Capital Rationing and Project Portfolio Optimization

Introduction

Capital budgeting is often presented as a ranking exercise: accept projects with positive net present value (NPV) and reject those with negative NPV. In practice, however, firms rarely possess unlimited capital, managerial attention, skilled labour, production capacity, data, or implementation time. A firm may therefore face several attractive projects but be unable to undertake all of them. This situation is known as capital rationing.

Capital rationing converts project appraisal into a portfolio decision. The question is no longer simply, “Is this project financially attractive?” It becomes, “Which combination of projects creates the greatest value while respecting financial, operational, strategic, risk, and timing constraints?” The distinction is important because the individually highest-NPV projects may not form the highest-value feasible portfolio.

This chapter develops an integrated framework for selecting projects under constraints. It explains profitability indices, linear and integer optimization, portfolio risk, project interdependencies, sequencing, capacity limits, scenario analysis, governance, and implementation. The emphasis is on graduate-level decision quality: sound valuation, explicit assumptions, transparent optimization, and managerial judgment.

Learning Objectives

By the end of this chapter, the reader should be able to:

1. Explain the causes and forms of capital rationing.
2. Calculate and interpret profitability indices and distinguish them from NPV.
3. Formulate project selection as a linear or integer optimization problem.
4. Incorporate financial, operational, strategic, risk, and timing constraints.
5. Analyze project dependencies, mutually exclusive alternatives, and indivisibilities.
6. Evaluate portfolio risk using covariance and scenario-based methods.
7. Apply sequencing and capacity analysis to multi-period investment programs.
8. Critically assess optimization results and communicate recommendations to decision-makers.

1. The Nature of Capital Rationing

Capital rationing exists when the firm imposes a limit on investment, whether because of external financing constraints or internal policy. It may be:

• Hard rationing, caused by binding external limits such as debt covenants, unavailable credit, regulatory capital requirements, or a borrowing ceiling.
• Soft rationing, caused by managerial policies such as annual spending limits, preservation of credit quality, or a desire to maintain strategic flexibility.
• Single-period rationing, in which the constraint applies to one budget period.
• Multi-period rationing, in which investments compete for capital over several years.

Rationing may concern more than cash. Typical constraints include engineering hours, production lines, information-technology capacity, scarce technical specialists, regulatory approvals, carbon emissions, working capital, and senior-management attention.

A project can have a positive NPV but still be excluded from the optimal portfolio. Exclusion does not necessarily imply that the project is poor; it may simply be inferior to competing uses of a scarce resource.

2. Value Measures for Constrained Selection

2.1 Net present value

For a conventional project with investment and operating cash flows, NPV is:

where I₀ is the initial investment at time 0, CF_t is the incremental after-tax cash flow in period t, r is the appropriate risk-adjusted discount rate per period, and T is the project life. Cash flows and the discount rate must use consistent timing and units.

NPV measures absolute value added. Under an unconstrained capital market, accepting every independent project with positive NPV maximizes value, assuming cash-flow estimates and the discount rate are appropriate.

2.2 Profitability index

When a single investment budget is binding, managers often use the profitability index (PI):

Equivalently:

A PI greater than 1 indicates positive NPV. PI is a relative measure of value per unit of initial investment, whereas NPV is an absolute measure of value. Under a single, divisible, one-period capital constraint, ranking by PI can be useful. It is not universally optimal when projects are indivisible, mutually dependent, multi-period, or subject to several resource constraints.

Worked example: PI ranking

Assume a firm has a $10 million budget and three independent projects:

Projects A and B together require $7 million and create $2.1 million of NPV. The remaining $3 million cannot fund C. A simple PI ranking therefore does not automatically identify the best feasible portfolio. If projects are indivisible, the feasible choices include A+B, C alone, or perhaps other projects not shown. A+B is preferred to C because its NPV is $2.1 million rather than $1.2 million.

This example illustrates two limitations of PI. First, equal or similar indices can conceal different absolute contributions. Second, choosing projects one at a time may leave unused capital and fail to identify the best combination.

2.3 Incremental analysis

When projects are mutually exclusive, the firm should compare their incremental cash flows rather than rank their individual PIs. A high-PI small project may be inferior to a larger alternative that creates more absolute value. Mutually exclusive projects may represent different technologies, plant sizes, or strategic routes; selecting one prevents selecting another.

