← 返回研究洞察 · 论文专区
Working Paper I · 预印本论文 · 系列基石
Published · SSRN Iterated version · v2 JEL · D21 · G31 · O41 · C61

Saturation Decoupling

Finite Firm Value under Unbounded Capital Accumulation
饱和解耦:无界资本积累下的有限公司价值
版本说明 / Version note. 论文 I 已在 SSRN 正式发表。本页展示的是其迭代版本(v2)——在已发表版基础上补充了最优控制推论(Corollary 1)、Amdahl 型微观基础与可识别性判据。若需正式引用,请以 SSRN 发表版为权威版本。This is a pure-theory working paper; all numbers are illustrative calibrations, not empirical results.

Abstract

This paper identifies and characterizes a structural property, hitherto unnoted, in a class of growth models: when a firm's accumulating factor (capital or a knowledge base) may grow without bound, but its value output enters profit only through a bounded saturating map, firm value is finite for every positive discount rate, and this finiteness is independent of the growth rate of accumulation (the level of value still depends on the path). We call this property saturation decoupling. The main theorem delivers three conclusions: (i) value is unconditionally finite, \(V\le \bar{\pi}/\rho<\infty\), even when the state diverges and even at the AK knife-edge or super-exponential growth; (ii) \(\lim_{\rho\to 0}\rho V=\Pi(x_\infty)\), the profit at the state's long-run limit; (iii) \(\lim_{\rho\to\infty}\rho V=\Pi(x(K_0))\) (initial profit). The latter two are the economic incarnations of the final- and initial-value theorems of the Laplace transform, and they furnish a falsifiable characterization that is opposite to standard AK endogenous growth. We generalize the model to the optimal-control problem of a firm that chooses investment and prove that saturation decoupling survives optimization. We provide a heuristic Amdahl-type micro-foundation for saturation, a distinguishability criterion, and a refutable econometric specification.

Keywords: firm value; endogenous growth; saturation; nonconvex dynamics; valuation; Laplace transform; AK model.   JEL: D21, G31, O41, C61.

AI disclosure. The author used an AI-based assistant (Claude, Anthropic) to support drafting, editing, and the symbolic verification of the closed-form results. All theoretical claims, definitions, proofs, and final text are the author's own and were reviewed and verified by the author, who takes full responsibility for the content.

1.Introduction

A class of firms is becoming common: they accumulate fast yet earn thin current profit. Standard discounted-value reasoning suggests that if a productive factor grows without bound, the firm's value should grow without bound too—unless the discount rate exceeds the growth rate. This is the logic of the AK family of endogenous-growth models, in which profit is linear in the state.

This paper isolates a structural feature that breaks that logic. Suppose the accumulating factor \(K_t\) may diverge, but value enters profit only through a bounded map—a coverage rate \(x(K)\in[0,\bar{x}]\) that saturates. Then firm value is finite for every positive discount rate, regardless of how fast \(K\) grows. We call this saturation decoupling: the finiteness of value decouples from the speed of accumulation.

The contribution is not technical depth—the proof rests on elementary properties of the Laplace transform. It is conceptual: value convergence depends on how value depends on the state, not on the speed of accumulation. This pinpoints a clean, falsifiable distinction between two worldviews—"value diverges with accumulation" (AK) and "value converges through saturation" (this paper)—and provides an anchor for valuing modern firms that accumulate faster than they monetize.

2.Model

Consider an infinite-horizon, continuous-time firm. Its state (an accumulating factor, interpretable as capital or a knowledge base) \(K_t\ge 0\) evolves according to

\[ \dot K_t = G(K_t), \qquad K_0 \text{ given,} \]

where \(G:\mathbb{R}_+\to\mathbb{R}\). We allow \(K_t\to\infty\); this is the core case of interest. Instantaneous profit depends on the state only through a bounded map:

\[ \pi_t=\Pi(x(K_t)), \qquad x:\mathbb{R}_+\to[0,\bar{x}], \quad \bar{x}<\infty, \]

where the coverage rate \(x(\cdot)\) is continuous, nondecreasing, and bounded above (typically \(x\le 1\)), and \(\Pi:[0,\bar{x}]\to\mathbb{R}\) is continuous. The boundedness \(x\le\bar{x}\) is the single key assumption: its economic meaning is that "the marginal value of accumulation saturates." Firm value is the present value of profit at discount rate \(\rho>0\):

\[ V(\rho)=\int_0^\infty e^{-\rho t}\,\Pi(x(K_t))\,dt. \]
Remark — Contrast with AK (the axis of the paper)

