Why Measure Hedge Effectiveness?
A catastrophe bond transfers some catastrophe risk away from its sponsor, but the amount transferred depends on the bond’s trigger and the event outcome. A bond can provide substantial expected recovery without reducing the variability of the sponsor’s retained losses by the same proportion. Measuring hedge effectiveness therefore requires comparing loss distributions, not just comparing expected payments.
The analysis below treats each scenario as an outcome over the same period, such as one year. Losses and recoveries are measured in the same monetary units and are assumed to be settled over that period without a timing adjustment. The scenarios and probabilities are hypothetical. The focus is the bond’s effect on sponsor losses, not investor returns or transaction structure.
This scenario-based variance is a probability-weighted measure, not a sample variance with an \(n-1\) denominator. When a recovery is larger than the gross loss in a scenario, \(L\) is negative; it represents the net financial result after the hedge, rather than a physical insured loss. The examples below have no such outcome.
Two Measures of Risk Reduction
A bond’s variance reduction compares the variance of retained loss with gross-loss variance. Its volatility reduction compares the corresponding standard deviations. Both are useful, but they answer different questions: variance is in squared monetary units, while standard deviation is in the original monetary units.
A positive value means the measure is lower after the hedge; zero means no change; and a negative value means the measure is higher. These ratios require \(\operatorname{Var}(G)>0\). Variance reduction is not volatility reduction: the latter uses standard deviations, so it does not generally equal the former.
To see why, subtract the means from \(L=G-R\), square, and take expectations:
The identity shows why recovery size alone is not enough. Recoveries that tend to be high when gross losses are high have positive covariance with gross loss, which tends to reduce retained-loss variance. A recovery pattern that is poorly aligned with gross loss can provide a smaller reduction—or even increase variance.
Worked Example: Calculating Volatility Reduction
A hypothetical sponsor models four equally likely annual catastrophe scenarios. Gross losses, bond recoveries, and retained losses are shown in thousands of dollars. Calculate the variance and volatility reductions.
| Scenario | Probability | Gross loss \(G\) | Recovery \(R\) | Retained loss \(L=G-R\) |
|---|---|---|---|---|
| 1 | 0.25 | 0 | 0 | 0 |
| 2 | 0.25 | 4 | 0 | 4 |
| 3 | 0.25 | 10 | 14/3 | 16/3 |
| 4 | 0.25 | 20 | 10 | 10 |
Method. Calculate each scenario variable’s probability-weighted mean and second moment, then subtract the squared mean from the second moment. This uses the full scenario distribution rather than treating the average recovery as a measure of volatility.
The gross-loss mean is \((0+4+10+20)/4=8.5\), and its second moment is \((0^2+4^2+10^2+20^2)/4=129\). Thus,
The retained-loss mean is \((0+4+16/3+10)/4=29/6\). Its second moment is \((0^2+4^2+(16/3)^2+10^2)/4=325/9\). Therefore,
The variance reduction is \(1-12.75/56.75=176/227\), or 77.53%. For volatility reduction, use the standard deviations: \(1-\sqrt{12.75}/\sqrt{56.75}=1-\sqrt{51/227}\). Rounded to two decimal places, the volatility reduction is 52.60%. Using the displayed rounded standard deviations gives \(1-3.571/7.533\approx 0.5260\), also 52.60% to two decimal places.
Check. The retained-loss variance, 12.75, is below the gross-loss variance, 56.75, so both reductions are positive. The volatility reduction is smaller than the variance reduction, as expected when the standard-deviation ratio is not confused with the variance ratio.
Residual Basis Risk and Net Results
A bond’s recovery may not match the sponsor’s loss pattern exactly. To quantify this mismatch, define a reference recovery \(R^*\): the recovery the sponsor would receive under a specified idealized or target schedule. The scenario-level basis difference is \(\Delta=R-R^*\). A positive value means the bond pays more than the reference in that scenario; a negative value means it pays less.
