Investment Decisions and Rating Volatility

Investment Choices and Rating Resilience

Investment choices affect more than expected yield. Two portfolios with similar expected income can have quite different volatility, market sensitivity, and potential capital impact. If an adverse movement reduces available capital, a company with little headroom above a rating objective may be more exposed to pressure on that objective.

The calculations below use hypothetical inputs to show the mechanics. A rating agency’s capital assessment may use prescribed stress factors, asset classifications, concentrations, and other adjustments; it should not be treated as a normal-distribution volatility calculation. The simplified measures here help compare choices, but do not reproduce any agency’s model. The framework is aligned with the CAS Exam 9 (Risk Management for Actuaries) syllabus.

Definition: Let $R_i$ be the one-period return on asset $i$, $w_i$ its portfolio weight, $\mu_i=E[R_i]$ its expected return, and $\sigma_i$ its standard deviation. For fixed weights that sum to 1, the portfolio return is $R_p=\sum_i w_iR_i$. Its expected return is $\mu_p=\sum_i w_i\mu_i$, while its variance depends on both asset variances and covariances.

Expected return describes the average result under the assumed model; standard deviation describes the spread around that average. Neither is a complete measure of investment risk. In particular, a portfolio’s volatility does not by itself describe the size of a tail loss, illiquidity, credit deterioration, or the sensitivity of liabilities to the same market movement.

Portfolio Volatility and Dependence

For two assets with fixed weights $w_1,w_2$, standard deviations $\sigma_1,\sigma_2$, and correlation $\rho_{12}$, the portfolio variance is

\operatorname{Var}(R_p)=w_1^2\sigma_1^2+w_2^2\sigma_2^2+2w_1w_2\rho_{12}\sigma_1\sigma_2.

This formula assumes the weights and return distributions are defined over the same time period and that the stated variances and correlation apply over that period. Correlation must be between $-1$ and $1$. For several assets, each distinct pair contributes a covariance term; correlations cannot be chosen independently if the resulting covariance matrix is not mathematically valid.

Derivation: Write each return as its mean plus a deviation: $R_i=\mu_i+(R_i-\mu_i)$. Then $R_p-\mu_p=\sum_iw_i(R_i-\mu_i)$. Squaring and taking expectations gives $\operatorname{Var}(R_p)=\sum_iw_i^2\sigma_i^2+2\sum_{i<j}w_iw_j\operatorname{Cov}(R_i,R_j)$. Since $\operatorname{Cov}(R_i,R_j)=\rho_{ij}\sigma_i\sigma_j$, the two-asset formula follows. The cross term is why adding asset standard deviations directly is generally incorrect.

Worked Example: Comparing Expected Income and Portfolio Volatility

A hypothetical insurer invests $100 million: 60% in bonds with an annual expected return of 4.5% and standard deviation of 6%, and 40% in equities with an annual expected return of 8% and standard deviation of 18%. Assume the annual return correlation is 0.20. Find expected investment income and the portfolio return standard deviation. Then calculate an illustrative 99.5% normal-model adverse deviation from expected income, using $z=2.576$. Assume returns are adequately represented by this model for the illustration; this is not an agency capital formula.

Method and setup. Use the weighted mean for expected return and the covariance formula for variance. The common mistake this method avoids is taking a weighted average of asset standard deviations as if that were the portfolio standard deviation.

Expected return. The bond contribution is $0.60(0.045)=0.027$, and the equity contribution is $0.40(0.08)=0.032$. Thus $\mu_p=0.027+0.032=0.059$, or 5.9%. Expected income is $100(0.059)=\$5.90$ million.

Variance and standard deviation. The bond variance contribution is $0.60^2(0.06^2)=0.36(0.0036)=0.001296$. The equity contribution is $0.40^2(0.18^2)=0.16(0.0324)=0.005184$. The covariance contribution is $2(0.60)(0.40)(0.20)(0.06)(0.18)=0.0010368$. Therefore the total variance is $0.001296+0.005184+0.0010368=0.0075168$, and the standard deviation is $\sqrt{0.0075168}=0.0866995$, or 8.6700%.

