Why Measure the Capital Change?
A capital model may be asked what happens to aggregate capital if a business unit is added, a treaty limit is reduced, or a block of policies is removed. The relevant quantity is the change in the portfolio's capital between the two specified states. It is not necessarily the capital that the exposure would require on its own, nor an allocation of the current portfolio's capital.
This distinction matters in pricing, portfolio decisions, and model analysis. The capital impact of an exposure can depend on what else is already in the portfolio and on how much exposure is added or removed. A result is therefore meaningful only when its baseline portfolio, change in exposure, and capital measure are clear. The methods here use discrete model outputs; they do not require a smooth capital function.
The exposure \(h\) may be a whole business unit, a contract, or a specified fraction of a portfolio. If the change has a meaningful size measure, one may also divide the capital change by that size. This gives an average capital change per unit of exposure over the particular interval; it is not automatically the effect of a different-sized change.
Calculating Discrete Contributions
The calculation requires two comparable model results: capital before the change and capital after it. The model configuration, capital measure, valuation date, and other assumptions should be held fixed unless the purpose is specifically to assess a change in those items. Otherwise, the difference combines the exposure change with other effects.
Identify the portfolio state and its aggregate capital \(K(x)\).
Name the exposure and the amount added or removed. Use the same exposure definition in both model runs.
Record capital for the portfolio after the change, using the same model settings and capital measure.
For addition, subtract baseline capital from post-addition capital. For a removal contribution, subtract post-removal capital from baseline capital.
State the portfolio baseline, exposure size, units, and whether the result is an addition or removal contribution.
Worked Example: Adding a Business Unit
A hypothetical insurer's model reports aggregate capital of $24.8 million for its current portfolio. After adding a new specialty unit, the model reports $29.6 million. The specialty unit's standalone capital, modeled by itself, is $6.2 million. Calculate its finite-difference contribution to the existing portfolio and compare it with standalone capital.
Method. Use the addition difference: capital after adding the unit minus capital before adding it.
Thus the unit increases capital by $4.8 million in this particular portfolio. The difference between its standalone capital and its portfolio contribution is
Relative to standalone capital, the contribution is \(4.8/6.2=0.77419\ldots\), or 77.42% to two decimal places. This comparison describes the model results; it does not establish that the unit should be allocated exactly $4.8 million under every allocation method. Check: adding the calculated contribution back to baseline capital gives \(24.8+4.8=29.6\) million, the reported post-addition result. The common mistake is to report $6.2 million as the contribution: that is standalone capital, not the change in aggregate capital when the unit joins this portfolio.
Why the Difference Formula Works
The finite difference follows directly from the meaning of a change. Let \(K(x)\) be the capital of the baseline portfolio and \(K(x+h)\) the capital after adding exposure \(h\). The amount by which the model's capital changes is the final value minus the initial value:
For removal, the starting value is \(K(x)\) and the ending value is \(K(x-h)\). The signed change is therefore \(K(x-h)-K(x)\). If instead we report the amount of capital associated with the removed exposure, with a positive sign when capital falls, we reverse that subtraction:
This is a finite comparison between two specified portfolio states. It does not require a derivative, and it remains usable when exposures change in whole units or the model's capital output is not smooth.
Worked Example: Removing All or Part of a Line
A hypothetical insurer has aggregate capital of $58.7 million with a regional liability line included. The model reports $53.9 million after the entire line is removed. It also reports $56.6 million if only 40% of the line is removed. Calculate the contribution of the full removal, the contribution of the 40% reduction, and the signed capital change for that partial removal.
Full removal. Subtract post-removal capital from baseline capital:
The removed line's finite-difference contribution, for removal of the whole line from this portfolio, is $4.8 million.
Partial removal. The capital contribution associated with removing 40% is
The signed change in capital from that removal is instead \(56.6-58.7=-2.1\) million dollars: capital decreases by $2.1 million. If the 40% is treated as a 0.40 fraction of the line, the average contribution per full-line fraction is \(2.1/0.40=5.25\) million dollars per full-line equivalent. This is an average over the specified partial change, not a guarantee that removing a different fraction would produce a proportional change.
Check. The partial-removal capital result reconciles as \(58.7-2.1=56.6\) million. A common sign error is to call \(-2.1\) million the removal contribution. That is the signed change in aggregate capital; under the stated removal-contribution convention, the contribution is positive $2.1 million.
Baseline, Change Size, and Interpretation
A finite-difference contribution answers a conditional question: what is the capital change for this specified exposure change from this specified starting portfolio? It is not an intrinsic number attached to an exposure regardless of context. A different baseline can produce a different change, and a larger change in exposure need not produce a proportionally larger capital difference.
