📈 economics

Comonotonic improvement under feasibility constraints

This paper demonstrates that while standard risk-sharing allocations are nondecreasing in aggregate loss, certain regulatory constraints like Value-at-Risk caps can destroy this property, and it proposes "componentwise convex-order solidity" as a sufficient condition to restore comonotonic improvement.

Original authors: Christopher Blier-Wong, Jean-Gabriel Lauzier

Published 2026-04-28

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Christopher Blier-Wong, Jean-Gabriel Lauzier

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The "Fair Share" Problem: When Rules Break the Logic of Risk

Imagine you and four friends are part of a community garden club. Every year, there’s a chance of a massive storm that could destroy the garden. To prepare, you all agree to a "Risk-Sharing Pool." If a storm hits, the total cost of repairs is split among you based on how much "risk" each person can handle.

In a perfect world, the math is simple: if the storm is huge, everyone’s "bill" goes up. If the storm is small, everyone’s "bill" stays low. This is called comonotonicity—the idea that everyone’s individual burden moves in the same direction as the total disaster. It feels fair, it’s easy to predict, and it prevents anyone from "gaming the system."

But this paper explores what happens when the rules of the game (regulations or contracts) get in the way of that natural logic.

1. The "Solid Ground" Rule (The Good News)

The authors first identify a "Golden Rule" for constraints. They call it Componentwise Convex-Order Solidity.

The Analogy: The Safety Net.
Imagine you are all walking across a tightrope. A "solid" rule is like a safety net that is wide and flexible. If you stumble slightly (meaning you take on a little less risk), the net is guaranteed to catch you and keep you on the rope. Because the net is "solid," the group can still find the most efficient way to share the risk without falling into chaos.

Most common rules are "solid." For example:

Hard Caps: "No one can pay more than $1,000."
Budget Limits: "The total pool can't spend more than $5,000."
Expected Shortfall: A sophisticated way of measuring risk that looks at the "worst-case scenarios."

When these rules are in place, the math still works beautifully. Even with limits, the best way to share the risk is still for everyone’s bill to move up and down together as the storm gets worse.

2. The "Broken Rules" (The Bad News)

The paper’s real breakthrough is showing how certain common rules actually break the logic of fairness. They identify two main ways this happens:

A. The "Secret Information" Problem (Private Deductibles)

Imagine if one friend says, "I won't pay a cent for any damage under $100; that's my personal responsibility."

Now, the group can't just look at the total storm damage to decide the split. They have to know exactly what happened to that specific person's patch of garden. Because the rule depends on private details rather than the big picture, the "fair share" math falls apart. You can no longer simplify the problem by just looking at the total disaster.

B. The "VaR" Trap (The Blind Spot)

This is the most important finding. Many big insurance companies and banks follow a rule called Value-at-Risk (VaR).

The Analogy: The "Almost-Perfect" Umbrella.
Imagine you have an umbrella that is amazing at blocking rain, but it has a tiny, specific hole that only lets water through if the rain is falling at a massive, once-in-a-century speed.

VaR is like that umbrella. It measures risk by looking at a specific "cutoff point" (say, the 99th percentile). It ignores everything in the extreme "tail" beyond that point.

Because VaR is "blind" to the absolute worst disasters, it creates a weird incentive. If a company is told, "You can't have a risk level higher than X," they might realize that they can take on massive, catastrophic risks as long as those risks happen in that "blind spot" where the VaR rule doesn't see them.

The Result: Instead of everyone's bill going up together during a disaster, the math might suggest that one person's bill should actually drop right when the disaster gets most intense, just to satisfy the "blind" rule. This is the opposite of what you want in a disaster recovery plan!

Summary: The Takeaway

The paper is a warning to regulators (the people who write the rules for banks and insurance companies).

It says: If you write rules that are "blind" to extreme risks (like VaR), you won't actually make the system safer. Instead, you will create a mathematical glitch where the most efficient way to manage risk is to act in ways that are unpredictable, counter-intuitive, and potentially dangerous.

To keep the system stable, regulators should use "solid" rules—rules that care about the entire spectrum of risk, from the small raindrops to the massive hurricanes.

Technical Summary: Comonotonic Improvement under Feasibility Constraints

Authors: Christopher Blier-Wong and Jean-Gabriel Lauzier
Date: April 28, 2026

1. The Problem

In classical risk-sharing theory (e.g., Borch, 1962; Ludkovski & Rüschendorf, 2008), when risk-averse agents share an aggregate loss $S$ , the Pareto-optimal allocation is comonotonic. This means each agent's share $X_i$ is a nondecreasing function of the aggregate loss $S$ . This property is mathematically powerful because it reduces a complex multi-dimensional optimization problem to a one-dimensional rearrangement problem.