In standard AK, profit is linear in the state, \(\pi_t=AK_t\), and \(V=\int_0^\infty e^{-\rho t}AK_t\,dt\) diverges with the diverging \(K_t\) unless \(\rho>g\). The profit equation severs profit's dependence on the state at a bounded map: however \(K_t\) diverges, \(x(K_t)\) and \(\pi_t\) stay confined to a bounded interval. This single modeling choice is the entire source of "decoupling."

2.1 · A firm that makes decisions (optimal-control formulation)

So far \(V\) is the present value of profit along a given state path—it answers "what is this path worth," but does not yet let the firm choose the path. A complete theory of firm value should contain a decision-maker. We generalize to an optimal-control problem, and prove that saturation decoupling does not depend on whether the firm optimizes.

Let the firm choose an investment rate \(i_t\ge 0\) as its control, with the state evolving as \(\dot K_t=i_t-\delta K_t\) (\(\delta>0\) depreciation). Investment carries a continuous, nonnegative, convex adjustment cost \(c(i)\ge 0\) (\(c(0)=0\)). Instantaneous net profit is value output minus investment cost, \(\pi_t=\Pi(x(K_t))-c(i_t)\). The firm solves

\[ V^*(\rho)=\sup_{\{i_t\ge 0\}}\int_0^\infty e^{-\rho t}\big[\Pi(x(K_t))-c(i_t)\big]\,dt \quad\text{s.t.}\quad \dot K_t=i_t-\delta K_t. \]

Coverage \(x(\cdot)\) still satisfies \(x\le\bar{x}\). The passive-cash-flow case is the special case with \(i_t\equiv 0\).

3.Main Result: The Saturation-Decoupling Theorem

Theorem 1 — Saturation decoupling

Let \(\bar{\pi}=\sup_{x\in[0,\bar{x}]}\Pi(x)<\infty\) and \(\underline{\pi}=\inf_{x\in[0,\bar{x}]}\Pi(x)>-\infty\). Then for every \(\rho>0\) and every dynamics \(G\) (even if \(K_t\to\infty\)):

(i) Unconditional finiteness. \(\underline{\pi}/\rho\le V(\rho)\le \bar{\pi}/\rho<\infty\); value is finite, and its finiteness is independent of the growth rate of \(G\) (the level of \(V\) still depends on the path \(\{K_t\}\)).

(ii) Low-discount limit. if \(x_\infty:=\lim_{t\to\infty}x(K_t)\) exists, then \(\lim_{\rho\to 0}\rho V(\rho)=\Pi(x_\infty)\). If \(K_t\to\infty\) then \(x_\infty=\bar{x}\); if \(K_t\to K^*\) finite, the limit is \(\Pi(x(K^*))\).

(iii) High-discount limit. \(\lim_{\rho\to\infty}\rho V(\rho)=\Pi(x(K_0))\) (initial profit).

(i) Since \(x(K_t)\in[0,\bar{x}]\) and \(\Pi\) is continuous, \(\Pi(x(K_t))\in[\underline{\pi},\bar{\pi}]\) for all \(t\). Hence \(V(\rho)\le \bar{\pi}\int_0^\infty e^{-\rho t}\,dt=\bar{\pi}/\rho<\infty\), and the lower bound likewise. The argument invokes no growth property of \(K_t\).

(ii) Write \(\phi(t)=\Pi(x(K_t))\). Then \(\rho V(\rho)\) is \(\rho\) times the Laplace transform of \(\phi\). Since \(\phi\) is bounded, the final-value theorem applies, giving \(\lim_{\rho\to 0}\rho V(\rho)=\lim_{t\to\infty}\phi(t)=\Pi(x_\infty)\).