3. Optimization Formulation

3.1 Binary project selection

For indivisible projects, define a binary decision variable:

If NPV_i is the NPV of project i, the basic objective is:

subject to a capital constraint:

where I_i is the required investment for project i and B is the available budget. The conditions x_i ∈ {0,1} make this a binary integer programming problem.

Additional constraints can be added directly. For resource k:

where a_{ki} is project i’s use of resource k and C_k is the available capacity.

3.2 Project selection example

Suppose the firm has a $12 million budget:

Assume engineering capacity is 10 hours. The feasible combinations include A+B, A+D, B+C, B+D, and C+D, subject to both constraints. Their results are:

The optimal feasible portfolio is A+C, with NPV of $6.8 million. A+B+D would have a higher apparent NPV of $7.1 million but violates engineering capacity. This demonstrates why both financial and operational constraints must be modeled.

3.3 Linear programming and divisibility

If a project can be scaled continuously, x_i may be allowed to take values between zero and one:

This produces a linear programming model. For example, a solar installation may be divisible into standardized capacity blocks, whereas a regulatory platform may be indivisible. Fractional solutions are economically meaningful only when partial implementation is technically and commercially feasible.

A linear programming solution can provide a useful upper bound even when projects are ultimately indivisible. Integer optimization is required when the decision variables must be whole projects, whole facilities, or whole contracts.

3.4 Common logical constraints

Optimization models become more realistic through logical relationships:

• Mutual exclusivity: x_A + x_B ≤ 1.
• Prerequisite: x_B ≤ x_A, meaning B can be chosen only if A is chosen.
• At least one project: x_A + x_B ≥ 1.
• Conditional selection: x_C ≤ x_D + x_E, meaning C requires D or E.
• Portfolio limit: ∑x_i ≤ m, limiting the number of projects.
• Minimum strategic allocation: ∑_{i∈S} I_i x_i ≥ L, where S is a strategic category.

These constraints should be justified operationally. Artificial constraints can produce precise but economically meaningless results.

4. Portfolio Risk and Correlation

Selecting projects solely by expected NPV can create excessive concentration. Projects may be exposed to the same commodity price, customer segment, technology, regulator, or macroeconomic condition.

Let R_p denote portfolio return, w_i the proportion of invested capital in project i, and R_i its return. Expected portfolio return is:

For two projects, portfolio variance is:

where σ_A and σ_B are the standard deviations of project returns and ρ_AB is the correlation coefficient. For n projects, the matrix form is:

where w is the vector of weights and Σ is the covariance matrix. The formula assumes that returns are represented consistently and that covariance estimates are meaningful. In capital budgeting, project cash-flow covariance may be more informative than accounting-return covariance.

Diversification reduces risk when project outcomes are not perfectly positively correlated. However, diversification is not automatically desirable: a strategically essential project may be retained despite high correlation, and weak projects should not be acquired merely to reduce measured variance.

Scenario-based portfolio analysis

Where probability distributions are unreliable, scenario analysis can be preferable. Let V_{ps} be the portfolio NPV under scenario s and p_s its probability. Then:

with ∑p_s = 1. A downside measure can be defined as the probability that NPV is negative, or as the expected shortfall below a specified threshold. Scenario analysis is especially useful for correlated shocks, such as a recession affecting several projects simultaneously.

5. Interdependencies and Real Options

Projects are often not independent. Their relationships may be:

• Complementary: one project increases the cash flows of another.
• Substitutive: selecting one reduces the value of another.
• Sequential: an initial investment creates the right, but not the obligation, to invest later.
• Enabling: an infrastructure project is required before commercial projects can proceed.
• Cannibalistic: a new product reduces sales of an existing product.

If project B creates an incremental benefit when combined with A, the portfolio NPV should include the interaction term:

where S_AB is the present value of the synergy, measured relative to valuing the projects separately. If S_AB is negative, it represents cannibalization or conflict.

The simple additive objective ∑NPV_i x_i is therefore incomplete when interactions are material. A quadratic formulation may be used:

This model remains exact only if interaction effects are adequately estimated and do not depend on unmodeled combinations.

Sequential projects may contain real options. A pilot project can generate information about demand or technology and permit expansion if conditions are favorable. Treating the pilot only as a conventional project may understate its option value; conversely, managers should avoid assigning speculative option value without a defensible decision tree or scenario model.

