Innes Moral-Hazard Debt Model

The Innes moral-hazard debt model is a foundational theory of financial contracting and corporate capital structure in business and economics theory, developed by Robert D. Innes in his 1990 Journal of Economic Theory article, "Limited Liability and Incentive Contracting with Ex-Ante Action Choices." Whereas the Townsend costly state verification model (Townsend, 1979; Gale and Hellwig, 1985) and the Hart–Moore incomplete-contracts debt model derive debt as a mechanism to minimize ex-post auditing costs or deter cash-flow diversion when output is privately observed, Innes analyzed the complementary polar benchmark: verifiable output q combined with unobserved ex-ante entrepreneurial effort e (moral hazard) and bilateral limited liability (0 ≤ R(q) ≤ q). Innes established a foundational duality result: if the investor's repayment schedule R(q) is only required to satisfy bilateral limited liability (0 ≤ R(q) ≤ q), the optimal contract under the Monotone Likelihood Ratio Property (MLRP) is a discontinuous "live-or-die" step contract that awards the investor 100% of output below a threshold q^* and 0% above q^*. However, because any declining segment in R(q) invites the entrepreneur to secretly borrow on the side to artificially inflate reported output above q^* (or invites the investor to sabotage output down to q^*), imposing the monotonicity constraint (R′(q) ≥ 0, or both parties' payoffs non-decreasing in q) uniquely selects the standard debt contract R(q) = min(q, D), making the entrepreneur the 100% residual claimant on all marginal output above face value D.

The Jensen–Meckling Outside Equity Distortion vs. Debt Overhang

In their classic 1976 paper on agency costs, Michael C. Jensen and William H. Meckling observed that when an entrepreneur raises external funds by issuing outside proportional equity (selling a fraction 1 − α ∈ (0, 1) of the firm's cash flows q to outside investors while retaining inside equity share α), the entrepreneur captures only α < 1 of the marginal return to costly effort e, causing underinvestment in effort relative to the first-best level e^FB. Yet Jensen and Meckling restricted attention to predetermined combinations of linear outside equity and risky debt rather than solving for the optimal nonlinear security R(q) over all feasible sharing rules.

In the canonical principal–agent problem (Holmström, 1979), if both the investor and the entrepreneur are risk-neutral, the first-best effort level e^FB can be achieved if the entrepreneur has unlimited wealth by paying a fixed lump sum D to the investor across all states (R(q) ≡ D) and retaining 100% of residual output q − D. In practice, however, entrepreneurs are wealth-constrained and protected by limited liability: whenever realized output q falls below D, repayment R(q) cannot exceed q (R(q) ≤ q).

Production Technology, Effort, and Probability Distributions

Consider a risk-neutral entrepreneur with zero initial wealth (W₀ = 0) who possesses an investment project requiring a capital outlay I > 0 at t = 0:

• Unobserved Effort and Verifiable Output: The entrepreneur chooses an unobservable effort level e ∈ [0, ē] at private disutility cost v(e) (v(0) = 0, v′(0) = 0, v′(e) > 0, v″(e) > 0). At t = 1, the project generates a court-verifiable cash flow q ∈ [0, Q] drawn from distribution F(q | e) with density f(q | e) > 0.
• Monotone Likelihood Ratio Property (MLRP): Higher effort e shifts the output distribution upward according to the Monotone Likelihood Ratio Property:

where f_e(q | e) ≡ ∂f(q | e) / ∂e. MLRP implies First-Order Stochastic Dominance (F_e(q | e) < 0 for q ∈ (0, Q)), with a single interior crossing point q₀(e) ∈ (0, Q) where f_e(q | e) changes sign from negative to positive. In the first-best benchmark, social surplus is maximized at e^FB where ∫₀^Q q f_e(q | e^FB) dq = v′(e^FB).

Feasible Contracts Under Bilateral Limited Liability

The entrepreneur and a competitive risk-neutral investor sign a contract specifying the investor's repayment schedule R(q) and the entrepreneur's residual payoff Y(q) ≡ q − R(q), subject to Bilateral Limited Liability (LL) (0 ≤ R(q) ≤ q for all q ∈ [0, Q]). Under the First-Order Approach, the entrepreneur's Incentive Compatibility (IC) constraint is:

Whenever ∫₀^Q R(q) f_e(q | e) dq > 0, the investor captures part of the marginal output gains from effort, causing underinvestment (e < e^FB). Maximizing effort incentives is therefore equivalent to choosing R(q) to minimize ∫₀^Q R(q) f_e(q | e) dq subject to the investor's break-even constraint (IR_I): ∫₀^Q R(q) f(q | e) dq ≥ I.

