Solvency Constraints and the Shadow Price of Capital

Why a Capital Constraint Changes a Decision

Two actions can have different economic values and require different amounts of capital. If capital is plentiful, the preferred action may simply be the one with the highest value. If available capital is limited, that action may not be feasible. The decision then depends on both the value of an action and whether it satisfies the solvency constraint.

A useful way to formalize this problem is to define a decision $a$, its value $V(a)$, its capital requirement $C(a)$, and available capital capacity $K$. The objective is to maximize value subject to $C(a)\leq K$. This framework can describe a choice among a finite set of actions or an adjustable decision such as the amount of risk transferred. The capital measure and valuation basis must be stated consistently: in the examples below, all monetary amounts are hypothetical, measured in thousands of dollars, and valued at the same date. No interest discounting, mortality assumption, or probability distribution is needed; the stated values and capital requirements are treated as deterministic inputs.

Definition: A solvency constraint requires a decision to use no more than the available capital capacity: $C(a)\leq K$. A decision is feasible if it satisfies this inequality and any other stated restrictions. A binding constraint holds with equality at the selected decision. The shadow price of capital capacity, when it exists, is the marginal increase in the optimized objective from a small increase in $K$.

Formulating the Constrained Decision

For a continuous decision, the basic problem is

\max_{a\in\mathcal{A}} V(a) \quad\text{subject to}\quad C(a)\leq K.

Here $\mathcal{A}$ is the set of otherwise allowable decisions. The constraint is not a penalty that can be ignored when an action has high value: an infeasible action cannot be selected. When the constraint binds, a decision-maker may have to choose a lower-value action that uses less capital, or adjust the decision to bring its capital requirement within capacity.

For differentiable functions and an interior optimum apart from the capital constraint, define the Lagrangian

\mathcal{L}(a,\lambda)=V(a)-\lambda\bigl(C(a)-K\bigr), \qquad \lambda\geq 0.

The nonnegative multiplier $\lambda$ is the shadow price associated with the capital constraint. Under the usual regularity conditions for a smooth constrained optimization problem, the first-order conditions include

V'(a)-\lambda C'(a)=0,\qquad C(a)\leq K,\qquad \lambda\geq 0,\qquad \lambda\bigl(C(a)-K\bigr)=0.

The last equation is complementary slackness. If the constraint is slack, $C(a)<K$, it requires $\lambda=0$. If $\lambda>0$, the constraint must bind. These conditions help explain the trade-off at a constrained optimum: the marginal change in value is balanced by the shadow-priced marginal change in capital use.

Why the Multiplier Measures Marginal Capacity Value

Let $W(K)$ be the maximum value attainable with capacity $K$, and suppose the optimum $a^*(K)$ is smooth and the capital constraint binds. Then $C(a^*(K))=K$. Differentiate the optimized value with respect to capacity:

\frac{dW}{dK} = V'(a^*)\frac{da^*}{dK}.

At an optimum satisfying the Lagrangian first-order condition, $V'(a^*)=\lambda C'(a^*)$. Also, differentiating the binding constraint $C(a^*(K))=K$ gives $C'(a^*)\,da^*/dK=1$. Substituting both results yields

\frac{dW}{dK} = \lambda C'(a^*)\frac{da^*}{dK} = \lambda.

Thus, under these assumptions, $\lambda$ is the local marginal value of relaxing the capacity constraint. It is not necessarily the total value of a large capital increase. The optimum may move as capacity changes, and the multiplier may change as well. Nor should a smooth multiplier be applied mechanically to a finite set of discrete choices: there, feasibility and optimized value can change in jumps.

Worked Example: Selecting Among Discrete Actions Under a Capital Limit

A hypothetical insurer compares four mutually exclusive actions. Their net economic values, relative to a common baseline, and their required capital are given below in thousands of dollars. Treat these figures as deterministic at one valuation date; no interest, mortality, or distributional assumptions are used. With capacity $K=70$, select the highest-value feasible action.

Action	Capital requirement	Net economic value
No action	62	0
Treaty A	68	6
Portfolio adjustment B	74	10
Portfolio adjustment C	66	8

Method. First test feasibility using the capital requirement, then compare values only among feasible actions.

Step 1: Test each requirement. No action is feasible because $62\leq70$. Treaty A is feasible because $68\leq70$. Adjustment B is infeasible because $74>70$. Adjustment C is feasible because $66\leq70$.

Step 2: Compare feasible values. The feasible values are 0 for no action, 6 for Treaty A, and 8 for adjustment C. The largest is 8, so adjustment C is preferred at $K=70$.

