GRIT: why welfare-destroying ratchets become irreversible after x*

GRIT proves a threshold: after x*, reversal costs more than the system can spend

The paper's general result is conditional: within its assumptions, any welfare-destroying process with monotonic accumulation, asymmetric reversal cost, and positive feedback becomes irreversible after a critical state x*. The process that creates the accumulated state also depletes or strains the resources needed to reverse it.

• Below x*, reversal is costly but still feasible
• Above x*, R(x, x - x0) > K(x) for every lower target x0 < x
• The theorem unifies biological, institutional, informational, and physical ratchets from the SAPM corpus

The paper tracks one accumulated state x(t), one reversal cost R, and one capacity limit K

GRIT models a ratchet as a state variable x(t) that grows under status quo drift, a reversal cost R(x, Δx), a feedback coefficient γ(x), and a bounded resource budget K(x). The theorem is about state accumulation, not one-time harm.

• Accumulation: dx/dt = f(x,t) + ε(t), where f is systematic drift and ε is noise
• Feedback: γ(x) = ∂f/∂x; when γ > 0, higher x raises future accumulation
• Capacity: K(x) is the maximum resource pool available to reduce x

The same formal x covers nicotine tolerance, debt, market share, pollutant burden, and legal precedent

The paper makes the model concrete by naming measurable state variables. Each example has an accumulated stock that can move past a reversal threshold.

• Biological x: nicotine tolerance in mg/day, tissue concentration of a persistent pollutant in ng/mL
• Institutional x: count of binding legal precedents, fiscal dependency, securitized obligations
• Informational x: platform market share, data volume, algorithmic optimization toward engagement

All three axioms must hold: nonnegative drift, costly unwinding, and self-reinforcing growth

GRIT requires the three axioms jointly. Many systems have one or two of them, but the theorem's irreversibility result requires all three at once.

• A1: E[dx/dt] = E[f(x,t)] ≥ 0, with strict accumulation on at least some interval
• A2: R(x,1) > C(x,1), with ∂R/∂x > 0 and ∂²R/∂x² ≥ 0
• A3: γ(x) = ∂f/∂x > 0, so higher x raises the rate of later accumulation

The paper's examples show why each axiom is not a cosmetic assumption

The axioms map to concrete mechanisms in the paper. The theorem is built from observed ratchet patterns rather than a single-sector analogy.

• A1 example: a smoker's tolerance may vary daily while trending upward; a contaminated aquifer may vary seasonally while accumulating
• A2 example: quitting after twenty years is harder than quitting after one year; displacing a platform at 80% share is harder than at 10%
• A3 example: higher tolerance increases cigarettes smoked per day; more users strengthen network effects; higher tissue burden weakens clearance

When any axiom fails, the paper treats the process as reversible

gives the non-ratchet cases. These are the control group for the theorem: stationary processes, symmetric-cost processes, and negative-feedback processes do not cross an impossibility threshold in the same way.

• No A1: commodity prices fluctuate around long-run equilibrium rather than systematically accumulating
• No A2: warehouse stockpiling can be unwound at the same marginal cost as stocking
• No A3: free-entry markets can self-correct because concentration attracts entry

The proof turns convex reversal cost and bounded capacity into a crossing at x*

The proof constructs total reversal burden as R_total(x) = ∫ from x0 to x of R(s,1) ds. Because marginal reversal cost rises with s, R_total is increasing and convex. Because K(x) is bounded or declining, the curves cross.

• x* is defined by R_total(x*) = K(x*)
• For x > x*, R_total(x) > K(x), so the system cannot fund reversal
• A steeper R lowers x*, stronger γ reaches x* faster, and higher K raises x*

Addiction, debt, and contamination lower x* by consuming the resources used for reversal

The paper's strongest version occurs when K falls as x rises. Then reversal burden rises while reversal capacity shrinks, so the threshold arrives earlier than in the constant-K case.

• Addiction consumes cognitive resources such as executive function and willpower that cessation requires
• Debt service consumes financial resources that principal reduction requires
• Ecosystem contamination degrades microbial metabolism, root filtration, and nutrient cycling used for remediation

Positive feedback makes the prevention window finite: T* = ∫ dx/f(x)

Corollary 5.1 states that under positive feedback, time to x* is finite. The practical point is timing: early intervention at x << x* can change the path; late intervention near x* may slow the path while still reaching the threshold.

• The prevention curve in Figure 2 keeps x(t) below x* only when intervention occurs early
• Strong γ shortens T* because f(x) rises with x
• After x*, the feasible menu shifts from reversal to bypass, containment, and accommodation

Shared capacity makes coal reform harder than any single-ratchet model predicts

5.2 says a compound ratchet crosses its threshold before any individual ratchet would cross alone: x* < min(x* , x* , ..., x* ). The reason is additive burden against a shared resource pool.

• Coal communities combine physical CO2, institutional fiscal dependency, and cultural identity around mining
• Total burden is ΣR_i(x_i), but K is shared across budget, legitimacy, state capacity, and organizational slack
• Sequential reform fails because each ratchet keeps advancing while another is addressed

The paper classifies ratchets by the mechanism that satisfies the same three axioms

GRIT's taxonomy explains why different domains share a formal structure while differing in their concrete pathway to x*.

• Biological: tolerance, tissue burden, species loss, and soil depletion; soil formation can require 500-1,000 years per centimeter
• Institutional: fiscal dependency, legal precedent, organizational scope, and securitization; precedent can require a higher court or constitutional amendment to reverse
• Informational: market share, data volume, and algorithmic optimization; users, data, predictions, and engagement reinforce each other

Counter-ratchets must change accumulation, feedback, or reversal cost before the crossing

The paper rejects post-threshold restoration as the default policy model. Once x* is exceeded, it points to bypass through substitution or acceptance through consequence management.

• Counter-ratchet instruments: external energy input through regulation, regime change, and technological bypass
• The firearms case names Heller in 2008 as the threshold crossing and Bruen in 2022 as later expansion
• Decision Accounting must record x, estimated x*, and the reversal cost curve before the decision is made