Natural Borrowing Limit: Theory & Implications

• The natural borrowing limit is defined as the maximum sustainable debt that can be repaid with certainty through the discounted present value of future income under no-Ponzi conditions.
• It contrasts with ad hoc liquidity, collateral, and interest-coverage constraints by emphasizing endogenous solvency bounds that vary with income risk and time horizons.
• Models incorporating natural borrowing limits reveal sensitivity to labor income shocks, equilibrium interest rates, and lifecycle factors, thereby influencing optimal consumption and policy design.

Searching arXiv for papers on “natural borrowing limit” and closely related borrowing-constraint terminology. Natural borrowing limit usually denotes the loosest lower bound on assets compatible with repayment under incomplete markets. In the standard household interpretation, it is the most debt an agent can hold while still being able to repay with certainty given the worst possible future income path; it is an intertemporal solvency bound arising from no-Ponzi logic and exogenous income risk, rather than a lender-imposed collateral or covenant limit. Recent work sharpens this distinction by showing that several objects often described as borrowing constraints—zero-asset liquidity constraints, collateral or leverage limits, interest-coverage covenants, and other endogenous debt ceilings—are conceptually different even when they also restrict borrowing (Chen et al., 5 Sep 2025, Roulleau-Pasdeloup, 5 Nov 2025, Camara et al., 2022).

1. Conceptual definition and boundaries of the term

The natural borrowing limit is best understood as a feasibility object. It is the debt level implied by future repayment capacity when the agent can trade only limited assets and must satisfy terminal or transversality-type solvency conditions. In this sense, it differs from exogenous policy bounds, from lender-imposed debt covenants, and from collateral rules tied to asset liquidation values.

Object	Representative formula	Status relative to the natural borrowing limit
Natural borrowing limit	$$w_t \ge -\int_t^L \exp\!\left(-\int_t^s r_u\,du\right) E[\eta_s \mid F_t]\,ds$$	Endogenous solvency bound
No-borrowing / liquidity constraint	$a(t)\ge 0$ or $0\le c(t)\le a(t)$	Constant lower bound
Collateral / leverage constraint	$B_t \le \theta^{LEV} p_{k,t} k_t$	Asset-value-based lender limit
Interest-coverage constraint	$\tilde r_t B_t^{IC} \le \theta^{IC}\varphi(L)\pi_t$	Cash-flow and interest-sensitive lender limit

A recurring misconception is to treat all lower bounds on wealth or debt as natural borrowing limits. The literature surveyed here rejects that equivalence. A constant floor such as $a(t)\ge 0$ is an ad hoc no-borrowing restriction; a collateral constraint is tied to pledgeable assets; an interest-coverage rule is tied to debt service out of cash flow. Only the first row is a natural borrowing limit in the canonical incomplete-markets sense (Chen et al., 5 Sep 2025, Roulleau-Pasdeloup, 5 Nov 2025, Camara et al., 2022).

2. Solvency logic in incomplete-markets models

A recent continuous-time OLG formulation makes the solvency logic explicit. Households face idiosyncratic labor-income risk, can trade only a single risk-free bond, and choose consumption subject to wealth dynamics

$$w_t^b(r)=w_b^b+\int_b^t r_s w_s^b\,ds+\int_b^t \eta_s^b\,ds-\int_b^t c_s^b\,ds.$$

With CRRA utility defined as $-\infty$ for negative terminal wealth, any optimal policy must satisfy $w_L\ge 0$ almost surely. Because consumption is also nonnegative, the most debt the household can carry is the amount that could still be repaid if future consumption were driven to its minimum feasible level. This yields the paper’s life-cycle lower bound

$$w_t \ge \underline w_t \coloneqq -\int_t^L \exp\!\left(-\int_t^s r_u\,du\right)E[\eta_s\mid F_t]\,ds,\qquad P\text{-a.s.},$$

and, cohort by cohort in the OLG model,

$a(t)\ge 0$0

The authors summarize this as the “discounted expected shortfall of future income,” although the actual formula is the negative present value of conditional expected future income (Chen et al., 5 Sep 2025).

The derivation is a finite-horizon analogue of no-Ponzi logic. Using variation of constants,

$a(t)\ge 0$1

Rearranging under $a(t)\ge 0$2 and $a(t)\ge 0$3 gives the lower bound. A crucial technical nuance is that this result is stated as a lower bound for the natural borrowing limit, not always the exact tight limit, because the argument uses Jensen’s inequality. The formula nonetheless captures the canonical content of the concept: debt capacity is backed by future labor income rather than by collateral liquidation or an externally imposed cap (Chen et al., 5 Sep 2025).

