THE LIMIT CASE · ECONOMIST ARC — PART 3 · CLIMAX

Pegging the Dollar to the Engine of the Economy

A mechanical rule anchored to people and joules, run against fifty years of history — and what the committee added instead.

---

1The question Part 2 left on the table

Part 2 ended with an argument about units. The production function was never labor plus capital — it was labor plus capital plus energy, with energy omitted because it was cheap and labor dominated the bill. Automation is driving the labor term toward zero, and energy efficiency — a pursuit as old as engineering, sharpened by every price shock and due to intensify the moment energy is taxed in earnest — stops at a floor that physics sets. Whatever the economy costs at the limit, it costs in energy. If energy is becoming the right unit of analysis — the unit in which production and trade are most honestly measured — then a question follows. Most readers will meet it with skepticism, and the rest of this essay exists to earn it:

Should the unit of account follow the unit of analysis?

Money is three things: a medium of exchange, a store of value, and a unit of account. The dollar does the first superbly and will keep doing it — nothing in this essay changes how you buy groceries. The second and third are where the trouble lives. A unit of account whose supply is decided eight times a year by committee, in response to conditions the committee is also trying to influence, is a measuring stick that changes length while you're using it. For two and a half centuries that was a tolerable defect, because the thing being measured — a labor-mediated economy — had no natural yardstick to offer as an alternative. At the limit, it does. The economy is becoming a machine that turns energy into output. You can peg the measuring stick to the machine.

This essay proposes exactly that, and then spends most of its length on the two questions any competent reader will raise: would it have worked? — which we can now answer with fifty years of data — and what breaks without the committee? — which turns out to be the most interesting question in the whole arc, because the honest answer is that the committee's flexibility is not the safety mechanism it appears to be. It is the mechanism by which the last several crises chose their targets.

2The mechanism, stated plainly

The proposal is a managed monetary regime in which the supply of dollars is constrained by a rule anchored to the physical productive base:

M = k · P · E · (1+i)ᵗ

• M is the monetary base target.
• k is a constant fixed once, at initialization, so that the formula reproduces the money supply on day one. Nothing is revalued; nothing deflates or inflates at the switchover. The peg constrains future expansion, not the existing stock.
• P is population — metered, not estimated.
• E is primary energy consumption within the United States — the productive base behind the dollar, as the EIA already meters and reports it.
• i is a small compounding term, set by legislation on a long review cycle — functionally, this is the inflation target, made explicit and owned. Section 4 is about what i physically is; it turns out to have a physical name and to be the single most clarifying object in the design.
• Existing federal debt remains defined in existing dollars. No retroactive re-marking.

In words: the number of dollars grows with the number of people, with the energy the economy runs on, and with a small legislated allowance for technology making energy go further. Nothing else moves it.

One definition matters enough to state here rather than in a footnote: E enters the formula at its high-water mark — the highest level of primary energy consumption achieved since the peg began, a ratchet, not the current year's flow. Consumption dips in every recession (2009: −5%; 2020: −8% ✦), and a rule pegged to the flow would contract the money supply in downturns — a miniature of the 1929–33 monetary contraction, automated. The ratchet makes the rule acyclic on the downside: in a recession E holds at its prior peak and issuance continues on the population-plus-i schedule, while in expansions E tracks consumption as before. And the ratchet looks forward only: at switchover E enters at current consumption, not at any historical peak — k absorbs the level either way, and a fresh lock lets the meter advance on the first new high above today's. Section 3 runs the fifty-year test with the ratchet applied: the fit barely moves, and the money supply grows in every single year of the path.

Enforcement is by supply control, not redemption. This is not a gold standard with energy substituted for gold: nobody exchanges dollars for kilowatt-hours at a window, and the system never faces a redemption run. The discipline is simpler and blunter — new dollars come into existence only on the rule's schedule, and (Section 7) they enter the economy through a different door than the one they use today.

Operationally, the dollar in your pocket is unchanged. What changes is how new dollars are born, how many, and who touches them first.

3The fifty-year test

Any monetary rule proposed in 2026 should be run against history, so we ran this one against 1975–2024. The ledger first:

The ledger nests cleanly: M2's 21.9× is nominal GDP's 17.4× times 1.26 of velocity drift, and nominal GDP is real GDP's 3.85× times the 4.51× price level. Stated in the units that matter: money grew 5.7× beyond real output (21.9 against 3.85). Of that over-expansion, 4.5× surfaced as the price level, and the remaining 1.26× was absorbed as idle balances — velocity fell (1.75 → 1.39) as dollars increasingly sat in savings deposits and money funds rather than turning over. Whether that velocity decline was benign (wealth grew relative to income, so households held more money) or itself a symptom of over-supply (cash injected beyond transactional need, parking where it landed) is a question the accounting identity cannot settle — and it does not need settling here, because Section 5 shows the parked half feeding the same asset channel either way. One more number worth extracting while the table is open: real GDP per person grew 2.45× — 1.83%/yr, just under two percent.