Illustrative adverse deviation. The dollar standard deviation is $100(0.0866995)=\$8.66995$ million. Multiplying by 2.576 gives $2.576(8.66995)=\$22.3338$ million. This is the modeled deviation below expected income at the stated normal quantile; it is not an absolute loss forecast. As a check, the calculated portfolio standard deviation is below the 18% equity standard deviation and above the 6% bond standard deviation, which is plausible for these positive weights and positive correlation.

For nonnegative weights, the weighted average of individual standard deviations is an upper bound on the portfolio standard deviation. It can be exact when the returns move perfectly together in the same direction, but it may overstate risk when dependence is less than perfect. Diversification does not mean that risk disappears: it means that covariance affects the total.

Worked Example: Testing a Rating Capital Threshold

A hypothetical insurer has $140 million of available capital and $100 million of required capital. Its management objective is to keep the available-to-required capital ratio at or above 1.25. It considers two mutually exclusive $100 million investment choices. Choice C has a 4% annual return standard deviation; Choice D has a 9% standard deviation. For this illustration only, assume an adverse capital stress reduces available capital by $2.576$ times the dollar standard deviation, while required capital remains fixed. Ignore expected return so the comparison isolates volatility.

Method and setup. Calculate each dollar standard deviation, apply the stated stress, and compare the resulting ratio with 1.25. The common mistake avoided is comparing volatility percentages without translating them into the amount of capital at risk.

For Choice C, the dollar standard deviation is $100(0.04)=\$4$ million. Its stress reduction is $2.576(4)=\$10.304$ million. Stressed available capital is $140-10.304=\$129.696$ million, so the ratio is $129.696/100=1.29696$, above the objective.

For Choice D, the dollar standard deviation is $100(0.09)=\$9$ million. Its stress reduction is $2.576(9)=\$23.184$ million. Stressed available capital is $140-23.184=\$116.816$ million, giving a ratio of $116.816/100=1.16816$, below the objective.

The threshold requires at least $1.25(100)=\$125$ million of available capital. Choice C leaves $129.696-125=\$4.696$ million of stressed headroom; Choice D is short by $125-116.816=\$8.184$ million. The arithmetic is internally consistent: the same initial capital and requirement lead to different threshold outcomes because the assumed volatility stress differs. Actual rating capital treatment can differ materially from this stylized calculation.

Market Movements and Asset Risk

Volatility is one lens on market risk. A specific market shock can also be evaluated directly using exposures and sensitivities. For a bond portfolio, a common first-order approximation for a small parallel yield change is

\frac{\Delta P}{P}\approx -D_{\mathrm{mod}}\Delta y,

where $D_{\mathrm{mod}}$ is modified duration and $\Delta y$ is the yield change expressed as a decimal. The approximation assumes a small, parallel change and ignores convexity, changes in credit spreads, and other effects. It estimates a change in economic market value; whether and when that movement appears in reported earnings depends on accounting classification and related rules.

Worked Example: Stressing a Mixed Investment Portfolio

A hypothetical insurer holds $60 million of bonds with modified duration 4, $20 million of equities, and $20 million of cash. Consider a parallel increase in yields of 0.75 percentage points and, at the same time, an equity-market decline of 12%. Assume the duration approximation applies, cash has no value change in this scenario, and there are no taxes, transactions, or offsetting liability movements. Estimate the total portfolio value change. Then consider reallocating $10 million from equities to cash.

Method and setup. Apply the duration approximation to the bond value and the stated percentage shock to equity value, then sum dollar changes. The common mistake avoided is applying the equity percentage loss to the entire portfolio rather than to the equity holding.

The yield change is $\Delta y=0.0075$. The bond percentage change is approximately $-4(0.0075)=-0.03$, or $-3\%$. The bond dollar change is $-0.03(60)=-\$1.80$ million. The equity change is $-0.12(20)=-\$2.40$ million, and cash changes by $0. The total estimated change is $-1.80-2.40+0=-\$4.20$ million.