Pointwise Variational Derivation of the Live-or-Die Contract

Suppose we impose only bilateral limited liability, 0 ≤ R(q) ≤ q, without requiring R(q) to be monotonic in q. Forming the Lagrangian for minimizing ∫₀^Q R(q) f_e(q | e) dq subject to ∫₀^Q R(q) f(q | e) dq ≥ I (multiplier γ > 0) gives:

Pointwise maximization with respect to R(q) ∈ [0, q] yields a bang-bang solution governed by the likelihood ratio LR(q | e) ≡ f_e(q | e) / f(q | e): R(q) = q when LR(q | e) < γ, and R(q) = 0 when LR(q | e) > γ. By MLRP, LR(q | e) crosses γ at a single threshold q^* ∈ (0, Q), yielding Innes's Unconstrained "Live-or-Die" Step Contract:

Why Live-or-Die Contracts Fail: Hidden Borrowing and Output Destruction

Any contract where R(q) decreases in q suffers from two implementation flaws:

1. Entrepreneurial Hidden Borrowing / Window-Dressing Arbitrage: If realized output is q = q^* − ε (for small ε > 0), the entrepreneur must turn over q^* − ε to the investor under R^LD(q). If the entrepreneur secretly borrows ε from a third party and reports total cash flow q̂ = q + ε = q^*, the investor's claim under R^LD(q^*) drops discontinuously from q^* − ε to 0, allowing the entrepreneur to repay ε and pocket q^* − ε > 0. Preventing secret output-padding arbitrage requires Y(q + ε) − ε ≤ Y(q), or non-decreasing investor repayments: R′(q) ≥ 0.
2. Investor Sabotage / Free Disposal: Conversely, if q = q^* + ε, the investor gets 0 under R^LD(q) instead of q^* − ε > 0 at q^* − ε. Requiring both parties' payoffs to be non-decreasing in q eliminates any incentive to destroy output.

To make the financial contract immune to window-dressing arbitrage and output destruction, Innes imposed the Bilateral Monotonicity Constraint (MC): both the investor's repayment R(q) and the entrepreneur's residual payoff Y(q) = q − R(q) must be non-decreasing in q, which means:

A Standard Debt Contract with face value D ∈ (0, Q) is defined by R_D(q) = min(q, D), whose slope is R_D′(q) = 1 for q < D (default region) and R_D′(q) = 0 for q > D (solvency region).

Single-Crossing Proof That Standard Debt Maximizes Effort Incentives

Theorem (Innes's Moral-Hazard Debt Optimality Theorem, 1990): Let R(q) be any feasible contract satisfying bilateral limited liability (0 ≤ R(q) ≤ q) and bilateral monotonicity (0 ≤ R′(q) ≤ 1) that gives the investor the same expected payoff as a standard debt contract R_D(q) = min(q, D) at effort e:

If R(q) ≠ R_D(q) on a set of positive measure, then under MLRP, the standard debt contract R_D(q) induces greater entrepreneurial effort incentives than R(q):

Proof via Single-Crossing and MLRP

1. Single-Crossing Property of Debt Minus Any Monotonic Contract: On the default interval q ∈ [0, D], the debt contract takes the maximum feasible amount R_D(q) = q, so R(q) ≤ q = R_D(q) for all q ≤ D. On the solvency interval q ∈ [D, Q], the debt contract is flat (R_D(q) = D), whereas R(q) is non-decreasing (R′(q) ≥ 0). Therefore, once R(q) rises above D at some crossing point q̂ ≥ D, R(q) remains ≥ D = R_D(q) for all q > q̂. Thus, Δ(q) ≡ R_D(q) − R(q) satisfies single-crossing around q̂ ≥ D:

1. Evaluating the Likelihood-Ratio Integral at q̂: Write the difference in the investor's marginal claim on effort using LR(q | e) = f_e(q | e) / f(q | e):

For q < q̂, R_D(q) − R(q) ≥ 0 and LR(q | e) < LR(q̂ | e); for q > q̂, R_D(q) − R(q) ≤ 0 and LR(q | e) > LR(q̂ | e). In both regions, the integrand is bounded above by [R_D(q) − R(q)] LR(q̂ | e), yielding:

Equivalently, integrating by parts gives ∫₀^Q [q − R(q)]f_e(q | e)*dq* = −∫₀^Q [1 − R′(q)]F_e(q | e)*dq*. Standard debt front-loads 100% of the investor's repayment into the lowest output states (q ≤ D) at the maximum monotonic slope (R′ = 1), freeing the entrepreneur to be the 100% residual claimant (R′ = 0) across all solvent states (q > D).

