantile dependence and testing directional predictability between time series (2016)
- Financial bubbles: a study of co-explosivity in the cryptocurrency market (2020)
- On use of the EM algorithm for penalized likelihood estimation (1990)
- Estimating global bank network connectedness (2018)
- Financial systemic risk measurement based on causal network connectedness analysis (2019)
- Exploring the dynamic relationships between cryptocurrencies and other financial assets (2018)
- Evolution of the bitcoin address graph (2017)
- Robust graphical modeling of gene networks using classical and alternative t-distributions (2011)
- Extended B ayesian information criteria for model selection with large model spaces (2008)
- High-dimensional I sing model selection with B ayesian information criteria (2015)
- Extended Bayesian information criteria for G aussian graphical models (2010)
- Mixtures of shifted asymmetriclaplace distributions (2013)
- Spillovers and Directional Predictability with a Cross - Quantilogram Analysis : The Case of U . S . and Chinese Agricultural Futures (2016) — link
- Asymmetric Risk Spillovers between Oil and Agricultural Commodities (2018) — link
- Connectedness of financial institutions in E urope: A network approach across quantiles (2020)
- Graphical models (1996)
- Regression Q uantiles (1978)
- Quantile R egression (2005)
- Conditional risk based on multivariate hazard scenarios (2018)
- On multivariate extensions of the conditional V alue-at- R isk measure (2015)
- A directional multivariate V alue at R isk (2015)
- Value at R isk, 3rd Ed.: The New Benchmark for Managing Financial Risk (2006)
- The sparse method of simulated quantiles: An application to portfolio optimization (2018)
- Market Risk Analysis, V alue at R isk M odels (2009)
- A M arkov regime switching approach for hedging energy commodities (2008)
- Quantile regression for longitudinal data (2004)
- Bayesian quantile regression (2001)
- Spatial quantile multiple regression using the asymmetric L aplace process (2012)
- On multivariate quantile regression (2003)
- VAR for V a R : Measuring tail dependence using multivariate regression quantiles (2015)
- MULTIVARIATE QUANTILES AND MULTIPLE-OUTPUT REGRESSION QUANTILES: FROM L $_ 1 $ OPTIMIZATION TO HALFSPACE DEPTH (2010)
- M-quantile regression for multivariate longitudinal data: analysis of the M illennium C ohort S tudy data (2016) — link
- Markov regime-switching quantile regression models and financial contagion detection (2016)
- The L aplace distribution and generalizations: a revisit with applications to communications, economics, engineering, and finance (2012)
- Multiple Risk Measures for Multivariate Dynamic Heavy-Tailed Models (2017)
- Conditional Risk Based on Multivariate Hazard Scenarios (2016)
- Multivariate M arkov- S witching models and tail risk interdependence (2013) — link
- C o V a R (2016)
- Joint estimation of conditional quantiles in multivariate linear regression models with an application to financial distress (2019)
- Mixed H idden M arkov models: an extension of the H idden M arkov model to the longitudinal data setting (2007)
- Latent M arkov models for longitudinal data (2012)
- Hidden M arkov models for time series: an introduction using R (2016)
- Mixed H idden M arkov quantile regression models for longitudinal data with possibly incomplete sequences (2018)
- Mixed hidden M arkov models for quantiles (2014)
- Generalized autoregressive score models with applications (2013)
- A non-homogeneous hidden M arkov model for precipitation occurrence (1999)
- Quantile regression for longitudinal data based on latent M arkov subject-specific parameters (2012)
- Maximum likelihood from incomplete data via the EM algorithm (1977)
- The adaptive lasso and its oracle properties (2006)
- Variable selection via nonconcave penalized likelihood and its oracle properties (2001)
- Nearly unbiased variable selection under minimax concave penalty (2010)
- Variable selection in robust regression models for longitudinal data (2012)
- Regularized partially functional quantile regression (2017)
- Quantile regression for longitudinal data using the asymmetric L aplace distribution (2006)
- Mixed-effects models for conditional quantiles with longitudinal data (2009)
- Numerical maximisation of likelihood: A neglected alternative to EM ? (2014)
- Regression Shrinkage and Selection via the L asso (1996)
- Mixed H idden M arkov models for longitudinal data: an overview (2011)
- Computational issues in parameter estimation for stationary hidden M arkov models (2008)
- Quantiles for counts (2005)
- Likelihood inference for a class of latent M arkov models under linear hypotheses on the transition probabilities (2006)
- Linear mixed models: a practical guide using statistical software (2014)
- Nonparametric maximum likelihood estimation of a mixing distribution (1978)
- Cross- C ountry A ssessment of S ystemic R isk in the E uropean S tock M arket: E vidence from a C o V a R A nalysis (2018) — link
- Selection of V alue at R isk models for energy commodities (2018)
- Regularization and variable selection via the elastic net (2005)
- The geometry of mixture likelihoods: a general theory (1983)
- A multivariate extension of the dynamic logit model for longitudinal data based on a latent M arkov heterogeneity structure (2009)
- Linear quantile mixed models (2014)
- Hidden M arkov M odels and the B aum- W elch algorithm (2003)
- Handbook of Q uantile R egression (2017)
- Longitudinal analysis of the strengths and difficulties questionnaire scores of the M illennium C ohort S tudy children in E ngland using M -quantile random-effects regression (2016)
- Model selection through sparse maximum likelihood estimation for multivariate G aussian or binary data (2008)
- Sparse inverse covariance estimation with the graphical L asso (2008)
- Forecasting V a R and ES using a joint quantile regression and its implications in portfolio allocation (2021)
- Quantile M ixed H idden M arkov M odels for multivariate longitudinal data: an application to children’s strengths and difficulties questionnaire scores (2022) — link
- Bayesian Tail Risk Interdependence Using Quantile Regression (2015)
- The structure and degree of dependence: A quantile regression approach (2013)
- Model selection and estimation in the G aussian graphical model (2007)
- High-dimensional graphs and variable selection with the lasso (2006)
- Estimation of symmetry-constrained G aussian graphical models: application to clustered dense networks (2015)
- Modeling systemic risk with Markov switching graphical SUR models (2019)
- The L aplace distribution and generalizations: a revisit with applications to communications, economics, engineering, and finance (2001)
2 of 148 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.