An Arbitrarily Precise Global Closed Form Approximation for the Neoclassical Growth Model

Scientific Impact Assessment

Core Contribution

The paper adds a new entry to the small catalogue of closed-form solutions for the Ramsey–Cass–Koopmans (RCK) model. The known tractable cases all share the feature that savings is linear in output (AK technology, full depreciation with log utility, Kurz/Barro–Mankiw/Smith parameter knife-edges), and recent work (Gün Polat & Özer 2021; Köster & Mühe 2025) suggests that under CRRA preferences with Cobb–Douglas technology the set of elementary-function-integrable cases is exhausted. The author's move is to swap CRRA for CARA utility (with a subsistence floor c̄ to prevent negative consumption). The key structural payoff is that the CARA Euler equation, ċ = (r(k)−ρ)/γ, is independent of c, which decouples the system. Using the time-elimination method, the saddle-path savings function satisfies an Abel ODE of the second kind — generically unsolvable — but becomes *separable* when ρ = n, yielding an explicit solution s(k) = (1 + W₀(ν e^{γψ(k)}))/γ in terms of the Lambert W function.

Because ρ = n makes the effective discount rate zero (objective diverges, transversality fails), the paper's second contribution is a continuity/approximation theorem: on compact subsets of the interior state space excluding the kink k̄, sup|s(k;ρ) − s(k;n)| ≤ C(ρ−n), established via smooth dependence of the stable manifold of a hyperbolic saddle on parameters (center-stable manifold construction in an augmented (y,ρ) system, backward flow from an anchor, global inversion). Two economic applications follow: (i) the consumption policy is strictly increasing and strictly *concave* (unlike CRRA, where curvature is ambiguous); (ii) a front-loading/back-loading dichotomy — a unique risk-aversion threshold γ̄ above which convergence is slower near the origin than near the steady state.

Methodological Rigor

The analytical work is careful and more thorough than typical for a short theory note. Lemma 1 establishes the feasibility set and the existence of k_min; the Kuhn–Tucker/costate argument at the subsistence junction (λ = μ − 1 > 0 in the constrained region, continuity of μ at the junction because the constraint is on the control, not the state) is handled correctly and explicitly. Lemma 4 is the strongest part of the appendix: it verifies the non-degeneracy conditions the approximation theorem needs (unique hyperbolic saddle, C^∞ vector field, s ∈ (0,1/γ) via a comparison/uniqueness argument that is honestly flagged as available only at ρ = n, and absence of homoclinic connections). The invocation of Wiggins' theorems is appropriate and the augmented-system trick for joint smoothness in (k,ρ) is standard but correctly executed.

Gaps: (1) Existence and uniqueness of the *optimal* path for ρ > n is asserted rather than proved; the theorem really compares an ODE solution branch, and the link to the planner's optimum for ρ > n rests on the terminal condition s(k*) = 0 being the selected branch. (2) There is no numerical verification of the approximation. A shooting solution of the exact ρ > n system compared against the Lambert-W formula at, say, ρ − n = 0.005 and 0.02 would have been cheap and would have quantified C — the paper's central selling point ("arbitrarily precise") is established only asymptotically, with no indication of the error magnitude at empirically relevant discount rates (ρ − n ≈ 0.02–0.04, which is *not* small relative to n). (3) Proposition 2's threshold depends on an arbitrary reference point k₀; the paper acknowledges this but it weakens the economic interpretation. (4) Minor inconsistency between the main-text and appendix statements of Theorem 1 (k_min ≤ k₁ vs 0 < k₁).

Potential Impact

Realistically limited. The value proposition is pedagogical and benchmarking: the author himself concludes "this approach could be used in teaching." Closed-form saddle paths are attractive for illustrating global transition dynamics that phase diagrams and log-linearization around the steady state cannot deliver, and they have genuine (if niche) value as exact benchmarks for testing global numerical solvers. But the binding constraints on uptake are severe: CARA preferences have no balanced growth path, are non-homothetic, imply absolute rather than relative risk aversion, and are essentially never used in applied growth or business-cycle work; and the result requires ρ ≈ n, a knife-edge with no empirical support. So the front-loading/back-loading taxonomy cannot be read as a statement about real economies' take-off speeds — it is a property of a CARA economy with a near-zero effective discount rate. The paper is best seen as a contribution to the "special functions in economic dynamics" thread (Boucekkine & Ruiz-Tamarit; the author's own JME paper on borrowing constraints), where it will likely accumulate modest citations.

Timeliness & Relevance

Not addressing a live bottleneck. Global solution of RCK-type models is a solved computational problem; contemporary frontier work is on heterogeneous-agent continuous-time models and neural-network solvers, where closed-form benchmarks are welcome but not scarce. The appearance of Köster & Mühe (2025) on the exhaustion of CRRA closed forms does give the paper a topical hook — it shows the classification can be escaped by changing the preference class rather than the technology or parameters — which is a genuinely useful framing.

Strengths & Limitations

Strengths: a clean, self-contained analytical result; the CARA-decoupling insight is elegant and the reason the Abel equation collapses is transparent; the paper is unusually scrupulous about what "closed form" means (Lambert W is non-Liouvillian) and about the failure of transversality at ρ = n, and does the work to convert the degenerate case into a rigorous approximation statement rather than hand-waving; the subsistence-floor treatment adds an economically meaningful take-off phase; fully analytic and hence trivially reproducible.

Limitations: the knife-edge ρ→n⁺ restriction is the central weakness and is not quantitatively probed; CARA preferences are a strong departure from applied practice and eliminate balanced growth; the "arbitrarily precise global" framing in the title overstates what a first-order continuity bound with an unquantified constant delivers; no empirical or numerical content beyond a single illustrative figure; the dynamical-systems machinery invoked is standard, so the technical novelty resides mainly in the ODE manipulation; the economic conclusions (concave consumption function, risk-aversion-dependent convergence speed) largely confirm or restate existing intuitions rather than overturning them.

Overall: a competent, honest, narrowly scoped theory note with pedagogical value and a small expected citation footprint.