• Dominance of Debt Over Outside Equity: Under any proportional equity contract R_eq(q) = (1 − α)q that raises the same expected revenue I as debt R_D(q) = min(q, D), the equity investor takes less than q in low states (q < D) and must therefore take a positive share (1 − α)q > D in high states (q > D/(1 − α)). By shifting investor repayments from low-output states (where f_e / f is low or negative) into high-output states (where f_e / f is high and positive), outside equity dulls the entrepreneur's incentive to exert effort relative to standard debt.
• Residual Debt Overhang (e^* < e^FB): Under the optimal debt contract R_D(q) = min(q, D), the entrepreneur's first-order condition is v′(e^*) = −∫_D^Q F_e(q | e^*) dq < −∫₀^Q F_e(q | e^*) dq. Because −F_e(q | e) > 0 on the bankruptcy interval q ∈ (0, D), the entrepreneur ignores the marginal social benefit of effort that accrues to bondholders in bankruptcy states—the classic Myers (1977) debt overhang effect. Innes's theorem shows that while standard debt suffers from debt overhang in default states (q < D), every other monotonic limited-liability contract suffers from an even larger effort distortion.
• Risk-Shifting and Convertible Securities: When the entrepreneur simultaneously controls both effort e (first-order stochastic dominance) and variance/risk-taking σ² (mean-preserving spreads), the call-option convexity of pure debt (Y_D(q) = max(q − D, 0)) creates an asset-substitution incentive (Jensen and Meckling, 1976). Green (1984) and Biais and Casamatta (1999) showed that combining debt with convertible debt or warrants tempers upper-tail convexity so the entrepreneur is deterred from variance-increasing gambles while preserving effort incentives.

How does the Innes (1990) derivation of debt differ from Townsend's (1979) costly state verification model?

In the Townsend costly state verification model, there is no ex-ante effort choice; output q is an exogenous random draw that only the entrepreneur observes unless the investor pays an auditing cost c > 0. Debt is optimal in Townsend (1979) because a flat repayment D in solvent states (q ≥ D) avoids paying the auditing cost c whenever the firm is solvent. In Innes (1990), by contrast, output q is publicly observable at zero auditing cost (c = 0), and the friction is ex-ante unobserved effort e (moral hazard). Debt is optimal in Innes (1990) because front-loading repayments into low-output states (R(q) = q for q ≤ D) maximizes the entrepreneur's marginal reward for effort in high-output states (R′(q) = 0 for q > D).

Why is the monotonicity constraint (0 ≤ R′(q) ≤ 1) needed to obtain standard debt in Innes's model?

Without monotonicity, minimizing the investor's claim on high-likelihood-ratio states under 0 ≤ R(q) ≤ q yields the discontinuous live-or-die contract (R(q) = q for q < q^* and R(q) = 0 for q ≥ q^*). However, at the discontinuity q = q^*, the investor's repayment drops vertically from q^* to 0. If the entrepreneur can secretly borrow a small amount Δq on the side and add it to reported output so q + Δq ≥ q^*, the entrepreneur wipes out the entire debt obligation q^* at negligible cost. Requiring R′(q) ≥ 0 eliminates all profit from secret output-padding loans, transforming the live-or-die contract into the standard debt contract R_D(q) = min(q, D).

Why does standard debt still result in underinvestment in effort (e^* < e^FB) in Innes's model?

Because the entrepreneur is protected by limited liability (R(q) ≤ q), the investor cannot collect D when output falls in the default region q < D. In those bankruptcy states, any incremental output generated by higher entrepreneurial effort goes 100% to the creditor (R′(q) = 1) rather than to the entrepreneur (Y′(q) = 0). Because the entrepreneur only internalizes marginal output gains in solvent states (q > D), the entrepreneur's private marginal return to effort is smaller than the social marginal return.

1. Innes, R. D. (1990). "Limited Liability and Incentive Contracting with Ex-Ante Action Choices." Journal of Economic Theory, 52(1), 45–67. https://doi.org/10.1016/0022-0531(90)90066-S90066-S)
2. Jensen, M. C., & Meckling, W. H. (1976). "Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure." Journal of Financial Economics, 3(4), 305–360. https://doi.org/10.1016/0304-405X(76)90026-X90026-X)
3. Myers, S. C. (1977). "Determinants of Corporate Borrowing." Journal of Financial Economics, 5(2), 147–175. https://doi.org/10.1016/0304-405X(77)90015-090015-0)
4. Biais, B., & Casamatta, C. (1999). "Optimal Leverage and Aggregate Investment." Journal of Finance, 54(4), 1291–1323. https://doi.org/10.1111/0022-1082.00147
5. Matthews, S. A. (2001). "Renegotiating Moral Hazard Contracts under Limited Liability and Monotonicity." Journal of Economic Theory, 97(1), 1–29. https://doi.org/10.1006/jeth.2000.2695