Now the test. Run the peg rule backward: set k in 1975, meter population and energy as they actually happened, and solve for the i that closes the physical base to real output. The answer is i = 1.21%/yr ✦. Now hold that i fixed — set once, in 1975, never touched — and trace the rule's path against real GDP year by year:

P·E·(1.0121)ᵗ tracks real GDP within ±8% for forty-nine years ✦ — through the Volcker disinflation, the eighties expansion, the dot-com boom and bust, the Global Financial Crisis, and COVID, ending 2024 in a dead heat. The endpoint match is by construction (that's how i was solved); the path staying inside ±8% is not. Nothing about solving on endpoints forces the middle years to behave. They behaved.

Read the chart above as two pairs of lines. The top pair — M2 and nominal GDP — travel together: the honest accounting, stated up front. The bottom pair — real GDP and the rule — sit on top of each other for five decades. And the vertical gap between the pairs is the price level, 4.5×. That gap is the entire content of monetary discretion, 1975–2024. A mechanical rule requiring zero meetings would have supplied the economy's money within eight percent of its real output for fifty years. What the meetings added was the 4.5×.

We frame this deliberately as an identity, not a causal claim. MV = PY is accounting, not theory — in words: the stock of money (M), times how many times each dollar changes hands in a year (V, velocity), necessarily equals the price level (P) times real output (Y), because every purchase is simultaneously a dollar moving and a good moving. The equation can't be wrong any more than a balance sheet can; only the causal stories told about it can. Monetarists read money driving prices, Keynesians read the Fed accommodating prices already in motion, and the chart is agnostic — whatever the causal story, the money supplied beyond the rule equals the inflation experienced, plus half a point of velocity. The argument does not need to win the causality war. It needs only this: discipline was feasible, and its absence was a choice with a measurable price.

The definition of E is itself backtested, not assumed. Pegging instead to jurisdictional supply — production plus imports, the wider base Part 4's tax uses — degrades the fit by nearly half (max error 11%, including a twenty-five-year stretch running 6–10% under output): gross supply swings with trade flows that don't power domestic production. The machine's output tracks what it burns, not what crosses the border, and the data picks the definition.

So is the ratchet. Re-run the same fifty years with E at its high-water mark and the endpoint residual solves to i = 1.12% ✦; the path holds within ±9%, with the worst year relocating to 2009 — where the rule runs high through the trough, which is the standing stabilizer doing its job and being scored as error for it — and the mean miss improving slightly (3.8% against 4.0%). The engagement statistic is the striking one: American energy consumption peaked in 2007 and has yet to re-touch that level, so the ratchet would have been engaged thirty-one of the fifty years — continuously since 2007 — and the rule's money supply grew every single year anyway, never slower than +1.3% (2021, and that was the census, not the meter). Since the rule this essay proposes is the ratcheted one, the ratchet-fit residual is the number the design carries forward: i = 1.12% (the chart below).

4The dial with a physical name

The obvious objection to the formula is i. "You've smuggled discretion back in through a compounding fudge factor and called the result physics." It's the right objection, and the answer is the most clarifying thing we found in the whole recompute.

Decompose real growth: real GDP growth = population growth + energy growth + growth in output per unit of population-energy. Over 1975–2024 ✦: 2.75 = 0.92 + 0.63 + 1.21. The residual — the 1.21 — is the rate at which the economy learned to wring more output from the same people and the same joules. It is the energy-productivity term: the technology residual. Population can be metered. Energy can be metered. Technological progress cannot — it is the one component of the base that must be an assumption, which is why it appears in the formula as the one legislated term. i is not a fudge factor. It is the honest residual of a physical decomposition, and 1975–2024 measured it at 1.21%/yr. Metered on the rule's own ratcheted series — the high-water mark, which grew slightly faster than the flow (0.71%/yr against 0.63) — the same march reads 1.12%/yr ✦, and since the ratcheted series is the one the rule actually meters, 1.12 is the number the design carries. Same physics, stricter yardstick.

Once you see i this way, three things fall out.

First, i is the inflation dial, made explicit. Set i at its neutral point — the actual energy-productivity rate — and money tracks real output: prices are flat (strictly, flat up to velocity drift; Section 14 folds that half-point into the operational setting). Set i one point above neutral and you get one point of trend inflation, on purpose, published in advance. The dollar's actual managers, 1975–2024, ran an effective i of neutral plus about 3.5 points — the 3.08 that surfaced as measured inflation plus the 0.47 that falling velocity absorbed; in supply terms the setting was the sum, even though only part of it reached the price level. The peg does not abolish the inflation choice. It drags the choice out of the fog of thousands of discretionary decisions and makes it one number, chosen in public, owned by whoever chose it. That is the difference between a thermostat setpoint and a hand on a valve.

Second, i is the same object Part 2 measured. The energy-productivity residual is precisely the "automating out energy" rate — LEDs, motor controllers, drivetrains, the fifty-year efficiency march. Part 2's thesis is that this march approaches a thermodynamic floor while output re-couples to energy. As that happens, the residual falls toward zero on its own. The monetary architecture and the physical argument are one claim viewed from two sides: the peg's neutral setting converges to zero exactly as the economy converges to the regime the peg was designed for. An earlier draft of this proposal recommended initializing i near zero, and that was wrong for the transition decades — at i = 0 the rule undershoots real output by the productivity residual, which is a structural deflation of about 1.2%/yr, the precise gold-standard failure this essay condemns in Section 10. i near zero is the destination setting, not the departure setting.

