State-Space Representation of INGARCH Models and Their Application in Insurance
Pith reviewed 2026-05-17 21:15 UTC · model grok-4.3
The pith
INGARCH models for count time series arise as special cases of a marginalized state-space model.
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
We first introduce the marginalized state-space model (M-SSM), defined solely through the marginal distribution of the observations, and show that INGARCH models arise as special cases of this framework. The M-SSM formulation facilitates the natural incorporation of covariates and missing data mechanisms, and this representation in turn provides a coherent way to incorporate these elements within the INGARCH model as well. We then demonstrate that an M-SSM can admit an observation-driven state-space model (O-SSM) representation when suitable assumptions are imposed on the evolution of its conditional mean, providing a natural setting for establishing weak stationarity even in the presence of
What carries the argument
The marginalized state-space model (M-SSM) defined through marginal observation distributions, which lifts to an observation-driven state-space model (O-SSM) under assumptions on conditional mean evolution.
If this is right
- INGARCH models gain a coherent way to include covariates through the state-space representation.
- Missing observations can be handled within INGARCH models without burdensome computation.
- Weak stationarity holds for INGARCH models even under heterogeneity and missing data.
- The Poisson and Negative-Binomial INGARCH(1,1) models admit this representation for predictive insurance analysis.
Where Pith is reading between the lines
- The framework could connect INGARCH to other state-space approaches for count data in fields like epidemiology.
- Estimation routines might be derived more directly from the state-space form than from the original INGARCH recursion.
- The approach may extend to higher-order INGARCH models or other discrete distributions beyond the two illustrated cases.
Load-bearing premise
Suitable assumptions must hold on the evolution of the conditional mean for the marginalized state-space model to admit an observation-driven state-space representation.
What would settle it
A direct comparison on data with covariates or missing values where the INGARCH likelihood or predictions diverge from those obtained via the M-SSM representation would falsify the claimed equivalence and lifting.
read the original abstract
Integer-valued generalized autoregressive conditional heteroskedastic (INGARCH) models are a popular framework for modeling serial dependence in count time-series. While convenient for modeling, prediction, and estimation, INGARCH models lack a clear theoretical justification for the evolution step. This limitation not only makes interpretation difficult and complicates the inclusion of covariates, but can also make the handling of missing data computationally burdensome. Consequently, applying such models in an insurance context, where covariates and missing observations are common, can be challenging. In this paper, we first introduce the marginalized state-space model (M-SSM), defined solely through the marginal distribution of the observations, and show that INGARCH models arise as special cases of this framework. The M-SSM formulation facilitates the natural incorporation of covariates and missing data mechanisms, and this representation in turn provides a coherent way to incorporate these elements within the INGARCH model as well. We then demonstrate that an M-SSM can admit an observation-driven state-space model (O-SSM) representation when suitable assumptions are imposed on the evolution of its conditional mean. This lifting from an M-SSM to an O-SSM provides a natural setting for establishing weak stationarity, even in the presence of heterogeneity and missing observations. The proposed ideas are illustrated through the Poisson and the Negative-Binomial INGARCH(1,1) models, highlighting their applicability in predictive analysis for insurance data.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper introduces the marginalized state-space model (M-SSM), defined solely via the marginal distribution of the observations, and shows that INGARCH models arise as special cases. It then asserts that an M-SSM admits an observation-driven state-space model (O-SSM) representation under suitable assumptions on the evolution of the conditional mean; this lifting is used to establish weak stationarity for INGARCH models even with heterogeneity, covariates, and missing observations. The ideas are illustrated on Poisson and Negative-Binomial INGARCH(1,1) models in an insurance context.
Significance. If the M-SSM to O-SSM lifting can be made rigorous with explicitly stated and verified assumptions that genuinely accommodate covariates and missing data without circularity, the framework would supply a missing theoretical justification for the INGARCH recursion and a practical route for handling incomplete insurance time series. The approach could also clarify the relationship between observation-driven and parameter-driven count models.
major comments (1)
- The central lifting step (M-SSM to O-SSM representation) relies on 'suitable assumptions' imposed on the evolution of the conditional mean, yet these assumptions are not stated explicitly enough to verify that they are satisfied by the Poisson and Negative-Binomial INGARCH(1,1) examples or that they deliver stationarity under heterogeneity and missingness without simply reproducing the original INGARCH recursion. This is the load-bearing claim for the paper's contribution to stationarity and missing-data handling.
minor comments (1)
- The abstract and introduction would benefit from a brief forward reference to the precise location where the 'suitable assumptions' are formalized and checked for the two INGARCH examples.