6. Timing, Sequencing, and Multi-Period Capacity

A portfolio may be financially feasible in total but infeasible in a particular year. Let I_{it} be project i’s investment in period t and B_t the available budget in that period:

where x_{it} indicates whether the relevant stage occurs in period t. Precedence can be represented as:

meaning stage B in period t requires stage A in the previous period. Timing affects value because delaying or accelerating cash flows changes their present value. If the project’s cash flow in period t is CF_t, discounting at rate r yields CF_t ÷ (1+r)^t.

Sequencing can also reduce risk. A firm may fund feasibility work before committing to full construction, or launch a product in one region before global expansion. The cost of delay must be compared with the value of information and flexibility.

Capacity constraints require careful definition. A resource may be:

• Renewable, such as staff hours available each year.
• Non-renewable, such as a fixed land allocation.
• Lumpy, such as a production line that becomes available only in whole units.
• Time-dependent, such as a specialist team whose capacity changes across periods.

Capacity slack is not necessarily wasteful. Maintaining unused capacity may protect resilience, enable rapid response, or reduce operational risk. Optimization should therefore distinguish economically valuable slack from avoidable underutilization.

7. Shadow Prices and Managerial Interpretation

In a constrained optimization model, the shadow price of a binding constraint estimates the marginal increase in objective value from relaxing that constraint by one unit, within the relevant range and subject to model assumptions. If the capital constraint has a shadow price of $0.18 million per additional $1 million of capital, obtaining an extra $1 million is valuable only if its financing cost, risk, and strategic consequences are less than that benefit.

Shadow prices are local estimates. They may change abruptly when the optimal portfolio changes, and they are not reliable outside the allowable range of the optimization solution. They also exclude factors omitted from the model, such as financing flexibility or reputational risk.

Sensitivity analysis should vary:

• Project NPVs and cash-flow forecasts.
• Discount rates and terminal values.
• Budget and capacity limits.
• Correlations and scenario probabilities.
• Synergy assumptions.
• Project completion dates.

A robust portfolio is one that remains acceptable across plausible assumptions, not merely one that is optimal under a single forecast.

8. Integrated Illustrative Case: Orion Mobility Systems

The following case is fictional and intended for instructional use.

Background

Orion Mobility Systems, a manufacturer of electric-vehicle components, has a three-year investment program. Management has $18 million available in year 1, $12 million in year 2, and $10 million in year 3. Engineering capacity is 12 specialist-years in year 1 and 10 specialist-years in year 2. The board also requires at least one project supporting low-carbon production.

The candidate projects are:

A and E are mutually exclusive because both require the same factory area. C cannot proceed unless A is selected. Project NPVs already incorporate the expected value of the pilot’s expansion opportunity.

Problem

A ranking by NPV would select A, C, B, E, and D in descending order, but that list violates several constraints. Total investment would exceed available capital, C requires A, and A conflicts with E.

Analysis

Let x_A through x_E be binary variables. The objective is:

Year-1 budget:

Year-2 budget:

Year-3 budget:

Year-1 engineering:

Logical constraints are:

and

The low-carbon requirement is:

because D is the only project satisfying the stated requirement. With D selected, A cannot be selected if the year-1 budget and engineering limits are binding alongside B. A feasible high-value portfolio is B+C+D only if C’s prerequisite A is satisfied; it is not, so that portfolio is invalid. The feasible candidates must therefore be checked carefully rather than assembled by superficial ranking.

A and D together require $20 million in year 1 and are infeasible. E and D require $15 million and 10 engineering units, leaving $3 million of year-1 capital but no additional candidate that can be funded. Their NPV is $6.8 million. B and D require $14 million and 9 engineering units; their NPV is $7.2 million. A alone plus D is infeasible, while A alone produces NPV of $7.0 million but fails the low-carbon requirement. Thus, under the stated policy constraint, B+D is the best feasible portfolio among these combinations.

Outcome and interpretation

The result is not a claim that project A is unattractive. A has the highest individual NPV, but it consumes scarce capital and engineering capacity and cannot coexist with the mandatory retrofit under the modeled limits. The recommendation should be presented with a sensitivity analysis: if the board relaxes the low-carbon requirement, A becomes a strong candidate; if year-1 capital increases by at least $2 million and engineering capacity is available, A+D may become feasible; if the retrofit can be phased, the optimal portfolio may change.