And a caution about the departure: the residual is a pre-AI measurement. It is the labor era's residual — closely related to Part 1's ~1.4%/yr per-worker constant, less the physical-base terms. AI plausibly moves it, and in the near term likely up: software-side productivity gains arrive with modest energy draw, so output can outrun the metered base before the mass physical build-out re-couples them. The residual's expected path is a hump — up through the early take-off, then down toward zero as coupling tightens — and its near-term value is genuinely uncertain in a way its fifty-year average never was. The governance consequence is that the review cadence should ramp like everything else in this arc: frequent early — the board in front of Congress with the meters annually or biennially through the transition decade, while the residual is moving — lengthening toward decadal as the coupling stabilizes and the dial's neutral point settles near zero. Tight cadence when the number is live, long cadence when it is physics.

Third, the governance question answers itself in cadence. i is one scalar, reviewed on a legislated cycle that itself ramps — annual-to-biennial through the transition decade while AI is moving the residual, lengthening toward decadal as coupling stabilizes — always brought by the central bank to Congress with the meters on the table, always benchmarked against the realized residual, with a known long-run direction (down). Yes, that is a lever, and yes, someone holds it. Section 13 counts the levers that remain and the ones that are abolished; the summary is that discretion's bandwidth falls by orders of magnitude — one number, in public, against a meter, versus eight meetings a year moving trillions through a reaction function nobody can state.

One honest limit belongs here rather than in a footnote: the fitted residual is a hindsight number. A 1975 policymaker could not have known it ex ante — though any choice in the 1–2% neighborhood keeps the fifty-year path far closer to real output than the 6.3% that actually happened, and the decadal review exists precisely to correct the drift a wrong guess produces.

5Where the harm actually lives: the entry point

The fifty-year test leaves a question owed to the reader: if the quantity of money roughly matched the nominal economy, what exactly was wrong with the discretionary era? Two things, and they are worth stating carefully because the rest of the essay's architecture answers them one-for-one.

The first harm is the price level itself. The 4.5× was not weather. It was a chosen, sustained tax of roughly three percent a year on everyone whose income adjusts with a lag — which is to say, on wage earners, whose pay is renegotiated annually if they're lucky, while the prices they pay adjust continuously. Fifty years of that lag compounds into the productivity–pay divergence that is now a named exhibit in the national argument: net productivity up 59.7% over 1979–2019 against 15.8% for median hourly compensation — the EPI's three-and-a-half-fold gap ✦. Both figures are real — inflation-adjusted — and the adjustments are what make them look small against lived memory. The 59.7% is net productivity: output per hour net of capital depreciation, economy-wide, over forty years — 1.18%/yr, sitting below Part 1's ~1.4%/yr gross per-worker constant because depreciation's bite grew as the capital stock tilted toward short-lived IT equipment. And the 15.8% is not the nominal wage path anyone remembers living through: production-worker pay ran from roughly $6 an hour in 1979 to $30 in 2024 — a 5× nominal rise — but the CPI did 4.3× of that ◦. Deflate, and five decades of raises compress to fifteen-odd percent of real gain. Not all of that gap is monetary; labor-share dynamics and bargaining power carry part of it. But a permanent, deliberate inflation whose incidence falls on lagged incomes is a standing transfer from wage earners, and it was a policy product.

Part 2 priced the CPI basket in hours of work and watched it split in two — the goods the machine reached falling, the labor-locked services (medical care, tuition, childcare) rising, energy holding flat between them. That split is not just a physical curiosity: it sets the headline "inflation rate," which blends two opposite regimes into one misleading number — falling prices where the machine took over, compounding increases where it couldn't. (Not all of the service-side inflation is labor scarcity; subsidy capture and regulatory cost carry part of medical and education. But the split is the pattern.) This is Section 12's good-deflation regime already running in half the basket — and a preview of the Limit Case's central tension. Automation keeps annexing categories, and the labor content of the basket keeps shrinking; but every category, automated or not, runs on energy. When falling prices are delivered by a machine that employs ever fewer people, cheap goods are no answer for a household with no wage to buy them — participation has to flow through a share of the energy the machine runs on, which is precisely what the peg's per-capita entry supplies (Section 7), before Part 4's tax stack even starts. An economy of falling prices and vanishing wages, with no such channel, has no peaceful equilibrium.

The second harm is the door the money comes through. New money today enters through the banking system: the central bank makes reserves abundant, banks lend against assets, asset prices rise, and the wealth effect is supposed to trickle into consumption. Cantillon's three-century-old observation was never about how much money is created — it is about who touches it first, because the first receivers spend it at old prices and the last receivers meet the new prices with unchanged incomes. The current transmission architecture is not accidentally Cantillon-shaped; it is Cantillon by design — banks first, asset markets second, wages last. And Section 3's parked half of the over-expansion is not idle in the way the word suggests: dollars sitting in deposits and money funds are the funding side of the same asset-credit machinery — the velocity decline and the asset-channel entry are one mechanism seen from its two ends. The Fed's own Distributional Financial Accounts show the top 1% share of household wealth rising from 22.8% (1989) to 30.8% (2024) ✦, with most of the gain arriving as asset-price appreciation. We no longer claim a naive quantity mechanism behind that number — Section 3 forbids us. We claim the channel: when every marginal dollar's first stop is an asset market, the escalator runs upward from the top.

