Allocation Bases, Constraints, and Reconciliation

Why Allocation Bases Must Reconcile

An aggregate capital target is useful for management only if it can be assigned to the portfolio components in a clear and controlled way. The assignment can inform performance assessment, pricing, or limits, but a chosen allocation basis does not change the aggregate target. It determines how that target is distributed among components, subject to the allocation policy and any applicable constraints.

The key accounting requirement is reconciliation: the amounts assigned to all components must add to the fixed aggregate target. A second requirement is feasibility: each assigned amount must satisfy the component’s bounds. These requirements are related but distinct. An allocation can add to the target and still violate a bound; alternatively, every component can satisfy its bounds while the amounts fail to add to the target.

The aggregate target \(K\) is the total amount to be assigned. The allocation basis \(b_i\) is a selected measure of component size or relevance, such as exposure, premium, expected loss, or a model-based indicator. The weight \(w_i\) is the share used to assign the target, and \(c_i\) is the resulting allocated amount. These are different quantities: a basis is an input to a policy, a weight is a share, and an allocated amount is expressed in the same units as \(K\).

An allocation basis must be chosen for a stated purpose and applied consistently. For example, a policy may use exposure because it is readily measured, or a model-based basis because it is intended to reflect the portfolio’s risk profile. The choice is a governance decision, not a consequence of the reconciliation equation. A reasonable basis should be documented along with the target, component definitions, and any constraints. Components should form a complete, non-overlapping partition for the allocation: otherwise some amount may be omitted or counted more than once.

From Bases to Reconciled Weights

If each raw basis \(b_i\) is nonnegative and their total is positive, one straightforward policy is to use each component’s share of the total basis. This produces weights that sum to one, so multiplying them by the fixed target automatically reconciles the component amounts. This is a basic normalization rule, not a claim that the selected basis is the only appropriate one.

The assumptions matter. The basis total must be positive for this normalization to be defined. Nonnegative bases give nonnegative weights; if a basis can be negative, its use as a share requires an explicit policy rather than automatic normalization. If the weights are specified directly rather than derived from bases, they must still sum to one for \(c_i=K w_i\) to reconcile. Rounding displayed weights can make them appear not to sum exactly to one, so calculations should use sufficient precision and reconcile the final amounts.

Constraints and Feasibility

An allocation policy may impose a floor \(L_i\), a cap \(U_i\), or both on a component. Bounds may represent a governance decision, a contractual or operational restriction, or a stated allocation rule. A proposed allocation must satisfy both the component bounds and the aggregate reconciliation condition. Simply clipping an unconstrained allocation at a cap or raising it to a floor generally changes the total.

Both parts of the condition are necessary. If an interval is empty, its component cannot satisfy its own bounds. If all bounds are satisfied, adding the lower bounds gives \(\sum_i c_i\geq\sum_i L_i\), while adding the upper bounds gives \(\sum_i c_i\leq\sum_i U_i\). Since reconciliation requires \(\sum_i c_i=K\), the target must lie between the two sums.

To see why the conditions are sufficient, start every component at its lower bound. The amount still to be assigned is \(R=K-\sum_i L_i\). The total room available above the lower bounds is \(\sum_i(U_i-L_i)=\sum_i U_i-\sum_i L_i\). The stated target condition gives \(0\leq R\leq\sum_i(U_i-L_i)\). Add the residual \(R\) across the components, never adding more than \(U_i-L_i\) to component \(i\). Because total available room is at least \(R\), this process can assign the entire residual without exceeding any cap. The resulting amounts satisfy the bounds and sum to \(K\). This proves the condition for continuous, real-valued allocations; extra requirements such as indivisible units or additional cross-component restrictions would need separate feasibility checks.

Reconciling Weights When Constraints Bind

An initial set of weights may produce amounts that violate a bound. The right response is not to keep the violation or silently alter the aggregate target. Instead, identify the required changes, find an allocation that satisfies all constraints, and verify the total. If more than one feasible allocation exists, the policy must specify how to choose among them; bounds and reconciliation alone do not determine a unique answer.

Common Reconciliation Traps

A frequent error is to apply a cap to one component and leave the other allocations unchanged. For example, with the initial amounts in the preceding example, suppose A instead has a $25 million cap, while B and C have the bounds stated there. Replacing A’s $36 million by $25 million but retaining B at $31.5 million and C at $22.5 million gives \(25+31.5+22.5=\$79\) million. The capped amounts are within the individual bounds, but $11 million of the fixed target is unallocated.

A feasible correction in that hypothetical cap case is \((\$25,\$35,\$30)\) million. The total is $90 million; A is at its cap, B is at its cap, and C is at its cap. The correction illustrates the two checks that must be made after any adjustment: bounds for each component and the sum across components. It does not prescribe a universal redistribution rule.

Another important trap is using the summed-bound test without checking each interval. For example, take \(L_1=10,U_1=0,L_2=0,U_2=10\), and \(K=10\). The sums happen to satisfy \(10\leq10\leq10\), but component 1 requires \(c_1\geq10\) and \(c_1\leq0\) at the same time. Its interval is empty, so no feasible allocation exists. Always check \(L_i\leq U_i\) for every component before relying on the aggregate bound sums.

For capital governance, a reconciliation check is a basic control, not a substitute for a sound allocation rationale. An allocation can reconcile perfectly while relying on a poorly chosen basis or an undocumented adjustment. Conversely, a well-motivated basis does not excuse amounts that fail the fixed-target or component-bound requirements. Clear definitions, explicit constraints, and arithmetic checks make the allocation interpretable and reproducible. These distinctions are aligned with the CAS Exam 9 (Risk Management for Actuaries) syllabus.