Why Diversification Must Be Recalculated
A risk transfer changes more than the standalone capital for the risk being transferred. It can also change how that risk combines with the rest of the portfolio. Consequently, a capital analysis must compare the portfolio before and after the transaction on a consistent basis. Adding a separately estimated diversification benefit to an aggregate capital-relief estimate can count the same reduction twice.
The calculations below use a specific model to make the comparison transparent: mean-zero, jointly normally distributed unexpected losses; a fixed confidence level; and capital measured by value at risk (VaR). A proportional cession scales the ceded line deterministically, and initially we assume no counterparty default, basis risk, or other new exposure. These assumptions are not universal. If a transaction introduces additional risks, those risks must be included in the post-transfer portfolio model.
For a mean-zero normal loss with standard deviation \(\sigma\), the \(q\)-VaR is \(z_q\sigma\), where \(z_q\) is the \(q\)-quantile of the standard normal distribution. For jointly normal risks, the aggregate loss is also normal, and its variance includes all variances and twice each pairwise covariance. The normal model is a simplifying assumption; the relation \(K_q=z_q\sigma\) does not generally apply to nonnormal losses.
A Consistent Capital Bridge
The key accounting identity follows directly from the definition of diversification benefit. Write \(S_{\mathrm{before}}\) and \(S_{\mathrm{after}}\) for the sums of standalone capitals before and after transfer, and \(B_{\mathrm{before}}\) and \(B_{\mathrm{after}}\) for the corresponding diversification benefits. Then portfolio capital is standalone capital less diversification benefit at each point in time.
Subtracting the second equation from the first gives the capital-relief bridge. This is an identity, not an additional source of benefit.
Thus, if a transfer reduces the sum of standalone capitals by 30 but the diversification benefit also falls by 12, aggregate capital falls by 18. The reduction in diversification benefit offsets part of the standalone reduction: a smaller portfolio has less risk available to diversify. If the benefit instead increases, that increase contributes additional aggregate relief. Calculate the two benefits afresh; do not assume either one stays fixed.
Worked Example: Independent Risks and a Proportional Cession
A hypothetical insurer models two independent, mean-zero normal loss components, measured in thousands of dollars. Their standard deviations are 30 and 40. It cedes 50% of the first component and retains the second unchanged. At \(q=97.725\%\), use \(z_q=2\). Find the portfolio capital before and after transfer, aggregate relief, and the diversification-benefit bridge.
Method. For independent components, aggregate variance is the sum of their variances. A 50% cession leaves half the first component, so its retained standard deviation is \(0.50(30)=15\).
Before transfer, the aggregate variance is \(30^2+40^2=900+1{,}600=2{,}500\), and the aggregate standard deviation is \(\sqrt{2{,}500}=50\). Therefore, \(K_{\mathrm{before}}=2(50)=100\) thousand dollars.
After transfer, the aggregate variance is \(15^2+40^2=225+1{,}600=1{,}825\). Its standard deviation is \(\sqrt{1{,}825}=42.7200\) thousand dollars, so \(K_{\mathrm{after}}=2(42.7200)=85.4400\) thousand dollars. Aggregate relief is \(100-85.4400=14.5600\) thousand dollars.
The standalone capital sum falls from \(2(30+40)=140\) to \(2(15+40)=110\) thousand dollars: standalone relief is 30. Before transfer, diversification benefit is \(140-100=40\); after transfer, it is \(110-85.4400=24.5600\). The benefit falls by \(40-24.5600=15.4400\). The bridge checks: \(30-15.4400=14.5600\) thousand dollars, matching the direct portfolio calculation.
Check. The post-transfer standalone sum less the post-transfer benefit is \(110-24.5600=85.4400\), exactly the post-transfer capital. The common mistake this avoids is treating the original benefit of 40 as unchanged after the first risk has been reduced.
Why Covariance Changes Under a Cession
A proportional cession changes the variance of the retained component and its covariance with every other component. If \(X_1\) is reduced to \(cX_1\), where \(c\) is the retained proportion, then variance and covariance obey the following relationships.
The variance identity follows from \(\operatorname{Var}(cX_1)=E[(cX_1-E[cX_1])^2]=c^2E[(X_1-E[X_1])^2]\). For covariance, expanding its definition gives \(\operatorname{Cov}(cX_1,X_j)=E[(cX_1-E[cX_1])(X_j-E[X_j])]=c\operatorname{Cov}(X_1,X_j)\). The covariance scales by \(c\), not \(c^2\), because only one of its two variables is scaled. If both variables are scaled, covariance scales by the product of their scaling factors.