Learning points

1. Prerequisites must be modeled explicitly.
2. Strategic requirements can materially alter financial rankings.
3. Feasibility must be tested period by period.
4. A mathematically optimal answer is conditional on assumptions.
5. Management should examine the value of relaxing binding constraints.

9. Governance, Ethics, and Technology

Capital allocation models influence employment, safety, environmental performance, and regional development. Governance requires clear ownership of assumptions, documented model versions, independent review, and approval thresholds. Projects should not be made artificially attractive by optimistic revenue forecasts, understated implementation costs, or selective treatment of risk.

Ethically, decision-makers should consider worker safety, environmental externalities, community effects, data privacy, and distributional consequences. A project with high NPV may be unacceptable if it violates law, safety standards, or the firm’s fiduciary and social responsibilities.

Optimization software can solve large integer programs, but software does not validate the economic model. Artificial intelligence can assist with forecasting, scenario generation, document review, and detection of anomalous assumptions. It should not replace managerial accountability. AI-generated forecasts require controls for data bias, explainability, confidentiality, model drift, and human approval. An auditable record should show the data, assumptions, transformations, constraints, solver settings, and overrides used in the final recommendation.

10. Best-Practice Decision Process

A disciplined portfolio process typically follows these steps:

1. Define strategic objectives and non-negotiable constraints.
2. Estimate incremental after-tax cash flows and risk-adjusted discount rates.
3. Calculate NPV and supporting measures such as PI, payback, and risk indicators.
4. Identify indivisibilities, dependencies, synergies, cannibalization, and timing.
5. Build a linear, integer, or mixed-integer model appropriate to the decisions.
6. Validate data, units, cash-flow timing, and logical constraints.
7. Solve the model and inspect feasibility, shadow prices, and sensitivity.
8. Stress-test the portfolio under adverse scenarios.
9. Apply strategic, ethical, regulatory, and resilience judgments not captured numerically.
10. Establish monitoring triggers and a process for reallocating capital as information changes.

Challenges and Limitations

Portfolio optimization is powerful but not omniscient. NPVs depend on forecasts, discount rates, tax assumptions, inflation treatment, and terminal values. Correlations can be unstable, particularly during crises. Synergies may be double-counted, and operational capacities may be measured inaccurately. Integer models can become computationally difficult as the number of projects and logical relationships grows.

Optimization also creates a risk of false precision. A portfolio with an objective value of $42.6 million is not necessarily meaningfully better than one valued at $42.2 million when forecast error is large. Qualitative judgment remains essential for strategic positioning, organizational capability, reputation, and unknown risks. The correct use of a model is to improve disciplined judgment, not to eliminate it.

Chapter Summary

Capital rationing arises when financial or non-financial resources limit the projects a firm can undertake. NPV measures absolute value creation, while the profitability index measures value per unit of investment and is most useful under a simple, single-resource constraint. It can mislead when projects are indivisible, mutually exclusive, interdependent, or subject to multiple constraints.

Binary and mixed-integer optimization models represent project selection explicitly. Financial budgets, engineering capacity, timing, prerequisites, mutual exclusivity, strategic allocations, and minimum requirements can be modeled as constraints. Portfolio risk should reflect covariance, common exposures, and adverse scenarios rather than project-by-project risk alone. Interactions, real options, sequencing, and capacity timing can materially change the optimal portfolio.

Optimization outputs require sensitivity analysis, shadow-price interpretation, governance, and ethical review. The strongest capital-allocation process combines rigorous valuation with strategic judgment, transparent assumptions, and continuous monitoring.

Key Terms

Capital rationing: Limitation on the resources available for investment.

Hard rationing: Resource limitation imposed by external financing or regulatory conditions.

Soft rationing: Internally imposed investment or resource limit.

Net present value (NPV): Present value of incremental future cash flows minus initial investment.

Profitability index (PI): Present value of future inflows divided by initial investment.

Binary variable: Decision variable restricted to zero or one.

Integer programming: Optimization with whole-number decision variables.

Linear programming: Optimization with linear objective and constraint relationships.

Mixed-integer programming: Model containing both continuous and integer variables.

Mutually exclusive projects: Alternatives for which no more than one may be selected.

Prerequisite: Project or stage that must be selected before another can proceed.

Synergy: Additional value created when projects are undertaken together.

Cannibalization: Reduction in one project’s cash flows caused by another project.

Covariance: Measure of how two project outcomes vary together.

Correlation: Standardized measure of co-movement between project outcomes.