(iii) Since \(\phi\) is bounded and right-continuous at \(t=0\), the initial-value theorem gives \(\lim_{\rho\to\infty}\rho V(\rho)=\Pi(x(K_0))\). ∎

Corollary 1 — Saturation decoupling under optimization

Consider the optimal-control problem with continuous, nonnegative adjustment cost \(c(i)\ge 0\), \(c(0)=0\). Then for every \(\rho>0\), the optimal firm value satisfies \(\underline{\pi}/\rho \le V^*(\rho) \le \bar{\pi}/\rho < \infty.\)

Upper bound: for any feasible control, net profit \(\pi_t=\Pi(x(K_t))-c(i_t)\le\bar{\pi}\) (since \(c\ge 0\)); taking the supremum gives \(V^*\le\bar{\pi}/\rho\). Lower bound: the control \(i_t\equiv 0\) is feasible with \(c(0)=0\), so net profit is \(\ge\underline{\pi}\); its present value is \(\ge\underline{\pi}/\rho\). ∎

Remark — the "passive cash flow vs. firm value" objection

The ceiling \(\bar{\pi}/\rho\) is entirely independent of whether the firm optimizes. A firm that actively invests cannot escape it either: optimization raises the level of value without changing its finiteness or its bound—this is saturation decoupling in the genuine sense of firm value.

4.A Concrete Growth Dynamics

Theorem 1 is almost assumption-free on \(G\). To show the mechanism concretely and mark the boundary with the nonconvex-growth literature, let

\[ G(K)=\eta_0 K^{\theta}x(K)-\delta K,\qquad x(K)=\bar{x}\big(1-e^{-bK}\big), \]

with \(\eta_0>0\), \(\theta\ge 0\), \(\delta>0\), \(b>0\).

Proposition 1 — Regimes and threshold

(i) for \(\theta<1\) all positive steady states are bounded; \(\theta=1\) is the AK knife-edge; \(\theta>1\) is super-exponential. (ii) For \(0<\theta<1\) and parameters in a certain range, \(G\) has exactly two positive roots \(0K_c\) ignites to \(K^*\).

Remark — Honest attribution

The bistability and unstable threshold are the standard result of the convex–concave / nonconvex growth literature since Skiba (1978); we claim no originality for it, using it only to show the saturating structure arises naturally. Our contribution is Theorem 1 on the state. Whichever regime obtains—including \(\theta\ge 1\) with \(K_t\to\infty\)—Theorem 1 still guarantees finite firm value.

Numerical verification

Take \(\Pi(x)=x\), \(\bar{x}=1\), \(K_0=2\), \(b=0.60\) (so \(x(K_0)\approx0.699\)). (i) For \(\theta\in\{0.3,1.0,1.5\}\) and \(\rho\in\{0.05,0.01,0.003\}\), \(V(\rho)\in[\underline{\pi}/\rho,\bar{\pi}/\rho]\) throughout—\(K_t\) diverges in the \(\theta\ge1\) cases yet value stays finite. (ii) On the diverging path \(\theta=1.0\), \(\rho V\) at \(\rho=0.05,0.01,0.003\) equals \(0.955,0.990,0.997\to\Pi(\bar{x})=1\). (iii) On the same path, \(\rho V\) at \(\rho=5,20,80\) equals \(0.708,0.701,0.700\to\Pi(x(K_0))\approx0.699\). All three match the theorem.

5.The Micro-Foundation of Saturation

The entire force of Theorem 1 rests on \(x\le\bar{x}\). Why assume saturation? Split the process converting state \(K\) into delivered value into a part that accelerates with \(K\) (execution: production, computation, replication), share \(p\); and a part that does not (judgment, review, physical-world interfacing), share \(1-p\). If the execution part is accelerated \(s(K)\)-fold (\(s\to\infty\) as \(K\to\infty\)), then by analogy with Amdahl's law (1967),

\[ x(K)\ \propto\ \frac{1}{(1-p)+p/s(K)},\qquad \lim_{K\to\infty}x(K)=\frac{1}{1-p}:=\bar{x}<\infty. \]
Remark — Heuristic observation, Amdahl analogy

As long as final value delivery depends on any stage that cannot accelerate without bound—human judgment, regulatory review, physical time, trust-building—saturation has an organizational source, not merely a mathematical-convenience one. Amdahl's law is classical and not original here; we use it as a heuristic analogy for a continuous value-output rate.