For two mean-zero jointly normal losses \(X_1,X_2\), portfolio capital is therefore calculated from their combined variance, not by adding their standalone VaRs. With standard deviations \(\sigma_1,\sigma_2\) and correlation \(\rho\), that variance is \(\sigma_1^2+\sigma_2^2+2\rho\sigma_1\sigma_2\). After a proportional cession, use the retained standard deviation and the revised covariance.
Worked Example: Correlated Risks and Corrected Aggregate Relief
Two hypothetical mean-zero jointly normal loss components, in thousands of dollars, have standard deviations 60 and 50 and correlation \(0.20\). A proportional cession leaves 50% of the first component and does not change the second. At \(q=99\%\), use \(z_q=2.3263\). Calculate pre-transfer and post-transfer portfolio capital, direct aggregate relief, and the diversification-benefit bridge. Report capital in thousands of dollars to four decimals.
Method. Calculate aggregate variance from both variances and the covariance term before and after transfer. The original covariance is \(0.20(60)(50)=600\). The retained first component has standard deviation \(0.50(60)=30\), so its covariance with the second component becomes \(0.20(30)(50)=300\).
Before transfer, variance is \(60^2+50^2+2(600)=3{,}600+2{,}500+1{,}200=7{,}300\). The standard deviation is \(\sqrt{7{,}300}=85.4400375\), so capital is \(2.3263(85.4400375)=198.7592\) thousand dollars.
After transfer, variance is \(30^2+50^2+2(300)=900+2{,}500+600=4{,}000\). The standard deviation is \(\sqrt{4{,}000}=63.2455532\), so capital is \(2.3263(63.2455532)=147.1281\) thousand dollars. Direct aggregate relief is \(198.7592-147.1281=51.6310\) thousand dollars, calculated from unrounded values.
The standalone capital sum falls from \(2.3263(60+50)=255.8930\) to \(2.3263(30+50)=186.1040\) thousand dollars. Thus standalone relief is \(69.7890\). The initial diversification benefit is \(255.8930-198.7592=57.1338\); the post-transfer benefit is \(186.1040-147.1281=38.9759\). Its decrease is \(18.1580\), using unrounded values. The bridge gives \(69.7890-18.1580=51.6310\) thousand dollars, consistent with the direct calculation.
Check. The post-transfer variance uses covariance 300, not 600: halving the first risk halves its covariance with the unchanged second risk. The direct capital figures and the standalone-capital bridge agree to the displayed precision.
Common Double-Counting Trap
A common error is to calculate the aggregate portfolio capital relief and then add the standalone relief for the ceded component as though it were a separate portfolio benefit. In the preceding example, this incorrect total would be \(51.6310+69.7890=121.4200\) thousand dollars. Both quantities describe the effect of reducing the same exposure; they are alternative views of the same transaction, not additive benefits.
A valid analysis uses either the direct difference between pre-transfer and post-transfer aggregate capital, or the bridge that starts with standalone relief and subtracts the decrease in diversification benefit. Do not add both results. Also keep other transaction effects separate and explicit: premium, expenses, counterparty risk, and basis risk may affect the decision, but none should be silently inserted into the Gaussian variance calculation above.
| Calculation | Interpretation | Use in the comparison |
|---|---|---|
| Pre-transfer aggregate capital minus post-transfer aggregate capital | Direct portfolio capital relief | Valid total relief measure |
| Standalone relief minus the decrease in diversification benefit | Reconciliation of the same portfolio relief | Valid bridge; should match the direct calculation |
| Aggregate relief plus standalone relief | Counts the same risk reduction through two measures | Invalid as total relief |
Three-Risk Reconciliation
A transfer can also change more than one covariance term. The following example shows how to apply the same method to a portfolio with three components. It is still essential to recalculate the entire post-transfer aggregate, rather than adjusting only the ceded component’s standalone capital.
Worked Example: Recalculating a Three-Risk Portfolio
A hypothetical portfolio has three mean-zero jointly normal loss components, all measured in thousands of dollars. Their standard deviations are 40, 30, and 20. The correlation between the first and second components is \(0.25\); the other two correlations are zero. A proportional cession leaves 50% of the first component. At a confidence level with \(z_q=2\), find aggregate capital relief and reconcile it to standalone relief.