The peg answers the first harm (the rule pins the quantity; i makes trend inflation an explicit, owned choice). Bottom-up issuance — Section 7 — answers the second. But before building the new door, it is worth studying what the old door does in reverse, because the demolition evidence is the strongest material in this essay.

6The blast radius: what 2001 and 2008 teach

Put the two great asset collapses of the era side by side. The dot-com crash, 2000–02, vaporized roughly eight trillion dollars of market value. The 2001 recession that accompanied it is nearly the mildest on record — eight months, shallow, and real household spending never declined at all, a distinction no other post-war recession can claim. Annual real GDP never went negative. On our Part 1 charts you can barely find it: a grey NBER band with no dent under it. Trillions in paper wealth burned off and the productive economy shrugged.

The housing collapse, 2007–09, destroyed a comparable amount of paper wealth — and produced the deepest recession since the Depression, a decade of impaired recovery, and the worst labor-market damage on the modern record.

Same order of wealth destruction. Opposite real-economy outcomes. The economics literature worked out why, and the answer is the entry-point argument running in reverse.

Mishkin and White, surveying fifteen U.S. crashes across a century, found that the crash itself is never the variable that matters — what matters is whether the crash produces financial instability, an impairment of credit intermediation. Crashes that stay out of the banking system are wealth events; crashes that get into it are economic catastrophes. The dot-com bubble was financed by equity — venture capital and IPO shares. Its losses landed on diversified shareholders as unlevered paper losses: no collateral chains, no margin spiral at systemic scale, no holes in bank balance sheets. And the losers were rich: equity is held overwhelmingly by the top decile, whose marginal propensity to consume out of stock wealth is famously small — Case, Quigley and Shiller found "at best weak evidence" of a stock-market wealth effect, in a paper first presented, with beautiful timing, in July 2001. The wealth was also of recent vintage; losing the 1998–2000 gains merely returned households to a wealth level their spending habits had never left. The recession that did occur was exactly where the bubble was and nowhere else: an investment-led downturn — telecom capital spending fell eighty percent — while the consumer never blinked.

The housing bubble inverted every one of those properties. It was financed by debt, intermediated by banks, collateralized by the homes of median households. Mian and Sufi's "levered losses" work made the incidence brutal and precise: the losses concentrated on leveraged, low-net-worth households with marginal propensities to consume of five to seven cents on the dollar and up, whose forced deleveraging and foreclosures fed back into the same collateral values, while the banking system that had levered the whole structure seized. Asset busts hurt in proportion to their entanglement with bank credit. The blast radius is set by whose balance sheet the bubble is wired through — not by how much paper wealth evaporates.

That is the Cantillon entry-point principle, stated for the downhill direction. The door the money comes through determines who captures it on the way up and who absorbs the detonation on the way down. Equity-financed bubble: rich, unlevered, low-MPC — a shrug. Bank-credit bubble: leveraged median households and impaired intermediation — a lost decade.

And now the part that indicts discretion directly. The two episodes are not independent. The Federal Reserve answered the equity bust with eleven rate cuts in 2001, taking the funds rate from 6.5% to 1.75% — and then held it at or below 1% into 2004, years after the recession had ended. We do not claim the counterfactual where the Fed sits on its hands and 2001 stays painless; nobody gets to run that experiment, and some of the mildness surely was the easing — the refinancing channel cushioning demand in real time. What the record does support is narrower and worse: the transmission evidence says an equity bust had limited real-economy teeth to begin with, the response was sized and sustained as if it had fangs, and the overshoot — the years below neutral after stabilization was achieved — is precisely what Mian and Sufi's data identifies as the engine of the 2002–06 household leverage build that defaulted into 2008. The discretionary response to an equity bubble overshot into manufacturing a bank-credit bubble. The administered interest rate is not merely a stabilization tool; it is the mechanism by which the monetary authority — with no such intent — chooses the transmission channel of the next crisis. Some of 2001's mildness was bought honestly. How much of the rest was paid for in 2008, the reader can weigh — the 2002–06 leverage build is in the record either way.

Under the peg with bottom-up entry, trend money never enters through bank leverage at all. The systemic channel is structurally starved — not abolished; Section 14 is honest about the credit superstructure — but starved. The bubbles that form in such a system are of the 2001 class, whose collapse the economy shrugs off, rather than the 2008 class, whose collapse it cannot. We know of no cleaner empirical validation of an architectural choice: the twenty-first century ran the controlled experiment for us, both arms.

(A live footnote for 2026: the AI capital build-out, unlike the dot-com one, is increasingly credit-financed — debt, special-purpose vehicles, vendor financing. Whatever one believes about AI revenues, the financing mix is quietly moving a hypothetical bust from the 2001 class toward the 2008 class. Watch it.)

7Bottom-up: inverting the entry point

Be fair to three centuries of central bankers: money entered through banks because there was no other rail. You cannot hand new currency to every citizen in 1870, or 1935, or even 1990 — the operational surface didn't exist. Issuing through the banking system was not a conspiracy; it was plumbing.

The plumbing constraint is gone. The pandemic proved it: the U.S. government placed funds in essentially every household's hands in days, repeatedly, using rails that already existed. What remains is to stop treating that capability as an emergency measure and make it the architecture.