However, real-world insurance and reinsurance markets are rarely unconstrained. Regulatory frameworks (like Solvency II or Basel III) and contractual terms (like reinsurance capacity limits or deductibles) impose constraints on individual exposures. The central problem addressed in this paper is: Under what conditions do these constraints preserve the comonotonicity of optimal allocations? If constraints break comonotonicity, the classical reduction fails, and optimal allocations may exhibit "counter-monotonic" behavior (where an agent's share decreases as the aggregate loss increases).

2. Methodology

The authors approach the problem through the lens of convex-order consistency. They assume agents have preferences (risk measures $\rho_i$ ) that are consistent with the convex order ( $\leq_{cx}$ ); essentially, if one distribution is less dispersed than another (mean-preserving contraction), the risk-averse agent prefers it.

The paper introduces a new structural property of the feasible set $\mathcal{X}$ called componentwise convex-order solidity.

Definition: A feasible set is "solid" if, whenever an allocation $X$ is feasible, any other allocation $Y$ that is "less risky" for every agent (i.e., $Y_i \leq_{cx} X_i$ for all $i$ ) is also feasible.

The authors use this property to bridge the gap between the unconstrained comonotonic improvement theorem and constrained optimization. They prove that if the feasible set is solid, the constrained inf-convolution (the minimum total risk) is equal to the comonotonic-restricted inf-convolution.

3. Key Contributions and Results

A. The Constrained Comonotonic Improvement Theorem (Theorem 2):
The authors prove that if the feasible set is nonempty and componentwise convex-order solid, then:

Every feasible allocation has a comonotonic counterpart that is also feasible and less risky in the convex order.
The optimal allocation for any convex-order-consistent preference can be found within the subclass of comonotonic allocations.

B. Identification of "Inadmissible" Constraints (Failure Modes):
The paper identifies three specific ways constraints break solidity and thus destroy comonotonicity:

Idiosyncratic State Dependence: Constraints that depend on an agent's own portfolio (e.g., a private deductible) rather than just the aggregate loss $S$ . This breaks the $\sigma(S)$ -measurability required for comonotonicity.
Non-Lipschitz Step-up Caps: Constraints that allow an agent's share to increase faster than the aggregate loss (e.g., a catastrophe top-up with a slope $> 1$ ). This forces the complementary agent's share to decrease as $S$ increases.
Value-at-Risk (VaR) Ceilings: Because VaR is not convex-order consistent (it is insensitive to the tail beyond the quantile), a VaR cap can reward agents for "pushing mass further out" into the tail. This leads to optimal allocations that are non-comonotonic.

C. Identification of "Admissible" Constraints (Theorem 3):
The authors prove that several standard operational constraints do preserve solidity:

Deterministic pathwise bounds (caps and floors).
Linear expectation constraints (e.g., premium budgets).
Orlicz-type bounds (e.g., $L^p$ norms, variance bounds).
Coherent risk-measure ceilings (e.g., Expected Shortfall/ES). Notably, replacing a VaR cap with an ES cap restores comonotonicity.

D. Application to Mean-Variance Risk Sharing:
The authors demonstrate the theory by solving the mean-variance problem under deterministic capacity caps. They derive a truncated-affine sharing rule, where agents share loss proportionally to their risk tolerances until they hit a cap, at which point they "saturate," and the remaining agents re-share the residual loss.

4. Significance

This paper provides a rigorous mathematical foundation for understanding how regulation distorts economic incentives in risk-sharing pools.

For Regulators: It provides a theoretical justification for using Expected Shortfall (ES) instead of Value-at-Risk (VaR). Using ES ensures that regulatory caps do not inadvertently incentivize "tail-pushing" or non-comonotonic behavior, which can destabilize the risk-sharing pool.
For Practitioners: It offers a "litmus test" (convex-order solidity) to determine if a complex reinsurance treaty or regulatory constraint will allow for simple, monotonic sharing rules or if it will result in complex, non-monotonic exposure patterns.
Theoretical Impact: It extends the classical literature on optimal risk sharing by moving from the "ideal" unconstrained case to a realistic, constrained framework, identifying the exact geometric property (solidity) required for mathematical tractability.