6.Distinguishability from AK: A Falsifiable Criterion

Theorem 2 — Distinguishability

Let \(\rho V(\rho)\) be the observable trajectory in \(\rho\). (This paper) under saturation decoupling, \(\rho V(\rho)\) is bounded over all \(\rho\in(0,\infty)\) within \([\underline{\pi},\bar{\pi}]\), and \(\lim_{\rho\to 0}\rho V=\Pi(x_\infty)<\infty\) exists. (AK) under \(V=AK_0/(\rho-g)\), \(\rho V=\rho AK_0/(\rho-g)\) diverges as \(\rho\to g^+\). The two models are qualitatively opposite in the boundedness of \(\rho V\) as \(\rho\) approaches its infimum.

Remark — Why this matters, and its identification limit

Distinguishability is an asymptotic property—ours bounded as \(\rho\to 0\), AK diverging as \(\rho\to g^+\). Over a realistically finite observable \(\rho\) range far from AK's \(g\), AK's \(\rho V\) may also look flat. The contrast is adjudicable in principle, not trivially separable in every finite sample.

7.Valuation Implications and a Testable Specification

Consider two firms with identical current profit, but one with state \(K\) far from saturation (\(x(K)\ll\bar{x}\)) and one near it. A cash-flow-anchored multiple assigns them similar values; but Theorem 1(ii) implies the former's \(\rho V\) will still rise to \(\Pi(\bar{x})\). The difference is precisely the value of the "unrealized saturation room," systematically missed by cash-flow-anchored methods.

Observable proxies

Value \(V\to\) enterprise value EV; discount rate \(\rho\to\) WACC; so \(\rho V\to \mathrm{EV}\times\mathrm{WACC}\); \(\Pi(\bar{x})\to\) long-run steady-state operating profit after saturation.

Testable predictions

Prediction 1. In a cross-section of accumulation-type firms, as WACC \(\downarrow\), \(\mathrm{EV}\times\mathrm{WACC}\) tends to a finite ceiling (curve saturates); AK predicts it rises without bound. Use monetary-policy shocks to the risk-free rate as an instrument for \(\rho\). Prediction 2. The EV of high-accumulation, low-current-cash-flow firms should lead their subsequent realized cash flows, with \(\beta>0\) around \(k\approx 3\)–\(5\) years. Falsifiability: if \(\mathrm{EV}\times\mathrm{WACC}\) rises without bound, or accumulation fails to lead cash flow, saturation decoupling is overturned.

8.Conclusion

We identified saturation decoupling: when an accumulating factor is unbounded but value output enters through a bounded saturating map, firm value is finite for every discount rate, and \(\rho V\) converges to initial profit as \(\rho\to\infty\) and to a finite limit \(\Pi(x_\infty)\) as \(\rho\to 0\). We showed it survives at the optimal firm value \(V^*\)—the ceiling \(\bar{\pi}/\rho\) being independent of whether the firm optimizes—lifting the result to a genuine statement about firm value. The property is oppositely signed to AK growth and adjudicable in principle by data. Its value lies in pinpointing that the finiteness of value depends on how value depends on the state, not on the speed of accumulation.

Remark — Limitations and next steps

(a) The final-value limit requires \(\lim\phi(t)\) to exist; persistent oscillation needs a Cesàro-mean generalization. (b) The Amdahl analogy is heuristic. (c) The valuation predictions require panel data—the key next step, and a natural entry point for collaboration.

References

Amdahl, G. M. (1967). Validity of the single processor approach to achieving large scale computing capabilities. AFIPS Conference Proceedings, 483–485.

Azariadis, C., & Stachurski, J. (2005). Poverty traps. In Handbook of Economic Growth, Vol. 1A, 295–384.

Garicano, L. (2000). Hierarchies and the organization of knowledge in production. Journal of Political Economy, 108(5), 874–904.

Rebelo, S. (1991). Long-run policy analysis and long-run growth. Journal of Political Economy, 99(3), 500–521.

Romer, P. M. (1986). Increasing returns and long-run growth. Journal of Political Economy, 94(5), 1002–1037.

Skiba, A. K. (1978). Optimal growth with a convex–concave production function. Econometrica, 46(3), 527–539.