Under the peg, the rule's new dollars — the P·E·(1+i)ᵗ increment, the money that today would be born as bank reserves — are issued instead as equal per-capita credits to a standing account for every person: call it the Uncle Sam account, cash out to any bank or payment app you like. The credit is fully fungible currency — spend it on rent or groceries or anything at all — and it is simultaneously, with no metaphor required, your share of the new productive base: it exists because new energy capacity came online, and it cannot exist otherwise. The "cash dividend" framing and the "ownership of the machine" framing collapse into the same object.

It is worth pausing on what this mechanism resolves, because the demand for it is already loud. "Pay an AI dividend" is by now a mainstream refrain — and every proposed vehicle stumbles on the same step. Hand out shares of today's AI companies and the state has picked winners: the incumbents of 2026 are capitalized into every citizen's portfolio, permanently, against every garage that would displace them — precisely the anti-market move the proposal's friends think they are avoiding. Tax "AI revenue" instead and you are chasing a definition that dissolves under your feet as models move on-device, into subscriptions, into open weights running at home. The energy meter does not care. Every form of automation — hosted, subscribed, self-run, or not yet invented — draws watts, and the peg's issuance (and Part 4's tax stack, where this argument gets its full hearing) attaches to the watts, not to a corporate register or a definition of "AI." The dividend the discourse keeps asking for turns out to need no new definition at all; it needs the meter that was already the honest unit.

Three consequences deserve emphasis.

The Cantillon effect inverts. New money's first stop becomes every household simultaneously, at the bottom of the income distribution's marginal-propensity gradient rather than the top of its asset-holding gradient. Three hundred years of structurally regressive money-creation becomes structurally progressive — not through a transfer program, but through a change in which door the money uses.

UBI's funding question dissolves. "Where does the money come from?" has always been the fatal question for universal income proposals, because they were framed as fiscal transfers requiring taxes or deficits. Here the answer is: from the same place new money comes from today. The same monetary expansion, a different first stop. What makes this different from money-printing populism is the rule: the peg caps the issuance at physical-base growth, so bottom-up never means unlimited. The discipline objection and the funding objection annihilate each other.

The precedent shelf is respectable and deep. Friedman's helicopter is the canonical thought experiment — direct distribution proposed by the century's most credentialed monetary conservative. Buiter formalized why it works (base money is an asset to holders and a liability to no one). Douglas's social credit movement proposed per-capita debt-free issuance a century ago and got a government elected on it; what it lacked was an anchor, which is exactly what the peg supplies. And direct-distribution architecture is now a mainstream CBDC design conversation in more than a hundred central banks. The novel move here is not the mechanism. It is making the mechanism the default channel of monetary expansion and anchoring the quantity to the productive base.

8Who sets interest rates? Nobody — that's the design

The sharpest question a monetary economist will ask about this system arrived in our own working session within an hour of the bubble-rotation finding: if there's no FOMC, what sets the borrowing rate?

The answer is a duality that organizes everything. Today, the central bank pins the price of money — the policy rate — and lets the quantity float to whatever hits the target. The peg pins the quantity and lets the price float. Nobody sets the borrowing rate, in precisely the sense that nobody sets the price of copper. It becomes a market-clearing price: the classical, Wicksellian configuration in which the interest rate gravitates to the natural rate — the price that balances the supply of real savings against the demand for investment — rather than to a committee's estimate of that number enforced through open-market operations.

What is the market rate made of, under the peg? Four layers. The natural real rate, whose steady-state anchor is the economy's growth rate — which under the peg is energy growth plus the i-residual, so that interest rates re-anchor to the physical growth rate of the productive base. An inflation expectation that is a published constant rather than a forecast — killing not only the Fisher premium but the inflation-risk premium and the monetary-policy-uncertainty premium that fatten long rates today; long-horizon capital, which is exactly what a generation-scale energy buildout requires, gets structurally cheaper term financing. Credit risk, per borrower, exactly as now. And a liquidity layer: interbank rates float on reserve scarcity, smoothed by collateralized repo operations that are temporary and self-reversing — compatible with the rule because only trend supply is pinned — with Bagehot's penalty rate standing as the ceiling in a panic: lend freely, on good collateral, at a price nobody pays voluntarily.

Two objections must be met head-on, and one of them comes from our own files.

"Volcker tried quantity targeting from 1979 to 1982, and it produced the wildest rate volatility on record — that is why the Fed abandoned it." True, and the episode differs from this proposal in every load-bearing respect. Volcker chased a moving M2 target using discretionary operations, amid double-digit inflation expectations that were entirely unanchored — the market was fighting the target because it did not believe it. The peg is a permanent, transparent, legislated rule; expectations anchor to it. The per-capita credit provides a continuous, predictable inflow of base money to household balance sheets — a steady feed of loanable funds, against 1979's lumpy reserve operations. And short-rate smoothing through repo is explicitly inside the rule. The lesson the Fed took from 1982 was "you cannot control the quantity, only the price." The lesson available was "you cannot control a quantity the market doesn't believe you'll keep controlling." A rule solves the belief problem that discretion could not.