Shadow price: Marginal value of relaxing a binding constraint.

Capacity constraint: Limit on a scarce operational resource.

Real option: Value of managerial flexibility to delay, expand, abandon, or switch an investment.

Scenario analysis: Evaluation of outcomes under alternative states of the world.

Portfolio resilience: Ability of an investment portfolio to withstand adverse conditions.

Review Questions

Short Answer Questions

1. Define hard and soft capital rationing.
2. Why can a project with positive NPV be excluded from an optimal portfolio?
3. State the profitability index and explain its principal limitation.
4. What is the purpose of a binary decision variable?
5. Explain the meaning of a shadow price in constrained optimization.

Long Answer Questions

1. Compare NPV ranking and profitability-index ranking under a single capital constraint. Under what conditions can PI fail?
2. Explain how financial, operational, strategic, and timing constraints can be incorporated into an integer programming model.
3. Discuss the role of covariance and correlation in project portfolio selection.
4. Explain how prerequisites, mutual exclusivity, synergies, and cannibalization affect portfolio valuation.
5. Evaluate the governance and ethical risks of relying on optimization software or AI in capital allocation.

Analytical / Application Questions

1. A firm has four indivisible projects with different NPVs, investments, and engineering requirements. Formulate a binary optimization model and identify the constraints needed to represent a prerequisite and two mutually exclusive projects.
2. Two projects have expected returns of 12% and 9%, standard deviations of 20% and 10%, and correlation of −0.25. Explain how correlation affects portfolio risk and identify additional information required before selecting the projects.
3. A project portfolio is feasible in total over three years but exceeds the engineering capacity in year 1. Design a sequencing approach and explain how the value of delay, information, and capacity expansion should be evaluated.

Multiple Choice Questions with Answers

Question 1

Which measure represents absolute value added by an investment project?

A. Profitability index

B. Net present value

C. Payback period

D. Accounting return

Correct Answer: B. Net present value

Question 2

A binary project-selection variable can take which values?

A. Any positive real number

B. Any value between −1 and 1

C. Only 0 or 1

D. Only values greater than 1

Correct Answer: C. Only 0 or 1

Question 3

A prerequisite relationship in a project model is correctly represented by which constraint?

A. x_B ≥ x_A

B. x_A + x_B ≤ 0

C. x_B ≤ x_A

D. x_A − x_B ≥ 2

Correct Answer: C. x_B ≤ x_A

Question 4

The profitability index is most directly interpreted as:

A. Value created per unit of initial investment

B. Total accounting profit divided by sales

C. The probability of project failure

D. The project’s standard deviation

Correct Answer: A. Value created per unit of initial investment

Question 5

If two projects have negative cash-flow correlation, including both may:

A. Increase diversification benefits

B. Guarantee positive NPV

C. Eliminate all project risk

D. Make their NPVs additive only when they are mutually exclusive

Correct Answer: A. Increase diversification benefits

Question 6

Which situation generally requires integer rather than continuous decision variables?

A. A project can be scaled smoothly by capacity

B. A facility must be built as a whole unit

C. A budget is measured in dollars

D. A discount rate is expressed annually

Correct Answer: B. A facility must be built as a whole unit

Question 7

If projects A and B are mutually exclusive, which constraint is appropriate?

A. x_A + x_B ≤ 1

B. x_A − x_B ≥ 1

C. x_A + x_B ≥ 2

D. x_Ax_B ≥ 1

Correct Answer: A. x_A + x_B ≤ 1

Question 8

A shadow price estimates:

A. The historical share price of the firm

B. The marginal objective improvement from relaxing a constraint

C. The tax rate on project cash flows

D. The probability of selecting a project

Correct Answer: B. The marginal objective improvement from relaxing a constraint

Question 9

If project B creates synergy with project A, the combined valuation should include:

A. Only the lower individual NPV

B. An interaction value in addition to separate project values

C. No cash flows from project B

D. The discount rate multiplied by both NPVs

Correct Answer: B. An interaction value in addition to separate project values

Question 10

Which statement about optimization results is most appropriate?

A. The solver removes the need for managerial judgment

B. The optimal solution is valid regardless of assumptions

C. Sensitivity analysis is needed because forecasts and constraints are uncertain

D. Positive NPV automatically means selection

Correct Answer: C. Sensitivity analysis is needed because forecasts and constraints are uncertain