Working Paper I · 预印本论文 · 系列基石
已发表 · SSRN 迭代版 · v2 JEL · D21 · G31 · O41 · C61

饱和解耦

无界资本累积下的有限企业价值
Saturation Decoupling: Finite Firm Value under Unbounded Capital Accumulation
版本说明。 论文 I 已在 SSRN 正式发表。本页展示的是其迭代版本(v2)——在已发表版基础上补充了最优控制推论(推论 1)、Amdahl 型微观基础与可识别性判据。若需正式引用,请以 SSRN 发表版为权威版本。本文为纯理论工作论文;所有数字均为示意性标定,而非实证结果。

摘要

本文识别并刻画一类增长模型中一项此前未被注意的结构性质:当企业的累积要素(资本或知识库)可无界增长,但其价值产出只通过一个有界饱和映射进入利润时,企业价值对每一个正贴现率都有限,且这种有限性独立于累积的增长率(价值的水平仍依赖于路径)。我们称此性质为饱和解耦。主定理给出三个结论:(i) 价值无条件有限,\(V\le\bar{\pi}/\rho<\infty\),即使状态发散、即使处于 AK 临界点或超指数增长;(ii) \(\lim_{\rho\to 0}\rho V=\Pi(x_\infty)\),即状态长期极限处的利润;(iii) \(\lim_{\rho\to\infty}\rho V=\Pi(x(K_0))\)(初始利润)。后两者是拉普拉斯变换终值定理与初值定理的经济学化身,并提供一个与标准 AK 内生增长相反的可证伪刻画。我们将模型推广到选择投资的企业的最优控制问题,并证明饱和解耦在优化下依然成立。我们为饱和提供一个启发式的 Amdahl 型微观基础、一个可区分性判据以及一个可证伪的计量设定。

关键词:企业价值;内生增长;饱和;非凸动态;估值;拉普拉斯变换;AK 模型。  JEL:D21, G31, O41, C61.

AI 披露。 作者使用了基于 AI 的助手(Claude,Anthropic)以辅助撰写、编辑及闭式结果的符号验证。所有理论主张、定义、证明及最终文本均为作者本人,并已由作者审阅与核实,作者对全部内容负责。

1.引言

一类企业正变得普遍:它们累积迅速,却只获得微薄的当前利润。标准的贴现价值推理暗示:若一个生产性要素无界增长,企业价值也应无界增长——除非贴现率超过增长率。这正是 AK 族内生增长模型的逻辑,其中利润对状态是线性的。

本文分离出一个打破该逻辑的结构特征。设累积要素 \(K_t\) 可发散,但价值只通过一个有界映射进入利润——一个饱和的覆盖率 \(x(K)\in[0,\bar{x}]\)。则企业价值对每一个正贴现率都有限,无论 \(K\) 增长多快。我们称之为饱和解耦:价值的有限性与累积速度解耦。

本文的贡献不在技术深度——证明仅依赖拉普拉斯变换的初等性质。它是概念性的:价值的收敛取决于价值如何依赖于状态,而非累积的速度。这精确地标定了两种世界观之间一个干净、可证伪的区分——"价值随累积发散"(AK)与"价值经由饱和收敛"(本文)——并为那些累积快于变现的现代企业的估值提供锚点。

2.模型

考虑一个无限期、连续时间的企业。其状态(一个累积要素,可解释为资本或知识库)\(K_t\ge 0\) 按下式演化:

\[ \dot K_t = G(K_t), \qquad K_0 \text{ 给定,} \]

其中 \(G:\mathbb{R}_+\to\mathbb{R}\)。我们允许 \(K_t\to\infty\);这是核心关注的情形。瞬时利润只通过一个有界映射依赖于状态:

\[ \pi_t=\Pi(x(K_t)), \qquad x:\mathbb{R}_+\to[0,\bar{x}], \quad \bar{x}<\infty, \]

其中覆盖率 \(x(\cdot)\) 连续、非减、上有界(通常 \(x\le 1\)),\(\Pi:[0,\bar{x}]\to\mathbb{R}\) 连续。有界性 \(x\le\bar{x}\) 是唯一的关键假设:其经济含义是"累积的边际价值饱和"。企业价值是利润在贴现率 \(\rho>0\) 下的现值:

\[ V(\rho)=\int_0^\infty e^{-\rho t}\,\Pi(x(K_t))\,dt. \]
评注 — 与 AK 的对照(本文的轴线)