"You've abandoned counter-cyclical policy." The rate cut is gone as a tool, yes. Consider what replaces it — and first, consider the trap the rule had to be designed around. A rule pegged to energy flow would be pro-cyclical: recession arrives, energy production dips with it, issuance shrinks or reverses exactly when the economy needs it least — a bad feedback loop, the 1929–33 contraction automated. This is why Section 2 defines E as the high-water mark: in a downturn E holds at its prior peak, and issuance continues — population growth plus i, paid per-capita, every month of the recession, unconditional on employment. The stabilizer is standing, not discretionary: a demand floor under every household through the trough, delivered without pushing the price of credit below its market level and re-inflating whichever asset class is nearest. (In recoveries, E catches up to its old peak before the ratchet advances — the rule is asymmetric by design, generous in the trough and patient at the rebound.) The honest tail-risk sits in the other branch of this arc's fork: if the natural real rate ever went persistently negative — a genuine secular-stagnation world, the creation-cap scenario — the rule cannot do negative rates; the cash floor binds. The take-off premise says returns on deployed capital at the limit are high, not negative. But the caveat belongs in print, not in a drawer.

One more lever must be named, because it does not disappear — it moves. The Fed loses its discretionary levers under the peg; Congress does not lose its power to legislate. In a genuine emergency, the government that made the rule can break it: appropriate funds and inject them, pandemic-style — and mechanically this is now trivial, because the peg already built the rail. The per-capita accounts of Section 7 are a standing national delivery mechanism, and 2020 proved the pattern reaches every household in days. What changes is the accounting and the ownership. Today, crisis stimulus arrives as a monetary-fiscal blend nobody fully owns — rate cuts here, asset purchases there, checks from somewhere. Under the peg, the counter-cyclical lever belongs to Congress alone, exercised from the Treasury, on Congress's own budget, in public, with a vote attached. Emergency money becomes fiscal policy with a signature on it rather than monetary policy with a committee behind it — the same move that made i one owned number in public makes crisis discretion one owned appropriation in public.

And the payoff loops back to Section 6: an administered rate is how discretion chooses the next crisis's transmission channel. Under the peg, no authority can hold the economy-wide price of credit below its natural level. Credit booms must be funded by real savings, or the rate rises and chokes them. Bubble rotation stops being a policy product. Friedman's k-percent-rule literature is the credentialed shelf — under a quantity rule, rates are market-determined, and Friedman spent a career arguing central banks never controlled real rates in the long run anyway.

9The dollar the world uses

Everything so far treats the dollar as America's money. It is also the world's, and the redesign has to answer for that too. The honest sizing surprised us.

The textbook story is official: foreign central banks hold roughly $6.5–7 trillion of dollar reserves, about 58% of the global total. But the official layer is the smaller half of the dollar system. Foreign private holdings of Treasuries now exceed official holdings (roughly $4.5T+ of the $8.45T total — which itself grew 128× since 1975 ✦, against M2's 22×). Offshore dollar credit to non-U.S. borrowers — the eurodollar system — runs $12–13 trillion, larger than all official reserves combined. The FX swap market carries $80+ trillion in dollar payment obligations, short-tenor and perpetually rolled — the BIS calls it "missing debt." Roughly half of world trade is invoiced in dollars — four to five times the U.S. share of world imports — and 88% of all foreign-exchange trades have a dollar on one side. This is not a system of governments being polite to the reserve issuer. It is the entire global trading apparatus — corporate working balances, trade receivables, hedging stacks — running on one country's unit of account.

Two consequences for the argument, cutting opposite ways.

The demand suppressed our measured inflation. All that external appetite for dollar assets propped the currency, cheapened imports, and absorbed Treasury issuance at suppressed yields (the "convenience yield" literature prices the subsidy at 25–60 basis points). The same 6.3%/yr of money growth bought less measured inflation than it would have in a non-reserve world — which means Section 3's 4.5× price level, the gap between the pairs, is a lower bound on what the discretionary era's issuance "deserved." The reserve privilege flattered discretion's report card. The counterfactual chart is conservative.

But the Fed never printed most of those dollars. Eurodollars are created offshore by non-U.S. banks. What the United States actually supplies is the settlement core — correspondent accounts, Treasuries as universal collateral — and the crisis backstop: when the offshore dollar system seizes, as in 2008 and 2020, Federal Reserve swap lines are what save it. That is the piece the peg genuinely touches, and it produces the sharpest transition requirement in the design: the first global dollar squeeze after the peg will demand swap lines — discretionary dollar creation — and if the rule has no answer, the rule dies in its first crisis. The answer is the same Bagehot valve that handles domestic panics — and it is worth saying concretely what a swap line is, because "discretionary dollar creation" makes it sound like trend money, which it is not. In a swap, the Fed wires newly created dollars to a foreign central bank and takes that bank's own currency at the market exchange rate as collateral, with a contract to unwind the trade at the same rate on a fixed date — days to months out — plus interest at an above-market penalty rate. The foreign central bank owes the dollars back regardless of what its commercial banks do with them; the Fed carries no exchange-rate risk and picks no borrowers. So the dollars go out, put out the fire, and are extinguished on return: the 2008 draw peaked near $580 billion and unwound to roughly zero within a year; 2020 peaked near $450 billion and was mostly repaid within months ✦. A loan that must reverse, priced so nobody keeps it a day longer than the emergency — that is what "collateralized and self-reversing" means, and it is why swap lines can live in the emergency channel, explicitly walled off from trend supply, without breaching the rule. That wall belongs in the statute itself; left to interpretation, the first global squeeze will redraw it under duress.

