Paper: arXiv 2508.20101
Authors: Atika Aouri, Philipp Otto
Abstract
We introduce a heterogeneous spatiotemporal GARCH model for geostatistical data or processes on networks, e.g., for modelling and predicting financial return volatility across firms in a latent spatial framework. The model combines classical GARCH(p, q) dynamics with spatially correlated innovations and spatially varying parameters, estimated using local likelihood methods. Spatial dependence is introduced through a geostatistical covariance structure on the innovation process, capturing contemporaneous cross-sectional correlation. This dependence propagates into the volatility dynamics via the recursive GARCH structure, allowing the model to reflect spatial spillovers and contagion effects in a parsimonious and interpretable way. In addition, this modelling framework allows for spatial volatility predictions at unobserved locations. In an empirical application, we demonstrate how the model can be applied to financial stock networks. Unlike other spatial GARCH models, our framework does not rely on a fixed adjacency matrix; instead, spatial proximity is defined in a proxy space constructed from balance sheet characteristics. Using daily log returns of 50 publicly listed firms over a one-year period, we evaluate the model’s predictive performance in a cross-validation study.
Complexity vs Empirical Score
- Math Complexity: 8.0/10
- Empirical Rigor: 4.0/10
- Quadrant: Lab Rats — high math complexity, limited empirical testing
Why this score: The paper introduces a highly advanced spatiotemporal GARCH model with spatially varying parameters, geostatistical covariance structures, and a latent economic space, requiring advanced stochastic calculus and local likelihood estimation. While it includes a cross-validation study, it is limited to a single year of data for only 50 firms, lacks implementation details for backtesting, and does not report robust performance metrics like those needed for production-ready strategies.
Research Flowchart
flowchart TD A["Research Goal:<br>Model & Predict Volatility in Financial Networks"] --> B["Data Input:<br>Daily log returns of 50 firms over 1 year"] B --> C["Methodology:<br>Heterogeneous Spatiotemporal GARCH Model"] C --> D["Computational Process:<br>Local Likelihood Estimation"] D --> E["Key Outcome 1:<br>Spatial volatility predictions at unobserved locations"] D --> F["Key Outcome 2:<br>Parsimonious spatial spillover & contagion effects"] E --> G["Validation:<br>Cross-validation study"] F --> G
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 heterogeneous spatiotemporal GARCH model that combines classical GARCH dynamics with spatially correlated innovations and spatially varying parameters, estimated via local likelihood methods.
Source sentence
We introduce a heterogeneous spatiotemporal GARCH model for geostatistical data or processes on networks, e.g., for modelling and predicting financial return volatility across firms in a latent spatial framework.
Assumes 6
- The innovation process {ηt(s)} is independent across time, strictly stationary and ergodic, with zero mean and unit variance. statistical
Source sentence
The innovation process {ηt(s)} is assumed to be independent across time, strictly stationary and ergodic, with zero mean and unit variance, and exhibits spatial dependence through a valid covariance function on D.
- Spatial dependence of innovations follows a Matérn covariance function. model
Source sentence
In what follows, we consider a covariance function of the Matérn class (Gneiting et al., 2010).
- The spatial domain is Euclidean D = [0,1]^2 and time index is discrete and equidistant. model
Source sentence
In the present work, without loss of generality, we focus on a Euclidean domain D = [0, 1]^2 ⊂R2 and assume a discrete, equidistant time index t ∈T = Z.
- Parameter functions ω(s), αi(s), βj(s) vary smoothly over space. model
Source sentence
where ω(s), αi(s), and βj(s) are location-specific parameters that vary smoothly over space
Show 2 more
- The spatial weighting kernel K integrates to one. statistical
Source sentence
The spatial weighting function is constructed from a kernel K : R2 →R satisfying ∫
- Balance‑sheet features are standardized across firms and missing values are imputed with zero. data
Source sentence
Each feature was standardised across firms, and missing values were imputed with zero
Data
- Financial dataset of 50 firms fundamentals public
Source sentence
The financial dataset consists of 50 firms traded on public markets, observed over 276 trading days in a single year.
- Balance‑sheet variables fundamentals public
Source sentence
For each firm, 66 balance sheet variables were extracted. After removing variables with more than 5% missing values, 44 features remained.
Costs
Not applicable
Finds 2
- Simulation results show low bias and RMSE, and accurate recovery of spatially varying parameters and volatility predictions. predictive-accuracy
Source sentence
Overall, the simulation results demonstrate that the proposed STGARCH estimation and prediction framework reliably recovers spatially varying parameters and accurately predicts both unconditional and conditional volatilities across a range of spatial sample sizes and time series lengths.
- Empirical application captures economically interpretable spatial heterogeneity and delivers competitive predictive performance with computational efficiency. performance
Source sentence
the model captured economically interpretable spatial heterogeneity in baseline risk, shock sensitivity, and persistence, and delivered competitive predictive performance while remaining computationally efficient via a two–stage estimation–kriging procedure.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 29 related papers). Coverage is partial, so this is neither replication nor refutation.
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 (32)
- Probabilistic insights into spatio-temporal GARCH model with location-dependent parameters (2025)
- A general framework for spatial GARCH models (2023)
- A review of regularised estimation methods and cross-validation in spatiotemporal statistics (2024)
- Stationary nonseparable space-time covariance functions on networks (2023)
- Spatio-temporal covariance and cross-covariance functions of the great circle distance on a sphere (2016)
- Mat 'e rn cross-covariance functions for multivariate random fields (2010)
- Dynamic spatiotemporal ARCH models (2024)
- Spatial GARCH : A spatial approach to multivariate volatility modeling (2012)
and 24 more
- Networks in risk spillovers: A multivariate GARCH perspective (2023)
- Proximity-structured multivariate volatility models (2015)
- Generalised spatial and spatiotemporal autoregressive conditional heteroscedasticity (2018)
- Scenario analysis of livestock-related PM2.5 pollution based on a new heteroskedastic spatiotemporal model (2024)
- A dynamic spatiotemporal stochastic volatility model with an application to environmental risks (2023)
- A class of stochastic volatility models for environmental applications (2011)
- Network log- ARCH models for forecasting stock market volatility — → 2303.11064
- Tail event driven networks of SIFIs (2019)
- On the past, present, and future of the Diebold–Yilmaz approach to dynamic network connectedness (2023)
- GARCH processes: structure and estimation (2003)
- Statistical inference for time-varying ARCH processes (2006)
- Spatial GARCH models for unknown spatial locations–an application to financial stock returns (2024)
- Spatial and spatiotemporal volatility models: A review — → 2308.13061
- A literature review of spatio-temporal data analysis (2021)
- Autoregressive conditional heteroscedasticity with estimates of the variance of United Kingdom inflation (1982)
- Generalized autoregressive conditional heteroskedasticity (1986)
- Spatial autoregressive conditional heteroskedasticity models (2017)
- Spatial GARCH models (2018)
- Spatiotemporal ARCH models (2018)
- A Stationary Spatio-Temporal GARCH Model (2020)
- Spatial and Spatiotemporal GARCH Models–A Unified Approach (2019) — link
- Statistical analysis of a spatio-temporal model with location-dependent parameters and a test for spatial stationarity (2008)
- Martingale limit theory and its application (1980)
- Local likelihood estimation (1987)
2 of 32 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.