Method. Include the first-second covariance in each aggregate variance. Before transfer, that covariance is \(0.25(40)(30)=300\). The first component’s retained standard deviation is 20, so after transfer its covariance with the second is \(0.25(20)(30)=150\). The other covariances remain zero.
Before transfer, aggregate variance is \(40^2+30^2+20^2+2(300)=1{,}600+900+400+600=3{,}500\). Thus aggregate standard deviation is \(\sqrt{3{,}500}=59.1608\), and capital is \(2(59.1608)=118.3216\) thousand dollars.
After transfer, aggregate variance is \(20^2+30^2+20^2+2(150)=400+900+400+300=2{,}000\). Thus aggregate standard deviation is \(\sqrt{2{,}000}=44.7214\), and capital is \(2(44.7214)=89.4427\) thousand dollars. Aggregate relief is \(118.3216-89.4427=28.8789\) thousand dollars.
The standalone capital sum falls from \(2(40+30+20)=180\) to \(2(20+30+20)=140\) thousand dollars, a reduction of 40. Diversification benefit falls from \(180-118.3216=61.6784\) to \(140-89.4427=50.5573\), a decrease of 11.1211. Therefore the bridge gives \(40-11.1211=28.8789\) thousand dollars of aggregate relief. The calculation checks against the direct portfolio difference.
Common mistake avoided. Keeping the original covariance of 300 after the first component is halved would overstate post-transfer variance. The covariance must be recalculated as 150, as follows from scaling only one variable.
Practical Checks Before Reporting Relief
Before presenting a capital comparison, verify that the two portfolio calculations use the same capital definition, confidence level, currency basis, and treatment of expected loss. In this tutorial, risks are centered unexpected losses and capital is normal VaR. If a model instead includes expected losses, premiums, or transaction costs, identify those items consistently rather than mixing them into one side of the comparison.
Also verify the post-transfer risk representation. A simple proportional cession scales the ceded component, but an actual contract may have limits, attachment points, exclusions, or counterparty exposure. Those features may make the retained loss nonlinear or introduce additional risk. In that case, model the post-transfer loss directly and recompute portfolio capital under the chosen model; do not apply the proportional-scaling formulas outside their assumptions.
A defensible conclusion reports direct aggregate relief and, when useful, its reconciliation into standalone relief and the change in diversification benefit. Agreement between those calculations is a useful control against double counting and covariance errors. The comparison is aligned with the CAS Exam 9 (Risk Management for Actuaries) syllabus.
Check Your Understanding
For each problem, use the stated model and check whether standalone and aggregate capital are being compared consistently.
- Two independent mean-zero normal loss components have standard deviations 25 and 60, in thousands of dollars. A proportional cession leaves 60% of the first component. At a confidence level with \(z_q=2\), calculate aggregate capital before and after transfer and the relief. (Answer: before \(=130.0000\), after \(=123.6932\), relief \(=6.3068\) thousand dollars.)
- Two mean-zero jointly normal risks have standard deviations 45 and 35, with correlation \(0.40\). A cession leaves two-thirds of the first risk; use \(z_q=2\). Recalculate the covariance and find aggregate relief. (Answer: pre-transfer variance \(=4{,}510\), post-transfer variance \(=2{,}965\), pre-transfer capital \(=134.3130\), post-transfer capital \(=108.9036\), relief \(=25.4094\) thousand dollars.)
- Three risks have standard deviations 40, 30, and 20, with only the first and second correlated, at \(0.25\). A cession leaves half of the first risk. At \(z_q=2\), what is the decrease in standalone capital, the decrease in diversification benefit, and the aggregate relief? (Answer: standalone relief \(=40.0000\), diversification-benefit decrease \(=11.1211\), aggregate relief \(=28.8789\) thousand dollars.)
- A portfolio’s standalone capital sum falls from 420 to 340, while aggregate capital falls from 300 to 250, all in thousands of dollars. Calculate the diversification benefit before and after, and verify aggregate relief using the bridge. (Answer: benefits \(=120\) before and \(90\) after; standalone relief \(=80\), benefit decrease \(=30\), aggregate relief \(=50\) thousand dollars.)
- An analyst reports aggregate capital relief of 45 thousand dollars and standalone capital relief of 70 thousand dollars for the same transfer. The analyst adds them and reports total relief of 115. Identify the error and state what additional bridge quantity would reconcile the two relief measures. (Answer: adding the two measures double counts; the diversification-benefit decrease is \(70-45=25\) thousand dollars, and aggregate relief remains 45 thousand dollars.)