Reverse Stress Testing a Capital Decision

From a Capital Decision to a Failure Boundary

A forward stress test asks what happens to a proposed capital decision under specified adverse scenarios. Reverse stress testing starts with the opposite question: what combination of events would make the decision fail its stated solvency or rating objective? That question can reveal whether failure requires an extreme single shock or a combination of stresses that management should monitor.

The calculation begins by writing the objective in a form that can be evaluated under stress. For a solvency objective, a simple measure is the solvency margin: available capital less required capital. For a rating objective, the measure might be the excess of available capital over a required multiple of the relevant capital requirement. The objective is met when the measure is nonnegative and breached when it is negative.

Definition: A reverse stress test identifies the combinations of adverse conditions that cause a specified objective to fail. The failure boundary is the set of combinations for which the objective is exactly at its threshold. Combinations on the failing side of that boundary breach the objective.

A reverse stress result is conditional on its model. The definition of capital, the stress variables, recovery terms, and any effects on required capital must be stated. The result identifies a failure condition; by itself, it does not estimate how likely that condition is.

Set Up the Objective Before Solving

Let $A$ denote available capital after stress and $K$ denote required capital after stress. For a minimum solvency requirement of $A\geq K$, define the margin

$$ M=A-K. $$

Then $M\geq 0$ means the objective is met, $M=0$ is the boundary, and $M<0$ means failure. A rating objective may instead require a capital coverage ratio of at least $q$, where $q>1$. Since $K>0$, the condition $A/K\geq q$ is equivalent to $A-qK\geq0$. Using the difference $A-qK$ makes it possible to solve for a failure boundary in the same way.

The stresses can affect both sides of the objective. A loss can reduce available capital and increase required capital; a market shock can reduce asset values while also changing the modeled requirement. Reinsurance recoveries can offset some losses, but they may be limited by attachment points and coverage limits, and actual recoveries can depend on counterparty collectibility.

Worked Example: Solvency Failure With a Limited Recovery

A hypothetical insurer has available capital of $120 million and required capital of $100 million before stress. Let $x$ be a catastrophe loss in millions of dollars and $y$ a nonnegative market-shock index. A reinsurance contract pays 80% of the layer of catastrophe loss above $20 million, subject to a $50 million layer limit. Assume 90% of the contract payment is collectible. The catastrophe loss increases required capital by $0.1x$ million. The market shock reduces available capital by $0.5y$ million and increases required capital by $0.2y$ million. Find the failure boundary and assess a stress with $x=10$.

Method. Write stressed available and required capital separately, using the collectible recovery rather than the contract’s nominal payment. With $\min$ and $\max$ denoting the smaller and larger of their arguments, respectively, the collectible recovery is

R(x)=0.9(0.8)\min\{\max(x-20,0),50\} =0.72\min\{\max(x-20,0),50\}.

Available capital after stress is $120-x+R(x)-0.5y$. Required capital after stress is $100+0.1x+0.2y$. Therefore,

$$ M=120-x+R(x)-0.5y-(100+0.1x+0.2y) =20-1.1x+R(x)-0.7y. $$

For $0\leq x\leq20$, the recovery is zero, so $M=20-1.1x-0.7y$. The boundary in this range is

y=\frac{20-1.1x}{0.7},

provided the right-hand side is nonnegative. At $x=10$, the boundary is $y=(20-11)/0.7=12.8571$. At $y=12$, the margin is $20-11-8.4=0.6$ million, so the objective is met. At $y=13$, it is $20-11-9.1=-0.1$ million, so the objective is breached. Thus, with $x=10$, a market shock above the boundary causes failure.

For $20<x<70$, the collectible recovery is $0.72(x-20)$, giving $M=5.6-0.38x-0.7y$. At $x=20$, the margin is already $-2-0.7y$, and it decreases as $x$ increases in this range. For $x\geq70$, the recovery is capped at $36$ million, and the margin is $56-1.1x-0.7y$, which is also negative throughout that range. These checks show that the treaty does not restore the objective for losses at or above the $20 million attachment point under these assumptions.

Check. At $x=10,\ y=12$, available capital is $120-10-6=104$, and required capital is $100+1+2.4=103.4$. Their difference is $0.6 million, matching the margin calculation. A common error is to use the nominal 80% recovery as if it were fully collectible; the correct calculation applies the 90% collectibility factor as well.

Solving a Rating Objective

A rating objective often involves a coverage multiple, so a loss may affect both available capital and the required capital against which it is measured. Solving $A-qK=0$, rather than comparing the loss only with starting available capital, keeps both effects in the test.

Worked Example: Reverse Stressing a Capital Coverage Threshold

A hypothetical insurer has starting available capital of $150 million and starting required capital of $100 million. A rating objective is represented by a capital coverage ratio of at least $1.25$. Let $L$ be a loss in millions, $R$ a collectible recovery in millions, and $z$ a nonnegative market-shock index. After stress, available capital is $150-L+R-0.4z$, and required capital is $100+0.2L+0.1z$. Find the market-shock boundary when $L=12$ and $R=6$.

Method. The objective is $A/K\geq1.25$, or $A-1.25K\geq0$. Substituting the stated stressed amounts gives

$$ A-1.25K =150-L+R-0.4z-1.25(100+0.2L+0.1z) =25+R-1.25L-0.525z. $$

Substitute $L=12$ and $R=6$. The margin relative to the rating threshold is

$$ 25+6-1.25(12)-0.525z =16-0.525z. $$

Setting this equal to zero gives $z=16/0.525=30.4762$. The rating objective is met for $z\leq30.4762$ and breached for larger $z$, given the other inputs. For example, at $z=30$, the threshold margin is $16-15.75=0.25$ million. At $z=31$, it is $16-16.275=-0.275$ million.