After moving $10 million of equities to cash, equities are $10 million and cash is $30 million; the bond holding is unchanged. The bond change remains $-\$1.80$ million. The equity change is now $-0.12(10)=-\$1.20$ million, while cash remains unchanged. The total is $-1.80-1.20=-\$3.00$ million. The reallocation reduces this particular scenario loss by $4.20-3.00=\$1.20$ million. It does not establish that the revised portfolio is always safer: other market movements, expected returns, liquidity needs, and liability sensitivities may change the comparison.

Interpreting Investment Risk in a Rating Decision

A rating-oriented investment decision should connect the portfolio change to both the insurer’s financial position and the way the relevant capital assessment treats assets. A higher expected yield can be attractive, but the additional return is not a guaranteed offset to a stressed loss. Likewise, a reduction in market-value volatility may not improve every capital measure if it raises concentration, credit, or reinvestment risk.

The portfolio formula also gives a useful diagnostic for proposed changes. If an asset is added, its effect depends not only on its own standard deviation but also on its covariance with existing holdings. Under the stated fixed-weight model, an asset with a lower correlation to the current portfolio can reduce aggregate volatility, even if its individual volatility is not low. This result is conditional on the assumed parameters remaining relevant; correlations can change in stressed markets.

Decision question	Useful calculation	Important limitation
How variable is modeled annual return?	Portfolio variance using weights and covariances	Depends on estimated parameters and return model
How much capital could a specified stress consume?	Apply the defined stress to available capital	Illustrative stress may not match a rating model
How could a market shock move asset values?	Apply asset-specific sensitivities, such as duration	Approximation may omit nonlinear effects and liability offsets
Will a rating objective still be met?	Recalculate the relevant capital measure and threshold margin	Qualitative factors and model adjustments may also matter

Common Mistake: Averaging Standard Deviations

Suppose 70% of a portfolio is in an asset with standard deviation 4%, and 30% is in an asset with standard deviation 10%; assume zero correlation. A tempting but incorrect calculation is $0.70(4\%)+0.30(10\%)=5.8\%$. That calculation is not the portfolio standard deviation.

The correct variance is $0.70^2(0.04^2)+0.30^2(0.10^2)+2(0.70)(0.30)(0)(0.04)(0.10)=0.000784+0.000900=0.001684$. The standard deviation is $\sqrt{0.001684}=0.0410366$, or 4.1037%. The 5.8% weighted average is an upper bound for nonnegative weights, not a generally accurate estimate; in this case it overstates the portfolio standard deviation.

Key takeaway: Compare investment choices using a consistent horizon and explicit assumptions. Calculate portfolio volatility from covariances, translate stresses into capital and threshold effects where appropriate, and distinguish economic value movements from reported earnings. A rating conclusion requires the relevant capital framework and cannot be inferred from yield or standard deviation alone.

Check Your Understanding

Work each problem using the stated assumptions; answers are provided for checking.

Two assets each have a 50% portfolio weight. Their annual standard deviations are 6% and 12%, and their correlation is $-0.25$. Calculate the portfolio standard deviation. (Answer: Variance is 0.0036 and standard deviation is 6.0000%.)
A hypothetical $50 million portfolio invests 80% in an asset with expected return 2% and 20% in one with expected return 9%. Find expected annual investment income and portfolio return. (Answer: Expected income is $1.70 million and expected return is 3.4000%.)
A $80 million portfolio has an annual return standard deviation of 5%. Under an explicitly illustrative normal stress with $z=2.576$, calculate the adverse deviation from expected return in dollars. (Answer: Dollar standard deviation is $4.00 million and the adverse deviation is $10.304 million.)
A portfolio holds 70% in an asset with standard deviation 4% and 30% in an asset with standard deviation 10%; correlation is zero. Calculate its standard deviation and explain why the weighted average of the two standard deviations is not the answer. (Answer: Portfolio standard deviation is approximately 4.1037%; the 5.8000% weighted average overstates it.)
A hypothetical insurer has available capital of $95 million, required capital of $70 million, and a minimum target ratio of 1.30. A specified stress reduces available capital by $5.6 million and leaves required capital unchanged. Does the stressed ratio meet the target, and by how much does stressed capital exceed or fall short of target capital? (Answer: The stressed ratio is $89.4/70=1.2771$, so it misses the target; stressed capital is $1.60 million below the $91.00 million target.)

What You'll Learn