Paper: arXiv 2310.11987

Authors: Georgios I. Papayiannis

Abstract

The problem of asset liability management (ALM) is a classic problem of the financial mathematics and of great interest for the banking institutions and insurance companies. Several formulations of this problem under various model settings have been studied under the Mean-Variance (MV) principle perspective. In this paper, the ALM problem is revisited under the context of model uncertainty in the one-stage framework. In practice, uncertainty issues appear to several aspects of the problem, e.g. liability process characteristics, market conditions, inflation rates, inside information effects, etc. A framework relying on the notion of the Wasserstein barycenter is presented which is able to treat robustly this type of ambiguities by appropriate handling the various information sources (models) and appropriately reformulating the relevant decision making problem. The proposed framework can be applied to a number of different model settings leading to the selection of investment portfolios that remain robust to the various uncertainties appearing in the market. The paper is concluded with a numerical experiment for a static version of the ALM problem, employing standard modelling approaches, illustrating the capabilities of the proposed method with very satisfactory results in retrieving the true optimal strategy even in high noise cases.

Complexity vs Empirical Score

  • Math Complexity: 9.5/10
  • Empirical Rigor: 3.0/10
  • Quadrant: Lab Rats — high math complexity, limited empirical testing

Why this score: The paper employs advanced mathematics, including stochastic differential equations, filtration theory, and Wasserstein barycenters for model uncertainty, indicating high mathematical complexity. However, the empirical section relies on a simple static numerical experiment with standard models and lacks detailed implementation, code, or backtesting data, resulting in low empirical rigor.

Research Flowchart

  flowchart TD
  A["Research Goal<br>Treat model uncertainty<br>in ALM problem"] --> B["Methodology<br>Wasserstein Barycenter Framework"]
  B --> C["Inputs<br>Multiple financial models<br>Liability & asset data"]
  C --> D["Computation<br>Robust optimization<br>reformulation"]
  D --> E{"Outcome"}
  E --> F["Findings: Robust portfolio<br>True optimal strategy recovered<br>High noise tolerance"]

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 framework based on the Wasserstein barycenter to robustly aggregate multiple prior models and derive optimal investment portfolios for the asset‑liability management problem under model uncertainty.

Source sentence
A framework relying on the notion of the Wasserstein barycenter is presented which is able to treat robustly this type of ambiguities by appropriate handling the various information sources (models) and appropriately reformulating the relevant decision making problem. The proposed framework can be applied to a number of different model settings leading to the selection of investment portfolios that remain robust to

Assumes 6

  • The financial market is assumed to be free of arbitrage, frictionless and complete. market
    Source sentence
    We typically assume that the market is free of arbitrage, frictionless and complete (however the latter assumption could be dropped in order to provide a more general framework).
  • The market consists of a risk‑free bond and n risky assets whose prices follow geometric Brownian motion dynamics. model
    Source sentence
    Assume that the financial market consists of a risk-free asset (a bond) and n risky assets. The typical model that is employed for describing the risky asset’s dynamics is the geometric Brownian motion (GBM) model.
  • The liability (or claim) process is modeled as a Brownian motion with drift, possibly including additional risk factors. model
    Source sentence
    Assume further that the financial institution/insurance company considers a Brownian Motion (BM) model with drift as an aggregate model for its liability process (or in other terms of its cumulative claim process) L(·) of the form dL(t) = αdt + βT dW (t) + γTdB(t), L(0) = ℓ0 > 0 (5) where α denotes the drift (average claim rate) for the aggregate liability process, β := (β1, ...,
  • The portfolio decision is static: the allocation is chosen at time 0 and held unchanged until the horizon T. execution
    Source sentence
    The construction of an investment portfolio X(t) to cover the liabilities that will occur at the future time t = T for the insurance company is considered here in a static setting. That means that at time t = 0 needs to be determined how a certain amount of wealth, x0 ∈R, should be optimally allocated by the risk management desk to the available assets in order to minimize the effect of future short positions due to
Show 2 more
  • The prior set of models is assumed to contain sufficient information to approximate the true (unknown) probability model. data
    Source sentence
    In what follows, the prior set is assumed that provides sufficient information to retrieve the true situation, i.e. the true probability model (unknown) should be able to be approximated by the models provided by the prior set.
  • When the prior set is sufficiently representative, the true optimal wealth allocations can be recovered. other
    Source sentence
    The successful handling of the multiple information from the proposed scheme becomes more clear in Table 3 where as the prior set size grows, the gap in the expected surplus amount and the surplus variance optimal estimations becomes smaller.

Data

  • Synthetic prior‑model set and aggregate model Q* synthetic unknown
    Source sentence
    The numerical experiment is performed by generating the prior set for each one of the cases under consideration, obtaining the aggregate model Q∗under equal weights and then calculating the optimal portfolio through the relation (8) after generating M = 100, 000 random samples from Q∗.

Costs

Not modeled A trading or execution result is reported without costs.

Source sentence
The static setting is in many cases much preferable than the continuous allocation of wealth amounts to the various available assets (especially in relative short time periods) to avoid extra transaction costs or due to other reasons, e.g. regulations, market imperfections, liquidity constraints, etc.

Finds 4

  • The proposed method retrieves the true optimal strategy even in high‑noise cases. theoretical
    Source sentence
    The paper is concluded with a numerical experiment for a static version of the ALM problem, employing standard modelling approaches, illustrating the capabilities of the proposed method with very satisfactory results in retrieving the true optimal strategy even in high noise cases.
  • Accuracy of the estimated optimal wealth allocations improves markedly when the prior set contains five or more models. performance
    Source sentence
    For prior sets constituting of five or more models, the method’s accuracy seems to be significantly improved, especially in recovering the true optimal wealth amount allocations to the various hedging instruments.
  • Larger prior‑set sizes reduce the gap between estimated and true surplus expectations and variances, indicating robust behaviour. performance
    Source sentence
    The successful handling of the multiple information from the proposed scheme becomes more clear in Table 3 where as the prior set size grows, the gap in the expected surplus amount and the surplus variance optimal estimations becomes smaller. This is an evidence of a robust behaviour indicating that the more the available models, the more robust the derived optimal decision becomes to model ambiguity.
  • Smaller prior‑set sizes lead to higher standard errors and wealth allocations that deviate significantly from the true ones. negative
    Source sentence
    When smaller prior set sizes are used, the standard errors seems to naturally increase due to the effect of possible extreme opinions, while the wealth proportion allocations significantly differ from the true ones, especially in cases of one prior opinions in the medium and low homogeneity scenarios.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 23 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.