3. State dependence, age dependence, and equilibrium dependence

In this continuous-time OLG setting, the natural borrowing limit is not constant. It is time-varying, state-dependent, age-dependent, cohort-dependent, and heterogeneous across agents. The dependence on $a(t)\ge 0$4 means that current and past income shocks affect the conditional expectation $a(t)\ge 0$5, hence current debt capacity. The dependence on $a(t)\ge 0$6 means that the remaining horizon matters directly: as the household ages and human wealth shrinks, the bound typically tightens. At the terminal date, the limit collapses to zero (Chen et al., 5 Sep 2025).

The interest rate path enters only through discounting. Higher future rates reduce the present value of future income and make the borrowing limit less negative; lower rates do the opposite. Because the equilibrium interest rate is determined by market clearing,

$a(t)\ge 0$7

the borrowing limit is endogenous twice: through the household’s income process and through the equilibrium rate path generated by demographics, aggregate savings, and capital supply. In that sense, the natural borrowing limit is not merely individual but equilibrium-dependent (Chen et al., 5 Sep 2025).

A concrete illustration is the geometric-Brownian-income example. If

$a(t)\ge 0$8

then

$a(t)\ge 0$9

and for constant $0\le c(t)\le a(t)$0,

$0\le c(t)\le a(t)$1

This makes the state dependence transparent: higher current income loosens the borrowing limit, low-income states tighten it, and the bound moves toward zero over the life cycle (Chen et al., 5 Sep 2025).

4. Ad hoc liquidity constraints and boundary behavior

A second line of work is useful mainly by contrast. In a continuous-time income-fluctuation problem with constant income $0\le c(t)\le a(t)$2, the exact borrowing restriction is

$0\le c(t)\le a(t)$3

treated in the appendix as equivalent to

$0\le c(t)\le a(t)$4

This is a nonnegative-asset or no-borrowing constraint, not a natural borrowing limit. It is a constant lower bound, not state-dependent, not income-dependent, and not endogenously derived from the present discounted value of future labor income. The paper does not discuss the natural borrowing limit by name, and does not derive the usual present-value formula (Roulleau-Pasdeloup, 5 Nov 2025).

The distinction matters operationally. The paper derives the integrated budget identity

$0\le c(t)\le a(t)$5

and this is the closest thing it contains to a solvency or no-Ponzi statement. But the inequality follows from the imposed nonnegativity of assets, not from an endogenous derivation of the loosest feasible debt floor. A plausible implication is that, with constant income $0\le c(t)\le a(t)$6 and $0\le c(t)\le a(t)$7, the usual natural borrowing limit in that environment would be

$0\le c(t)\le a(t)$8

whereas the paper imposes the tighter bound $0\le c(t)\le a(t)$9. For $B_t \le \theta^{LEV} p_{k,t} k_t$0, the standard infinite-horizon present-value formula is not finite, which further separates the model from the canonical natural-borrowing-limit framework (Roulleau-Pasdeloup, 5 Nov 2025).

The paper is nevertheless informative about boundary behavior under a hard lower asset bound. If assets are exhausted at time $B_t \le \theta^{LEV} p_{k,t} k_t$1, optimal consumption is

$B_t \le \theta^{LEV} p_{k,t} k_t$2

with $B_t \le \theta^{LEV} p_{k,t} k_t$3. At the boundary, $B_t \le \theta^{LEV} p_{k,t} k_t$4, while for $B_t \le \theta^{LEV} p_{k,t} k_t$5 the marginal propensity to consume out of assets diverges as $B_t \le \theta^{LEV} p_{k,t} k_t$6. That pattern is characteristic of an ad hoc liquidity constraint, not of a derived natural borrowing limit (Roulleau-Pasdeloup, 5 Nov 2025).