在标准 AK 中,利润对状态是线性的,\(\pi_t=AK_t\),而 \(V=\int_0^\infty e^{-\rho t}AK_t\,dt\) 随发散的 \(K_t\) 发散,除非 \(\rho>g\)。利润式在一个有界映射处切断了利润对状态的依赖:无论 \(K_t\) 如何发散,\(x(K_t)\) 与 \(\pi_t\) 都被限制在一个有界区间内。这一个建模选择就是"解耦"的全部来源。

2.1 · 一个做决策的企业(最优控制表述)

迄今为止,\(V\) 是利润沿给定状态路径的现值——它回答"这条路径值多少",但还不允许企业选择路径。一个完整的企业价值理论应包含一个决策者。我们将模型推广为一个最优控制问题,并证明饱和解耦不依赖于企业是否优化。

设企业选择投资率 \(i_t\ge 0\) 作为其控制,状态按 \(\dot K_t=i_t-\delta K_t\)(\(\delta>0\) 为折旧)演化。投资带有连续、非负、凸的调整成本 \(c(i)\ge 0\)(\(c(0)=0\))。瞬时净利润是价值产出减去投资成本,\(\pi_t=\Pi(x(K_t))-c(i_t)\)。企业求解

\[ V^*(\rho)=\sup_{\{i_t\ge 0\}}\int_0^\infty e^{-\rho t}\big[\Pi(x(K_t))-c(i_t)\big]\,dt \quad\text{s.t.}\quad \dot K_t=i_t-\delta K_t. \]

覆盖率 \(x(\cdot)\) 仍满足 \(x\le\bar{x}\)。被动现金流情形是 \(i_t\equiv 0\) 的特例。

3.主结果:饱和解耦定理

定理 1 — 饱和解耦

设 \(\bar{\pi}=\sup_{x\in[0,\bar{x}]}\Pi(x)<\infty\)、\(\underline{\pi}=\inf_{x\in[0,\bar{x}]}\Pi(x)>-\infty\)。则对每一个 \(\rho>0\) 与每一种动态 \(G\)(即使 \(K_t\to\infty\)):

(i) 无条件有限。 \(\underline{\pi}/\rho\le V(\rho)\le\bar{\pi}/\rho<\infty\);价值有限,且其有限性独立于 \(G\) 的增长率(\(V\) 的水平仍依赖于路径 \(\{K_t\}\))。

(ii) 低贴现极限。 若 \(x_\infty:=\lim_{t\to\infty}x(K_t)\) 存在,则 \(\lim_{\rho\to 0}\rho V(\rho)=\Pi(x_\infty)\)。若 \(K_t\to\infty\) 则 \(x_\infty=\bar{x}\);若 \(K_t\to K^*\) 有限,极限为 \(\Pi(x(K^*))\)。

(iii) 高贴现极限。 \(\lim_{\rho\to\infty}\rho V(\rho)=\Pi(x(K_0))\)(初始利润)。

(i) 由于 \(x(K_t)\in[0,\bar{x}]\) 且 \(\Pi\) 连续,对所有 \(t\) 有 \(\Pi(x(K_t))\in[\underline{\pi},\bar{\pi}]\)。故 \(V(\rho)\le\bar{\pi}\int_0^\infty e^{-\rho t}\,dt=\bar{\pi}/\rho<\infty\),下界同理。该论证不援引 \(K_t\) 的任何增长性质。

(ii) 记 \(\phi(t)=\Pi(x(K_t))\)。则 \(\rho V(\rho)\) 是 \(\phi\) 的拉普拉斯变换的 \(\rho\) 倍。由于 \(\phi\) 有界,终值定理适用,给出 \(\lim_{\rho\to 0}\rho V(\rho)=\lim_{t\to\infty}\phi(t)=\Pi(x_\infty)\)。

(iii) 由于 \(\phi\) 在 \(t=0\) 处有界且右连续,初值定理给出 \(\lim_{\rho\to\infty}\rho V(\rho)=\Pi(x(K_0))\)。∎

推论 1 — 优化下的饱和解耦

考虑最优控制问题,其调整成本 \(c(i)\ge 0\) 连续、非负且 \(c(0)=0\)。则对每一个 \(\rho>0\),最优企业价值满足 \(\underline{\pi}/\rho\le V^*(\rho)\le\bar{\pi}/\rho<\infty.\)