Check. At the boundary, available capital is $150-12+6-0.4(30.4762)=131.8095$ million, and required capital is $100+0.2(12)+0.1(30.4762)=105.4476$ million. Multiplying the latter by $1.25$ gives $131.8095$ million, as required at the boundary. A common mistake is to solve $A=K$, which tests a 1.00 coverage ratio, not the specified 1.25 ratio.

Include Interaction and Collectibility Effects

Some stress variables affect the size of a recovery as well as capital. For instance, a market stress may worsen the financial condition of a reinsurer. When the model specifies such an interaction, a reverse stress test should preserve it rather than treating recovery as fixed. The resulting boundary may be nonlinear even when each individual effect is simple.

Worked Example: Market Stress Reduces Collectible Recovery

A hypothetical insurer starts with a $40 million solvency margin. Let $c$ be catastrophe loss, $r$ reserve deterioration, and $m$ a nonnegative market-stress unit; all losses are in millions of dollars. Before recoveries, each $1 million of catastrophe loss reduces the margin by $1.2 million, each $1 million of reserve deterioration reduces it by $1.3 million, and each market-stress unit reduces it by $1.4 million. A treaty pays 80% of catastrophe loss in excess of $10 million, subject to a $40 million layer limit. The collectible fraction is $1-0.02m$, for $0\leq m\leq50$. Find the market-stress failure boundary for $c=30$ and $r=5$.

Method. The collectible recovery is $0.8\min\{\max(c-10,0),40\}(1-0.02m)$. Since $c=30$, the loss in the covered layer is $30-10=20$ million, below the $40 million limit. The recovery is therefore $0.8(20)(1-0.02m)=16(1-0.02m)$ million. The stressed margin is

$$ M=40-1.2(30)-1.3(5)-1.4m+16(1-0.02m) =13.5-1.72m. $$

Set $M=0$ to find the boundary: $m=13.5/1.72=7.8488$. At $m=7$, the margin is $13.5-12.04=1.46$ million, so the objective is met. At $m=8$, it is $13.5-13.76=-0.26$ million, so the objective is breached. The recovery changes with $m$: at the boundary, the collectible fraction is $1-0.02(7.8488)=0.8430$, and the recovery is approximately $16(0.8430)=13.49$ million.

Check. Substituting the unrounded boundary into $13.5-1.72m$ gives zero. Ignoring the collectibility interaction would instead give $M=13.5-1.4m$ and a boundary of $9.6429$, overstating the market stress the insurer can withstand.

Interpret the Result, Not Just the Equation

A reverse stress boundary is most useful when it is translated back into the decision context. Report which stresses are held fixed, which combinations breach the objective, and what features of the proposed action matter to that result. For a reinsurance decision, these features may include attachment, limit, recovery percentage, or collectibility. For a capital decision, they may include the amount of capital raised and the market risks affecting its value.

The boundary is conditional on the chosen model and stress units. A threshold such as $y=12.8571$ is not a probability, nor does it say that failure is likely. It says that, under the assumptions specified, that combination reaches the objective’s boundary. To assess plausibility, decision-makers need additional evidence, such as historical experience, expert judgment, or a separately specified probability model.

State the objective.
Write the solvency margin or rating-threshold margin, and define precisely when it is breached.

Model stressed amounts.
Separate effects on available capital from effects on required capital; specify recovery limits and collectibility.

Solve the boundary.
Set the objective margin equal to zero, respect any piecewise limits, and identify which side represents failure.

Check and interpret.
Substitute boundary and nearby passing and failing values, then explain what the result does—and does not—establish.

A frequent exam trap is to report a single loss amount without stating the other stresses held fixed. With multiple stress variables, a boundary is generally a set of combinations, not one universal threshold. Another trap is to use a recovery above its limit or assume full collection when the problem specifies a counterparty adjustment. These errors can materially change the decision interpretation.

Key takeaway: Reverse stress testing turns a capital objective into a boundary of failure combinations. The boundary is meaningful only when the capital measure, stress effects, recovery assumptions, and breach direction are explicit.

Check Your Understanding

For each problem, identify the boundary using the stated model, then distinguish passing from failing values.

A hypothetical firm has a stressed solvency margin $M=18-0.9x-0.6y$, where $x$ is a loss in millions and $y$ is a nonnegative market-stress index. At $x=10$, find the boundary for $y$. (Answer: $y=15$; larger values fail.)
A rating-threshold margin is $25+R-1.25L-0.525z$, in millions. If $L=16$ and collectible recovery $R=8$, find the boundary for nonnegative $z$. (Answer: $z=24.7619$; larger values fail.)
A solvency margin is $M=15-1.1x+0.6\min\{\max(x-10,0),20\}-0.5y$, in millions. For $x=14$, find the market-stress boundary. (Answer: $y=4$; larger values fail.)
A stress model gives $M=32-0.4c-1.3r-[1.4+0.016(c-10)]m$, valid for $10\leq c\leq50$. For $c=30$ and $r=5$, find the boundary for $m$. (Answer: $m=7.8488$; larger values fail.)

What You'll Learn