5. Distinction from collateral, leverage, and cash-flow borrowing limits

Much of macro-finance uses “borrowing constraint” language for objects that are not natural borrowing limits. In emerging-market corporate finance, the benchmark collateral constraint is

$B_t \le \theta^{LEV} p_{k,t} k_t$7

or in the DSGE formulation

$B_t \le \theta^{LEV} p_{k,t} k_t$8

The same paper compares this with cash-flow constraints

$B_t \le \theta^{LEV} p_{k,t} k_t$9

and argues that the empirically relevant limit for Argentine firms is closer to interest coverage than to collateral. Using credit-registry data from 1998–2020, only about $\tilde r_t B_t^{IC} \le \theta^{IC}\varphi(L)\pi_t$0 of firm debt is collateral-based, while $\tilde r_t B_t^{IC} \le \theta^{IC}\varphi(L)\pi_t$1 is cash-flow-based. This is analytically important, but it is conceptually different from the classic household natural borrowing limit: the operative bound is lender-imposed, endogenous to cash flow or asset values, and directly interest-sensitive (Camara et al., 2022).

A related distinction appears in robust portfolio-choice models with different borrowing and lending rates. There the wealth process contains the financing term

$\tilde r_t B_t^{IC} \le \theta^{IC}\varphi(L)\pi_t$2

borrowing occurs when $\tilde r_t B_t^{IC} \le \theta^{IC}\varphi(L)\pi_t$3, and the admissible set imposes an exogenous leverage cap $\tilde r_t B_t^{IC} \le \theta^{IC}\varphi(L)\pi_t$4. The resulting “borrowing limit” is indirect: a higher borrowing rate $\tilde r_t B_t^{IC} \le \theta^{IC}\varphi(L)\pi_t$5 makes leverage less attractive, and the upper portfolio bound $\tilde r_t B_t^{IC} \le \theta^{IC}\varphi(L)\pi_t$6 caps debt-financed risky investment. Again, this is not a natural borrowing limit derived from future endowments and repayment feasibility (Yang et al., 2017).

The same holds in insurance models with portfolio constraints

$\tilde r_t B_t^{IC} \le \theta^{IC}\varphi(L)\pi_t$7

If $\tilde r_t B_t^{IC} \le \theta^{IC}\varphi(L)\pi_t$8, borrowing to invest is allowed, but the amount borrowed cannot exceed $\tilde r_t B_t^{IC} \le \theta^{IC}\varphi(L)\pi_t$9 times current surplus. This is a wealth-proportional leverage cap, introduced as an admissibility or regulatory restriction. It behaves like a state-dependent debt ceiling because the maximum dollar position scales with current surplus, but it is not derived from no-Ponzi logic or from the present value of future premia net of claims (Belkina et al., 2011).

6. Endogenous analogues and terminological extensions

Outside the canonical household setting, “natural borrowing limit” is sometimes used more loosely to denote an endogenous debt ceiling generated by optimization or market microstructure. With proportional transaction costs, for example, a finite maximum sustainable leverage emerges because rebalancing costs rise with leverage. In the risk-neutral limit, the paper identifies the asymptotic maximum multiple

$a(t)\ge 0$0

This is “natural” in the sense that it is not imposed by regulation; it arises endogenously because sufficiently high leverage lowers long-run performance. But it is not the household natural borrowing limit of incomplete-markets theory (Guasoni et al., 2015).

DeFi lending provides another analogue. There the effective debt ceiling is determined not only by collateral valuation but also by liquidation incentives, close factors, and market depth. The key non-toxicity condition is

$a(t)\ge 0$1

and the paper argues that positions can be formally overcollateralized according to oracle prices yet economically unliquidatable in a safe way. The resulting debt ceiling is endogenous to liquidation technology and liquidity, not to future labor income or no-Ponzi logic (Warmuz et al., 2022).

A further extension appears in unsecured lending via delegated underwriting. The operative account-level limit is

$a(t)\ge 0$2

with aggregate conservation

$a(t)\ge 0$3

Here borrowing power is backed by delegated and earned credit capacity in a sponsor forest. This too is an endogenous borrowing ceiling, but its backing resource is network credit state rather than discounted future labor income (Estevez, 5 May 2026).

These extensions are analytically valuable, but they should not be conflated with the canonical natural borrowing limit. The former are endogenous ceilings generated by frictions, protocol rules, or delegated backing; the latter is the loosest solvency-compatible lower bound on assets in an incomplete-markets consumption-savings environment. Precision in terminology is therefore substantive rather than cosmetic: it determines whether the relevant object is a present-value solvency bound, an ad hoc liquidity floor, a lender covenant, a leverage cap, or an endogenous market-design limit.