Paper: arXiv 2310.06151
Authors: Silvana M. Pesenti, Pietro Millossovich, Andreas Tsanakas
Abstract
Differential sensitivity measures provide valuable tools for interpreting complex computational models used in applications ranging from simulation to algorithmic prediction. Taking the derivative of the model output in direction of a model parameter can reveal input-output relations and the relative importance of model parameters and input variables. Nonetheless, it is unclear how such derivatives should be taken when the model function has discontinuities and/or input variables are discrete. We present a general framework for addressing such problems, considering derivatives of quantile-based output risk measures, with respect to distortions to random input variables (risk factors), which impact the model output through step-functions. We prove that, subject to weak technical conditions, the derivatives are well-defined and derive the corresponding formulas. We apply our results to the sensitivity analysis of compound risk models and to a numerical study of reinsurance credit risk in a multi-line insurance portfolio.
Complexity vs Empirical Score
- Math Complexity: 8.5/10
- Empirical Rigor: 3.0/10
- Quadrant: Lab Rats — high math complexity, limited empirical testing
Why this score: The paper presents a high-level theoretical framework with rigorous mathematical proofs involving generalised functions, quantile differentiation, and weak convergence, which is dense and advanced. While it includes numerical studies on reinsurance credit risk and compound models, the focus is on theoretical derivations rather than detailed backtesting protocols or public datasets.
Research Flowchart
flowchart TD A["Research Goal: Derive differential sensitivity for discontinuous models"] --> B["Key Methodology: Quantile-based risk measures with step-function inputs"] B --> C["Key Methodology: Derivatives w.r.t. random input distortions"] C --> D["Input: Compound Risk Models &<br>Multi-line Insurance Portfolio"] D --> E["Computational Process: Verify well-defined derivatives<br>& derive sensitivity formulas"] E --> F["Outcome: General framework for sensitivity analysis"] F --> G["Outcome: Application to reinsurance credit risk"]
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 general framework for differential sensitivity analysis of quantile‑based risk measures in models with discontinuities, deriving well‑defined derivative formulas.
Source sentence
We present a general framework for addressing such problems, considering derivatives of quantile-based output risk measures, with respect to distortions to random input variables (risk factors), which impact the model output through step-functions. We prove that, subject to weak technical conditions, the derivatives are well-defined and derive the corresponding formulas.
Assumes 2
- For all ε in a neighbourhood of 0 the distribution function Fε is continuously differentiable at the α‑quantile. model
Source sentence
For all ε in a neighbourhood of 0 the distribution function Fε is continuously differentiable at F −1(α). Assumption 2. Let 0 ≤α ≤1.
- The quantile function qα(ε) of the stressed loss is differentiable with respect to ε. model
Source sentence
The quantile function at level α of the stressed loss Lε, qα(ε), is differentiable with respect to ε, that is ∂ ∂εqα(ε) exists.
Data
- Compound model simulation synthetic unknown
Source sentence
The frequency W follows a Negative Binomial distribution with mean E[W] = 5 and over-dispersion Var(W)/E[W] = 2.5, truncated at the 99.9th percentile. The severities Yℓfollow Gamma distributions with shape parameter θ = 5, corresponding to a skewness coefficient of 0.894.
- Reinsurance credit risk simulation synthetic unknown
Source sentence
We use our sample of (X,Z,L | L > 0), which contains 500,000 simulated scenarios, and estimate the sensitivities using bootstrap with replacement and a bootstrap size of 450,000.
Costs
Not applicable
Finds 1
- In the compound model, the scaled sensitivities to loss frequency and severity are 0.414 and 0.429 respectively, indicating roughly equal impact on portfolio risk. theoretical
Source sentence
With these choices, we find that the sensitivities, scaled by the portfolio risk, take values ˜ SW [ ESα ] ESα(T ) = 0.414 and ˜ SY [ ESα ] ESα(T ) = 0.429.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 5 related papers). Coverage is partial, so this is neither replication nor refutation.
1 paper in the archive
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.