On the transition itself, honesty requires a harder statement than the usual Triffin paragraph. Official reserves can diversify by policy decision; trade invoicing is a network-effect equilibrium — everyone prices in dollars because everyone else does — and network equilibria unwind on decades-scale or chaotically, never by memo. Sterling's displacement took roughly half a century and two world wars. A pegged dollar starts a long re-invoicing transition whose speed it does not control, during which the emergency valve must remain dollar-scale. The costs of exiting the privilege are real and should be printed: the convenience-yield subsidy on foreign-held Treasuries (order $30–40B/yr ◦), and the seigniorage stock — roughly $1.1 trillion of foreign-held physical currency ✦, real goods the world traded us for paper. Against the stakes of the regime, these are small; and Triffin's own point, made in 1960 and vindicated in 1971, is that the privilege was always self-terminating. The peg does not destroy the exorbitant privilege. It chooses the exit, rather than having the exit chosen for it.

10The Cross of Gold, resolved

This proposal will be pattern-matched to gold-buggery, so meet the pattern head-on with the era that created it.

Bryan's 1896 "Cross of Gold" was a correct critique. Under the classical gold standard, the monetary base grew with mined gold — about 1.5%/yr — while the American population grew at 2.3%/yr. Per-capita money contracted almost one percent a year for a generation ◦, against per-capita productivity growth of 1.5–2%: a structural deflation of two to three points, compounding, borne overwhelmingly by debtors — the farmers behind the Free Silver movement — and punctuated by the panics of 1873, 1893, and 1907 in a system with no lender of last resort. "You shall not crucify mankind upon a cross of gold" was sound monetary economics in one sentence.

The twentieth century's answer was discretionary fiat, which broke the other way — the 4.5×, the asset-channel entry, the wage lag: Cantillon crucifying the same class by subtler means. A hundred and thirty years of monetary debate has been a pendulum between an anchor that starves and a discretion that skims.

The energy peg is the third path, and the arithmetic distinguishes it from gold decisively. Gold's supply is indifferent to the economy — that is the point of gold, and its fatal flaw: its value is scarcity. Energy's value is use; its supply is the productive base — so the peg's base grows at the pace of the economy it serves: a modest 1.55%/yr in our mature test window, four-plus percent in an industrializing one. Gold grows at gold's pace regardless — a flat ~1.5%, no population term, no productivity term, whatever the economy underneath it is doing. Run the full rule — population × energy × (1+i)ᵗ — through Bryan's own era, and the contrast is not subtle. Population grew ~2.3%/yr. American energy production grew 2–3%/yr through the late nineteenth century as coal displaced wood and the railroads industrialized ◦. The peg's base therefore expands at roughly 4.5–5%/yr before the i-term — right at the era's real output growth of about 4%/yr ◦, if anything a touch loose, in exactly the direction the debtors were marching for — where gold managed a flat 1.5%. Instead of per-capita money contracting one percent a year for a generation, the supply keeps up with people and growth: no structural deflation, no debt spiral grinding the Plains, no silver debasement required to get there. Bryan asked for silver, which was merely a looser scarcity. He should have asked for the anchor that tracks the machine. Anchored like gold, scaling like the economy, distributed bottom-up: anti-inflation, anti-deflation, anti-Cantillon — one instrument, all three.

11The deflation objection, honestly

The most serious technical objection to any supply-constrained regime is the one the gold record teaches: if the base grows slower than transaction demand, you get deflation, debt-burden spirals, and liquidity panics. Does the peg import that failure mode?

Partially — and the mitigations are structural, not hopeful. In order of load-bearing:

The take-off is the primary mitigation, and the integration is the point. This essay sits inside an arc whose Part 1 diagnosis is an AI-driven take-off in which energy consumption grows at multiples of its historical rate — data-center demand is the leading edge. Under take-off, the peg's base grows fast enough that the deflation question inverts into "is the base growing too fast?" (to which the answer is: that is real capacity, and Section 7 distributes it). The regime and the diagnosis are coupled by necessity: without the take-off, the peg faces nineteenth-century pressure; without the peg, the take-off's gains pool at the entry point. One system, not five reforms.

i is the deflation offset. This is what the dial is for. The historical neutral was ~1.1%; a legislated i covers the gap between physical-base growth and real-output growth by construction, reviewed decadally against the meters. The creation-cap branch — where the take-off underdelivers and the residual stays high — is handled by i staying high; the destination regime — coupling tight, residual near zero — is handled by i ramping down. The Cross-of-Gold failure required an anchor with no such term. This one has it as a first-class object.

Initialization headroom, Bagehot, velocity, and the credit superstructure complete the stack: k set with deliberate slack so the rule doesn't bind on day one; emergency liquidity at penalty rates against collateral (the panics of 1873–1907 happened in a system without a central bank — we are not proposing to abolish the fire department, only the arsonist's day job); modern settlement supporting far more nominal activity per unit of base than 1890's; fractional-reserve credit continuing to operate on top of the base exactly as it did under gold; and bottom-up distribution itself raising velocity, since money at median households turns over faster than money at asset managers. Existing nominal debts stay nominal, and if persistent deflation nonetheless emerged, indexed-conversion options for legacy borrowers (TIPS run in reverse) prevent the debt-deflation spiral Fisher named.