上界:对任何可行控制,净利润 \(\pi_t=\Pi(x(K_t))-c(i_t)\le\bar{\pi}\)(因 \(c\ge 0\));取上确界得 \(V^*\le\bar{\pi}/\rho\)。下界:控制 \(i_t\equiv 0\) 可行,其下 \(c(0)=0\),故净利润 \(\ge\underline{\pi}\);其现值至少为 \(\underline{\pi}/\rho\)。∎

评注 — 这回答了"被动现金流 vs. 企业价值"的质疑

天花板 \(\bar{\pi}/\rho\) 完全独立于企业是否优化。一个主动投资、主动决策的企业同样无法逃脱它:优化抬高价值的水平,而不改变其有限性或其界——这正是企业价值意义下真正的饱和解耦。

4.一个具体的增长动态

定理 1 对 \(G\) 几乎不设假设。为在一个具体的内生增长模型中展示该机制、并标出与非凸增长文献的边界,令

\[ G(K)=\eta_0 K^{\theta}x(K)-\delta K,\qquad x(K)=\bar{x}\big(1-e^{-bK}\big), \]

其中 \(\eta_0>0\)、\(\theta\ge 0\)、\(\delta>0\)、\(b>0\)。

命题 1 — 区制与阈值

(i) 对 \(\theta<1\),所有正稳态有界;\(\theta=1\) 是 AK 临界点;\(\theta>1\) 为超指数。(ii) 对 \(0<\theta<1\) 及某一参数范围,\(G\) 恰有两个正根 \(0K_c\) 点燃至 \(K^*\)。

评注 — 诚实归属

命题 1(ii) 的双稳态与不稳定阈值是 Skiba (1978) 以来凸–凹/非凸增长文献的标准结果;我们不主张其原创性,仅用以表明饱和结构在内生增长模型中自然出现。我们的贡献是关于状态的定理 1。无论哪一区制成立——包括 \(\theta\ge 1\) 且 \(K_t\to\infty\)——定理 1 仍保证企业价值有限。

数值验证

取 \(\Pi(x)=x\)、\(\bar{x}=1\)、\(K_0=2\)、\(b=0.60\)(故 \(x(K_0)\approx0.699\))。(i) 对 \(\theta\in\{0.3,1.0,1.5\}\) 与 \(\rho\in\{0.05,0.01,0.003\}\),\(V(\rho)\) 始终位于 \([\underline{\pi}/\rho,\bar{\pi}/\rho]\) 内——在 \(\theta\ge1\) 情形 \(K_t\) 发散而价值保持有限。(ii) 在发散路径 \(\theta=1.0\) 上,\(\rho V\) 在 \(\rho=0.05,0.01,0.003\) 处等于 \(0.955,0.990,0.997\to\Pi(\bar{x})=1\)。(iii) 在同一路径上,\(\rho V\) 在 \(\rho=5,20,80\) 处等于 \(0.708,0.701,0.700\to\Pi(x(K_0))\approx0.699\)。三者均与定理吻合。

5.饱和的微观基础

定理 1 的全部力量都依赖于 \(x\le\bar{x}\)。为何假设饱和?将"把状态 \(K\) 转化为已交付价值"的过程拆为两部分:一部分随 \(K\) 的累积而加速(执行:生产、计算、复制),占比 \(p\);另一部分不加速(判断、审查、与物理世界对接),占比 \(1-p\)。若执行部分被加速 \(s(K)\) 倍(\(K\to\infty\) 时 \(s\to\infty\)),则类比 Amdahl 定律(1967),

\[ x(K)\ \propto\ \frac{1}{(1-p)+p/s(K)},\qquad \lim_{K\to\infty}x(K)=\frac{1}{1-p}:=\bar{x}<\infty. \]
评注 — 启发式观察,Amdahl 类比

只要最终价值交付依赖于任一无法无界加速的环节——人类判断、监管审查、物理时间、信任建立——饱和就有组织层面的来源,而不仅是数学便利的来源。Amdahl 定律是经典结果,在此非原创;我们将其用作连续价值产出率的启发式类比。