Step 3: Check the threshold effect. At $K=64$, no action remains feasible because $62\leq64$, but Treaty A and both adjustments are infeasible. At $K=74$, adjustment B becomes feasible and its value of 10 exceeds the value of C. Thus, the preferred action changes when capacity crosses a feasibility threshold.

Common mistake avoided. It is wrong to say that no action is infeasible at $K=70$: its requirement is 62, and $62\leq70$. The inequality must be checked in the correct direction, separately for every alternative. This finite-choice example also shows why a single smooth shadow price need not describe the value of capacity across a threshold.

A Binding Constraint and a Continuous Decision

For a continuous decision, a binding constraint can change the selected decision even when the unconstrained optimum is well-defined. The next example makes the comparison explicit. The value function represents net economic value, so the capital requirement is not itself subtracted from that value.

Worked Example: Finding the Constrained Optimum and Shadow Price

A hypothetical risk-transfer program is indexed by $x$, the fraction of a specified layer transferred, where $0\leq x\leq1$. Its net value and capital requirement, both in thousands of dollars, are

V(x)=48x-30x^2,\qquad C(x)=100-40x.

Available capital capacity is $K=66$. Treat the functions as deterministic, measured at the same date; no interest, mortality, or probability distribution is assumed. Find the optimal $x$, the achieved value, and the local shadow price if the constraint binds.

Step 1: Find the unconstrained optimum. Differentiate the value function:

$$ V'(x)=48-60x. $$

Setting $V'(x)=0$ gives $x=48/60=0.8$. Since $V''(x)=-60<0$, this is the unconstrained maximum on the permitted range. Its capital requirement is $C(0.8)=100-40(0.8)=68$, which exceeds $K=66$. The unconstrained optimum is infeasible.

Step 2: Convert the capacity limit into a restriction on $x$.

100-40x\leq66 \quad\Longrightarrow\quad -40x\leq-34 \quad\Longrightarrow\quad x\geq0.85.

The feasible values therefore begin at $x=0.85$. Since $V'(x)=48-60x<0$ for every $x\geq0.85$, value decreases over the feasible interval. The constrained maximum is consequently the smallest feasible choice, $x=0.85$, and the capital constraint binds.

Step 3: Compute and check the achieved value.

$$ V(0.85)=48(0.85)-30(0.85)^2 =40.8-30(0.7225) =40.8-21.675 =19.125. $$

Also, $C(0.85)=100-40(0.85)=100-34=66$, exactly the available capacity. The optimal value is 19.125 thousand dollars, or $19,125.

Step 4: Calculate the multiplier. At the binding optimum, $V'(0.85)=48-51=-3$ and $C'(0.85)=-40$. The first-order condition $V'-\lambda C'=0$ gives

-3-\lambda(-40)=0 \quad\Longrightarrow\quad \lambda=\frac{3}{40}=0.075.

The local shadow price is 0.075 thousand dollars of value per additional thousand dollars of capacity: equivalently, 7.5 cents of value per dollar of capacity, for a sufficiently small increase around $K=66$. It is a local marginal quantity, not a claim that every additional thousand dollars of capacity will always create 75 dollars of value.

Common mistake avoided. Do not retain $x=0.8$ just because it maximizes the unconstrained value. It requires 68 thousand dollars of capital, more than the available 66. Nor should the constraint be rearranged as $x\leq0.85$: dividing by the negative coefficient $-40$ reverses the inequality, giving $x\geq0.85$.

Using a Shadow Price to Compare Capital-Intensive Choices

A shadow price can express the opportunity cost of using scarce capacity. In a Lagrangian comparison, an action with value $V$ and capital requirement $C$ has a capacity-adjusted score $V-\lambda C$. A larger $\lambda$ penalizes capital-intensive actions more heavily. This adjusted score is useful for interpreting a multiplier or comparing marginal alternatives; it does not replace the original feasibility test.

Worked Example: Finding the Shadow-Price Crossover Between Actions

A hypothetical firm must choose either action A or action B, or retain the baseline. Their incremental net values and capital requirements are in thousands of dollars: A has value 7 and requires 18 of capital; B has value 4 and requires 6; the baseline has value and capital requirement both equal to 0. All values are deterministic at the same valuation date, with no interest, mortality, or distributional assumptions. Determine when the shadow-price-adjusted comparison favors each action, and then identify the feasible choice at $K=10$.

Step 1: Compare adjusted scores. The scores for A and B are $7-18\lambda$ and $4-6\lambda$, respectively. Action A is preferred to B when

7-18\lambda>4-6\lambda \quad\Longleftrightarrow\quad 3>12\lambda \quad\Longleftrightarrow\quad \lambda<0.25.

