Question
Two departments earn uncertain net revenues \(X\) and \(Y\), measured in thousands of dollars. Their risk estimates are
\[
\operatorname{SD}(X)=10,\qquad \operatorname{SD}(Y)=6,\qquad
\operatorname{Cov}(X,Y)=-48.
\]
The business cares about the standard deviation of the combined revenue \(X+Y\). Management wants to allocate that combined risk back to the two departments according to the marginal effect of each department's position, rather than in proportion to its standalone standard deviation.
Introduce temporary exposure weights \(a,b\) on \(X,Y\), differentiate \(\operatorname{SD}(aX+bY)\), and use the derivatives at \(a=b=1\) to obtain each department's Euler risk contribution. Does one department have a negative contribution? Check that the two contributions sum to total risk, and that the covariance numbers are possible.
Get hint
Hint 1
First calculate \(\operatorname{Var}(aX+bY)=100a^2+36b^2-96ab\). Differentiate its square root, not just the variance.
Check your answer
Solution
Define
\[
\sigma(a,b)=\sqrt{100a^2+36b^2-96ab}.
\]
At the actual exposure \(a=b=1\), total variance is \(100+36-96=40\), so \(\sigma=\sqrt{40}\) thousand dollars. When \(\sigma>0\), differentiation gives
\[
\frac{\partial\sigma}{\partial a}=\frac{100a-48b}{\sigma},\qquad
\frac{\partial\sigma}{\partial b}=\frac{36b-48a}{\sigma}.
\]
The (weight times marginal derivative) Euler contributions at unit weights are
\[
RC_X=1\cdot\frac{100-48}{\sqrt{40}}=\boxed{\frac{52}{\sqrt{40}}\approx8.222},
\]
\[
RC_Y=1\cdot\frac{36-48}{\sqrt{40}}=\boxed{-\frac{12}{\sqrt{40}}\approx-1.897}.
\]
Their sum is \((52-12)/\sqrt{40}=40/\sqrt{40}=\sqrt{40}\), exactly the portfolio's standard deviation. This is an instance of the Euler identity for a positively homogeneous risk measure.
A negative marginal risk contribution does not mean Y has negative standalone risk: Y alone has standard deviation 6. It means adding a small amount of Y to the current mix reduces total standard deviation because the covariance with X is sufficiently negative. The portfolio risk remains nonnegative.
Finally, the implied correlation is \(\rho=-48/(10\cdot6)=-0.8\), within \([-1,1]\). The covariance matrix is positive definite because \(100\cdot36-48^2=1296>0\). Thus the proposed risks can arise jointly.
Assumptions
The exposures are exactly one unit of each department; the temporary weights exist only to define derivatives.
The provided standard deviations, covariance and combined revenue exhaust the risk model; total risk is nonzero.
Risk contribution means weight times marginal contribution to portfolio standard deviation, in thousand-dollar SD units—not a literal cash transfer.
Final Answer
Total SD = \(\sqrt{40}\approx6.325\) thousand dollars. Euler contributions: \(X:52/\sqrt{40}\approx8.222\) and \(Y:-12/\sqrt{40}\approx-1.897\) thousand dollars; they sum to \(\sqrt{40}\). Correlation is −0.8, hence the covariance is valid.