A simple summary is the probability-weighted mean absolute basis difference, \(E[|\Delta|]\). It treats overpayment and underpayment as mismatches without allowing them to cancel. The signed mean \(E[\Delta]\) can also be informative about average excess or shortfall, but it can conceal offsetting scenario errors. Neither summary replaces the retained-loss variance calculation: they describe different aspects of the hedge.
Worked Example: Separating Basis Mismatch from Volatility Reduction
A hypothetical sponsor compares actual bond recoveries with a target indemnity-style schedule that attaches at 5 thousand dollars and pays at most 10 thousand dollars. For gross loss \(G\), the reference recovery is \(R^*=\min(\max(G-5,0),10)\). Four annual scenarios are equally likely, and all amounts are in thousands of dollars. Find the actual retained-loss volatility reduction and summarize the basis mismatch.
| Scenario | \(G\) | Reference \(R^*\) | Actual \(R\) | \(L=G-R\) | \(\Delta=R-R^*\) |
|---|---|---|---|---|---|
| 1 | 0 | 0 | 0 | 0 | 0 |
| 2 | 5 | 0 | 2 | 3 | 2 |
| 3 | 12 | 7 | 6 | 6 | -1 |
| 4 | 20 | 10 | 8 | 12 | -2 |
Basis calculation. The absolute differences are \(0,2,1,2\), so \(E[|\Delta|]=(0+2+1+2)/4=1.25\) thousand dollars. The signed mean is \(E[\Delta]=(0+2-1-2)/4=-0.25\) thousand dollars. The negative signed average indicates an average shortfall relative to the reference, but it is smaller in magnitude than the mean absolute mismatch because positive and negative differences offset.
Volatility calculation. Gross loss has mean \((0+5+12+20)/4=9.25\) and second moment \((0+25+144+400)/4=142.25\), so \(\operatorname{Var}(G)=142.25-(9.25)^2=56.6875\). Retained loss has mean \((0+3+6+12)/4=5.25\) and second moment \((0+9+36+144)/4=47.25\), so \(\operatorname{Var}(L)=47.25-(5.25)^2=19.6875\).
Check. The actual bond lowers standard deviation because retained-loss variance is below gross-loss variance. The scenario mismatches also show why a favorable volatility result does not mean every outcome follows the reference schedule: there is an overpayment in scenario 2 and underpayments in scenarios 3 and 4.
Compare Scenario Patterns, Not Only Expected Recovery
Two bonds can have different expected recoveries and different effects on volatility. The weighted scenario probabilities matter: a large payment in an infrequent high-loss scenario can affect risk differently from a payment concentrated in a more common moderate-loss scenario. A comparison should therefore use consistent gross-loss scenarios, probabilities, and measurement periods for every candidate hedge.
Worked Example: Comparing Two Hypothetical Hedge Patterns
A sponsor considers two hypothetical bonds under the same three annual catastrophe scenarios. Scenario probabilities are 0.50, 0.30, and 0.20. Gross losses are 0, 10, and 30 thousand dollars. Bond A pays 0, 8, and 15 thousand dollars; Bond B pays 0, 0, and 20 thousand dollars. Compare expected retained losses and volatility reductions.
Gross portfolio measure. The expected gross loss is \(0.50(0)+0.30(10)+0.20(30)=9\) thousand dollars. The second moment is \(0.50(0^2)+0.30(10^2)+0.20(30^2)=210\), giving \(\operatorname{Var}(G)=210-9^2=129\) and standard deviation \(\sqrt{129}\approx11.358\) thousand dollars.
Bond A. Retained losses are 0, 2, and 15. Their mean is \(0.50(0)+0.30(2)+0.20(15)=3.6\) thousand dollars. Their second moment is \(0.50(0^2)+0.30(2^2)+0.20(15^2)=46.2\), so \(\operatorname{Var}(L_A)=46.2-(3.6)^2=33.24\). The standard deviation is \(\sqrt{33.24}\approx5.765\) thousand dollars, and the volatility reduction is \(1-\sqrt{33.24/129}\approx49.24\%\).