6.与 AK 的可区分性:一个可证伪判据

定理 2 — 可区分性

设 \(\rho V(\rho)\) 为关于 \(\rho\) 的可观测轨迹。(本文)在饱和解耦下,\(\rho V(\rho)\) 在所有 \(\rho\in(0,\infty)\) 上有界于 \([\underline{\pi},\bar{\pi}]\) 内,且 \(\lim_{\rho\to 0}\rho V=\Pi(x_\infty)<\infty\) 存在。(AK)在 \(V=AK_0/(\rho-g)\) 下,\(\rho V=\rho AK_0/(\rho-g)\) 随 \(\rho\to g^+\) 发散。两模型在 \(\rho\) 趋于其下确界时 \(\rho V\) 的有界性上定性相反。

评注 — 为何重要,及其识别极限

可区分性是一个渐近性质——我方在 \(\rho\to 0\) 时有界,AK 在 \(\rho\to g^+\) 时发散。在一个现实有限的、远离 AK 之 \(g\) 的可观测 \(\rho\) 范围内,AK 的 \(\rho V\) 也可能看似平坦。该对照原则上可裁决,而非在每个有限样本中都可轻易分离。

7.估值含义与可检验设定

考虑两家当前利润相同的企业,但一家状态 \(K\) 远离饱和(\(x(K)\ll\bar{x}\)),另一家接近饱和。现金流锚定的利润倍数赋予它们相近的价值;但定理 1(ii) 意味着前者的 \(\rho V\) 仍将上升至 \(\Pi(\bar{x})\)。差额恰是"未实现饱和空间"的价值,被现金流锚定方法系统性遗漏。

可观测代理

价值 \(V\to\) 企业价值 EV;贴现率 \(\rho\to\) WACC;故 \(\rho V\to \mathrm{EV}\times\mathrm{WACC}\);\(\Pi(\bar{x})\to\) 饱和后的长期稳态营业利润。

可检验预测

预测 1. 在累积型企业的横截面中,当 WACC \(\downarrow\) 时,\(\mathrm{EV}\times\mathrm{WACC}\) 趋向一个有限天花板(曲线饱和);AK 预测其无界上升。用对无风险利率的货币政策冲击作为 \(\rho\) 的工具变量。预测 2. 高累积、低当前现金流企业的 EV 应领先其随后实现的现金流,\(\beta>0\),在 \(k\approx 3\)–\(5\) 年附近显著。可证伪性:若 \(\mathrm{EV}\times\mathrm{WACC}\) 无界上升,或累积未能领先现金流,则饱和解耦被推翻。

8.结论

我们识别了饱和解耦:当累积要素无界但价值产出经由一个有界饱和映射进入时,企业价值对每一贴现率都有限,且 \(\rho V\) 在 \(\rho\to\infty\) 时收敛于初始利润、在 \(\rho\to 0\) 时收敛于一个有限极限 \(\Pi(x_\infty)\)。我们表明饱和解耦在最优企业价值 \(V^*\) 处依然成立——天花板 \(\bar{\pi}/\rho\) 独立于企业是否优化——从而将结果提升为关于企业价值的真陈述。该性质与 AK 增长符号相反,原则上可由数据裁决。其价值在精确指出:价值的有限性取决于价值如何依赖于状态,而非累积的速度。

评注 — 局限与下一步

(a) 终值极限要求 \(\lim\phi(t)\) 存在;持续振荡时需 Cesàro 均值推广。(b) Amdahl 类比是启发式的。(c) 第 7 节的预测需以面板数据实证检验——这是将理论锚定于证据的关键下一步,也是合作的自然切入点。

参考文献

Amdahl, G. M. (1967). Validity of the single processor approach to achieving large scale computing capabilities. AFIPS Conference Proceedings, 483–485.

Azariadis, C., & Stachurski, J. (2005). Poverty traps. In Handbook of Economic Growth, Vol. 1A, 295–384.

Garicano, L. (2000). Hierarchies and the organization of knowledge in production. Journal of Political Economy, 108(5), 874–904.

Rebelo, S. (1991). Long-run policy analysis and long-run growth. Journal of Political Economy, 99(3), 500–521.

Romer, P. M. (1986). Increasing returns and long-run growth. Journal of Political Economy, 94(5), 1002–1037.

Skiba, A. K. (1978). Optimal growth with a convex–concave production function. Econometrica, 46(3), 527–539.

链接已复制到剪贴板