Simulated Author's Rebuttal
We thank the referee for the careful and constructive report. The major comment correctly identifies that greater explicitness is needed around the lifting assumptions; we will address this directly in revision.
read point-by-point responses
-
Referee: The central lifting step (M-SSM to O-SSM representation) relies on 'suitable assumptions' imposed on the evolution of the conditional mean, yet these assumptions are not stated explicitly enough to verify that they are satisfied by the Poisson and Negative-Binomial INGARCH(1,1) examples or that they deliver stationarity under heterogeneity and missingness without simply reproducing the original INGARCH recursion. This is the load-bearing claim for the paper's contribution to stationarity and missing-data handling.
Authors: We agree that the assumptions must be stated and verified explicitly rather than left as 'suitable.' In the revised manuscript we will add a dedicated subsection that lists the precise conditions on the conditional-mean evolution (a linear autoregressive structure with a uniform contraction factor less than one, together with a Lipschitz condition on the link function) required for an M-SSM to admit an O-SSM representation. We will then substitute the Poisson and negative-binomial INGARCH(1,1) recursions into these conditions and confirm that they hold. Stationarity will be established from the resulting state-space contraction mapping, which continues to apply when covariates enter the mean equation and when observations are missing at random; the missing-data case is handled by replacing the missing observation with its conditional expectation under the marginal distribution, without presupposing the INGARCH update rule. This derivation is therefore non-circular: the INGARCH recursion is recovered as the marginal evolution of the conditional mean, while stationarity follows from the state-space properties. revision: yes
Circularity Check
M-SSM defined via marginals with INGARCH as special case; O-SSM lift under independent assumptions
full rationale
The derivation begins by defining the M-SSM solely through the marginal distribution of the observations, an independent starting point that does not presuppose the INGARCH recursion. INGARCH models are then shown to arise as special cases within this framework, which is a containment result rather than a definitional equivalence. The lifting to an O-SSM representation is achieved by imposing additional assumptions on the evolution of the conditional mean; these assumptions are external to the marginal definition and enable the stationarity analysis with heterogeneity and missing data. No equation or step reduces the claimed results to the inputs by construction, and the structure remains self-contained against external benchmarks for count time series.
Axiom & Free-Parameter Ledger
axioms (1)
- domain assumption The marginal distribution of observations defines the model
invented entities (2)
-
Marginalized state-space model (M-SSM)
no independent evidence
-
Observation-driven state-space model (O-SSM)
no independent evidence
Lean theorems connected to this paper
-
IndisputableMonolith/Cost/FunctionalEquation.leanwashburn_uniqueness_aczel echoes?
echoesECHOES: this paper passage has the same mathematical shape or conceptual pattern as the Recognition theorem, but is not a direct formal dependency.
We then demonstrate that an M-SSM can admit an observation-driven state-space model (O-SSM) representation when suitable assumptions are imposed on the evolution of its conditional mean. This lifting from an M-SSM to an O-SSM provides a natural setting for establishing weak stationarity... mt(Θt)|Z ≤t ⪯cx Θt+1 |Z ≤t
What do these tags mean?
- matches
- The paper's claim is directly supported by a theorem in the formal canon.
- supports
- The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
- extends
- The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
- uses
- The paper appears to rely on the theorem as machinery.
- contradicts
- The paper's claim conflicts with a theorem or certificate in the canon.
- unclear
- Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.
Reference graph
Works this paper leans on
-
[1]
Abdallah, A., Boucher, J.-P., and Cossette, H. (2016). Sarmanov family of multivariate distributions for bivariate dynamic claim counts model.Insurance: Mathematics and Economics, 68:120–133. Ahn, J. Y., Jeong, H., Lu, Y., and Wüthrich, M. V. (2025). An observation-driven state-space count model for experience rating.Insurance: Mathematics and Economics, ...
work page 2016
-
[2]
Jasiak, J. (1999). Persistence in intertrade durations.Available at SSRN 162008. Jørgensen, B. (1987). Exponential dispersion models.Journal of the Royal Statistical Society Series B: Statistical Methodology, 49(2):127–145. Knape, J., Jonzén, N., and Sköld, M. (2011). On observation distributions for state space models of population survey data.Journal of...
work page 1999
-
[3]
Lee, C.-C., Hsu, Y.-C., and Lee, C.-C. (2010). An empirical analysis of non-life insurance consump- tion stationarity.The Geneva Papers on Risk and Insurance-Issues and Practice, 35(2):266–289. Leskelä, L. and Vihola, M. (2017). Conditional convex orders and measurable martingale couplings. Bernoulli, 23(4A):2784–2807. Nelder, J. A. and Verrall, R. J. (19...
work page 2010
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.