The size of the step must also be reported. For a change from \(x\) to \(x+h\), the average change per unit of exposure is
when \(h\) is a positive numerical exposure amount. This ratio summarizes the interval from \(x\) to \(x+h\). It should not be presented as the effect of every possible additional unit unless that constancy has been established.
Worked Example: Comparing Sequential Additions
A hypothetical model reports capital for a portfolio with a fixed set of other exposures and \(n\) identical contract blocks as follows. Capital is in millions of dollars.
| Number of blocks \(n\) | Aggregate capital \(K(n)\) |
|---|---|
| 0 | $18.0 million |
| 1 | $19.1 million |
| 2 | $20.5 million |
| 3 | $22.4 million |
Calculate the contribution of each block when it is added in sequence, the contribution of the third block when added to the two-block portfolio, and the total change from zero to three blocks.
First block: \(19.1-18.0=1.1\) million dollars. Second block: \(20.5-19.1=1.4\) million dollars. Third block: \(22.4-20.5=1.9\) million dollars. The third block's contribution is $1.9 million specifically when added to the portfolio that already contains two blocks.
Adding the three sequential contributions gives \(1.1+1.4+1.9=4.4\) million dollars. Directly comparing the endpoints gives \(22.4-18.0=4.4\) million dollars, so the totals agree. The per-block average over the full three-block change is \(4.4/3=1.4666\ldots\), or $1.47 million per block to the nearest cent in millions. This average is not the contribution of each individual block: the model results show different increments at different portfolio sizes. Check: the three differences telescope to the endpoint difference. The common mistake is to use the first block's $1.1 million contribution for all three blocks; that would give $3.3 million, which does not reconcile to the modeled endpoint change.
Sequential Changes and Practical Controls
For a sequence of discrete additions, let the successive portfolio states be \(x_0,x_1,\ldots,x_m\). The change at step \(j\) is \(K(x_j)-K(x_{j-1})\). Adding these differences cancels each intermediate capital result once with a positive and once with a negative sign:
This identity is useful as a reconciliation check. It does not mean that the contribution assigned to each exposure is independent of the order in which exposures are added. If portfolio interactions matter, changing the order can change the individual step differences even though the total endpoint change remains the same.
When using a capital model for a decision, document at least the following:
- the baseline portfolio and the exact exposure change;
- the capital measure, model version, and material assumptions held constant;
- whether the output is an addition contribution, a removal contribution, or a signed capital change;
- the units and size of the exposure step; and
- the model results before and after the change, so the subtraction can be checked.
Common Mistakes
- Substituting standalone capital. Standalone capital measures an exposure in isolation. A finite-difference contribution compares the aggregate portfolio before and after the exposure changes.
- Leaving out the baseline. A contribution is conditional on the portfolio state from which the exposure is added or removed.
- Reversing the sign for removal. The signed change from removal is post-removal minus baseline; the removal contribution under the convention here is baseline minus post-removal.
- Ignoring step size. A contribution for a partial reduction is not necessarily a fixed fraction of the contribution for full removal.
- Calling a contribution an allocation. A finite difference is a model comparison for a specified change. It does not, by itself, define a complete allocation of aggregate capital across all exposures.
Check Your Understanding
For each problem, identify the relevant baseline and exposure change before calculating.
- A hypothetical portfolio has capital of $36.5 million. Adding a new unit produces capital of $41.2 million, while the unit's standalone capital is $6.0 million. What is its addition contribution, and how does it compare with standalone capital? (Answer: $4.7 million; it is $1.3 million below standalone capital.)
- A portfolio has capital of $73.4 million before a line is removed and $69.8 million afterward. State the removal contribution and the signed capital change from removal. (Answer: Removal contribution $3.6 million; signed change \(-$3.6\) million.)
- With other exposures fixed, model capital is $12.0 million with zero blocks, $12.8 million with one, $14.0 million with two, and $15.5 million with three. Find the sequential contribution of each block and check the total. (Answer: $0.8 million, $1.2 million, and $1.5 million; total $3.5 million, equal to \(15.5-12.0\).)
- A 25% reduction in a hypothetical portfolio segment lowers aggregate capital from $50.0 million to $48.9 million. Calculate the contribution for the reduction and the average change per percentage point of the segment reduced. Does this establish the capital effect of a full removal? (Answer: $1.1 million contribution; $0.044 million per percentage point; no, the partial-change average does not establish the full-removal effect.)
- Starting from a portfolio with capital of $20.0 million, adding exposure P first raises capital to $23.4 million, then adding Q raises it to $27.1 million. If Q is added first, capital becomes $24.2 million, and adding P then gives the same $27.1 million endpoint. Find each exposure's sequential contribution in both orders. (Answer: P then Q: P contributes $3.4 million and Q $3.7 million. Q then P: Q contributes $4.2 million and P $2.9 million. Both sequences total $7.1 million.)