They tie at $\lambda=0.25$. For $\lambda>0.25$, B has the larger adjusted score because it uses less capital. The crossover has units of thousands of dollars of value per thousand dollars of capital, or dollars of value per dollar of capital.

Step 2: Apply the actual capacity limit. At $K=10$, A is infeasible because $18>10$. B is feasible because $6\leq10$, and the baseline is feasible because $0\leq10$. Among these feasible choices, B has the largest net value, 4 rather than 0, so B is selected.

Step 3: Interpret without overclaiming. The crossover at 0.25 describes the comparison of adjusted scores. It is not, by itself, a proof that the exact discrete constrained problem has a smooth multiplier equal to 0.25. Discrete decisions can have threshold effects, and the feasibility test at $K=10$ remains decisive.

Common mistake avoided. Choosing A solely because its unadjusted value, 7, exceeds B’s value, 4, ignores feasibility: A cannot be selected with capacity 10. Conversely, a shadow-adjusted score should not be treated as permission to select an infeasible action.

Interpreting Constraints and Avoiding Exam Traps

A defensible constrained-capital analysis separates three questions: what is being maximized, what capital measure is constrained, and which alternatives satisfy that constraint. The value objective should use a consistent basis across alternatives. The capacity limit must use the same capital measure and valuation basis as the requirements being compared. If an action changes both value and capital, its net value and capital requirement should be evaluated together rather than comparing only one of those changes.

The shadow price adds an interpretation to the optimization: it estimates the marginal value of a small relaxation of a binding constraint under stated smoothness and regularity assumptions. When the constraint is slack, the multiplier is zero in the standard formulation because a small increase in capacity does not change the selected optimum. When choices are discrete, the value function may be flat and then jump as an action becomes feasible. In that setting, report the actual feasible choices and their values; do not present a local derivative as an exact measure of a finite change.

Check the inequality direction. An action is feasible when its capital requirement is less than or equal to capacity.
Separate feasibility from ranking. First exclude infeasible actions; then compare the objective among those that remain.
State units. A multiplier is measured in units of objective value per unit of capacity.
Distinguish a local from a finite change. A shadow price may change as capacity changes and may not apply across a discrete threshold.
Keep the constraint measure consistent. A comparison is not meaningful if requirements and capacity use different definitions or valuation bases.

Key takeaway: A solvency limit can change which action is preferred by changing the feasible set. The shadow price quantifies the local value of easing a binding limit in a smooth model, but discrete choices require explicit feasibility and value comparisons.

Check Your Understanding

Work each problem by checking the constraint before ranking alternatives. All monetary amounts are hypothetical and in thousands of dollars; unless specified otherwise, inputs are deterministic at a common valuation date, with no interest, mortality, or distributional assumption.

A firm has capacity $K=50$. No action requires 41 of capital and has value 0; action A requires 47 and has value 4; action B requires 53 and has value 9; action C requires 49 and has value 6. Which action is preferred, and which listed actions are feasible? (Answer: No action, A, and C are feasible; C is preferred with value 6.)
A continuous choice $x\in[0,1]$ has $V(x)=30x-20x^2$ and $C(x)=60-20x$. With $K=48$, find the optimal $x$, its capital requirement, and its value. (Answer: The unconstrained optimum $x=0.75$ is feasible; $C(0.75)=45$ and $V(0.75)=11.25$, so $x=0.75$ is optimal and the constraint is slack.)
A continuous decision has $V(x)=36x-24x^2$, $C(x)=90-30x$, $0\leq x\leq1$, and $K=66$. Find the constrained optimum, value, and local shadow price. (Answer: Feasibility requires $x\geq0.8$; the unconstrained optimum $0.75$ is infeasible, so $x=0.8$, $V(0.8)=13.44$, and $\lambda=(36-48)/(-30)=0.4$ thousand dollars of value per thousand dollars of capacity.)
Two actions have value-capital pairs A $(7,18)$ and B $(4,6)$. At what shadow price do their adjusted scores tie, and which has the greater adjusted score above that value? (Answer: They tie at $\lambda=0.25$; B has the greater adjusted score for $\lambda>0.25$.)
Three mutually exclusive choices have requirements and values: baseline $(0,0)$, X $(12,8)$, and Y $(8,6)$. Find the preferred choice at $K=9$ and at $K=12$, and describe the value change. (Answer: At $K=9$, Y is preferred with value 6; at $K=12$, X is preferred with value 8. The optimized value increases by 2 when capacity rises from 9 to 12, but that finite change does not establish a constant marginal shadow price across the interval.)

What You'll Learn