Bond B. Retained losses are 0, 10, and 10. Their mean is \(0.50(0)+0.30(10)+0.20(10)=5\) thousand dollars. Their second moment is \(0.50(0^2)+0.30(10^2)+0.20(10^2)=50\), so \(\operatorname{Var}(L_B)=50-5^2=25\). Its standard deviation is 5 thousand dollars, and its volatility reduction is \(1-5/\sqrt{129}\approx55.98\%\).
Interpretation and check. Bond A produces the lower expected retained loss, 3.6 rather than 5 thousand dollars, while Bond B produces the lower retained-loss standard deviation, 5 rather than about 5.765 thousand dollars. Both variances are below the gross-loss variance of 129. This comparison does not make either bond universally preferable; it demonstrates that mean loss and volatility are distinct evaluation criteria.
Common Exam Traps and Practical Interpretation
A calculation is meaningful only if gross losses and recoveries are paired on the same scenarios and assigned the same probabilities. Applying one scenario distribution to gross loss and a different one to recovery breaks the link that determines retained-loss volatility. If scenario probabilities are estimated from a model or data, the result also depends on those assumptions; it is not a guarantee of future performance.
- Using the wrong reduction formula. Calculating \(1-\operatorname{Var}(L)/\operatorname{Var}(G)\) answers the variance-reduction question. For volatility reduction, calculate \(1-\operatorname{SD}(L)/\operatorname{SD}(G)\).
- Subtracting expected recovery from a risk measure. In general, the standard deviation of \(G-R\) is not the standard deviation of \(G\) minus the expected recovery. Calculate retained loss scenario by scenario, or use the variance identity with the covariance term.
- Ignoring probability weights. When outcomes are not equally likely, use the stated probability for each scenario in both the mean and second moment.
- Letting signed basis differences cancel. A near-zero \(E[\Delta]\) can coexist with large scenario mismatches. Report a mismatch measure such as \(E[|\Delta|]\) when the size of deviations is the concern.
- Mixing units or horizons. Do not compare an annual gross-loss standard deviation with a multi-year recovery measure or mix dollars with thousands of dollars.
Hedge effectiveness should be interpreted as a portfolio-level result under a stated scenario set and recovery rule. Expected loss, retained-loss volatility, and residual basis mismatch are complementary descriptions; none alone fully characterizes the hedge. These calculations are aligned with the CAS Exam 9 (Risk Management for Actuaries) syllabus.
Check Your Understanding
For each problem, use the stated scenario probabilities and consistent units. Treat all amounts as thousands of dollars unless stated otherwise.
- A hypothetical bond has two equally likely scenarios: gross losses are 4 and 12, and recoveries are 0 and 4. Find retained-loss variance and volatility reduction. (Answer: retained-loss variance \(16\); volatility reduction \(50.00\%\).)
- Three equally likely scenarios have gross losses 0, 8, and 16, with recoveries 0, 4, and 8. Calculate gross-loss and retained-loss variances, then find variance reduction and volatility reduction. (Answer: gross variance \(128/3\); retained variance \(32/3\); variance reduction \(75.00\%\); volatility reduction \(50.00\%\).)
- A reference recovery is defined by \(R^*=\min(\max(G-4,0),8)\). In three equally likely scenarios, \(G\) is 0, 10, and 20, and actual recoveries are 0, 3, and 6. Find the reference recoveries and the mean absolute basis difference. (Answer: reference recoveries \(0,6,8\); mean absolute basis difference \(5/3\), or \(1.667\).)
- A portfolio has gross losses 0, 10, and 20 with probabilities 0.50, 0.30, and 0.20. A bond pays 0, 5, and 10. Find the mean retained loss, retained-loss variance, and volatility reduction. (Answer: mean retained loss \(5.5\); retained-loss variance \(44.25\); volatility reduction approximately \(50.25\%\).)
- In two equally likely scenarios, gross losses are 0 and 10. Bond X pays 0 and 8; Bond Y pays 0 and 5. Compare their expected recoveries and volatility reductions. (Answer: Bond X expected recovery \(4\), volatility reduction \(60.00\%\); Bond Y expected recovery \(2.5\), volatility reduction \(50.00\%\).)