What we do not offer is a guarantee of quarter-by-quarter price stability. Velocity is not constant — our own ledger shows 0.47%/yr of drift — and the rule pins supply, not demand. The claim is fifty-year discipline with explicit, owned trend inflation, not a price-level thermostat. Say it plainly and the objection loses its best ammunition, which is the suspicion that we haven't thought about 1893.

12What it would have felt like

Counterfactuals are treacherous — under a real peg the last fifty years unfold differently in a thousand ways, and the chart shows the gap, not the alternative history. But the texture is worth one section, because the regime's daily experience is deeply unfamiliar to anyone raised inside permanent inflation, and unfamiliarity is most of the resistance.

With i at neutral, the price level is roughly flat across a working lifetime ◦. Nominal wages are roughly flat too — and real purchasing power grows anyway, because productivity gains arrive as falling prices instead of nominal raises perpetually chasing them: the "good deflation" Schumpeter and Hayek described, in which the cheapening of goods is the distribution mechanism. The per-capita credit arrives monthly on top — your share of new capacity. Asset prices track real productivity rather than money growth; the escalator still runs, but only as fast as the machine underneath it. Savings hold value without financial engineering; a retiree's cash is not a melting ice cube; the two-earner household's arithmetic is not silently rewritten every decade by a lag structure nobody voted on. First-pass sizing of the wage counterfactual (◦ — the seed's estimate, awaiting re-derivation against the recompute) put broad-based real gains near 2%/yr against the realized ~0.7% median — cumulatively, a median household at roughly twice its realized purchasing power. We hold that number loosely and the mechanism tightly: the productivity–pay gap is not a law of capitalism; it is a property of a particular monetary architecture, and architectures are choices.

And the architecture's central choice — the i-dial — can be laid out as a published menu, with the velocity drift priced in (Section 14): the operational zero-inflation setting is 1.6% (the ratchet-metered residual of 1.12 plus 0.47 of drift), and each row above it adds a stated point of net inflation. Every row is computed on today's money supply and today's physical growth ✦ — $21.9T of M2 across 341 million people, about $64,200 per head, growing at the metered 1.42%/yr plus i. Every row sets inflation below the discretionary era's realized average of ~3.1%/yr — a number the pandemic barely moves: stop the meter in 2019 and the average is still ~3.0%/yr ◦, because the era's inflation was front-loaded in the 1970s, not manufactured in 2020. And in every row the entire proceeds of issuance arrive per-capita instead of through the banking channel:

The memo row is the argument. Households paid roughly 3.1% a year of inflation for five decades and received none of the proceeds — the seigniorage entered through the banks and pooled where the entry point aimed it. And the memo row survives its best defense. Grant that the pandemic years were an anomaly and strike them from the record — stop the meter in 2019: the effective setting barely moves (~4.5% ◦ against ~4.7%), and measured inflation barely moves with it (~3.0%/yr ◦ against 3.08). The era's excess was not manufactured in 2020; it was front-loaded in the 1970s, and the pandemic M2 surge was partly reabsorbed by the 2022–23 contraction that followed it. Every row on this menu undercuts the discretionary record under either accounting. There is no setting here that isn't a strict improvement on the status quo along both axes: less inflation than the realized average, and a check. If the design must name its number, it is the second row: 2.6% total — the drift-corrected neutral plus one point. The extra point buys a standing margin against the deflation floor and the lubrication nominal wages need, undercuts the discretionary era's delivered inflation even in its best-behaved decades, and writes a family of four a roughly $10,300 annual check. Calibrate the checks against real household stress: the Federal Reserve's own survey work finds more than a third of American adults unable to cover a $400 emergency from cash — the zero-inflation setting retires that entire class of stress at $162 per person per month, and the recommended setting hands a family of four roughly $860 a month — most of a grocery budget, or a large bite of the rent — unconditionally, scaling with the buildout. Note also what happens to incidence at any setting above neutral: the inflation-tax side of the ledger scales with nominal holdings, which concentrate at the top, while the payout side is flat per head — the net transfer is automatically progressive for every household holding less than the per-capita average of nominal assets, which is to say nearly all of them. And this is the smaller of the regime's two streams: Part 4's tax-funded energy credit (~$5,100 ◦ per person at Phase 3 maturity) stacks on top.

The political reading writes itself, and Part 5 will collect it: to the conservative, this is hard money, near-zero income tax (Part 4), market allocation untouched, and a Fed stripped to a fire department; to the progressive, it is universal monetary participation, an inverted Cantillon effect, and the distributional contract restored by structure rather than transfer. Both are describing the same machine.

13The Fed that remains

Name what survives, because "abolish the Fed" is not the proposal and the distinction is load-bearing.

Two levers remain. i — one scalar, legislated, decadal, public, benchmarked against a metered residual, with an expected downward glide as coupling tightens. And the Bagehot valve — fast, unlimited-in-the-crisis, collateralized, penalty-priced, self-reversing, extended explicitly to the swap-line theater the global dollar system requires. Between them: a thermostat and a fire department. A third lever exists but sits outside the building: Congress's emergency power to break its own rule from the Treasury, on its own budget, through the per-capita rails (Section 8) — fiscal, voted, and owned, rather than monetary and diffuse.

What is abolished is the middle — the reaction function, the meeting cycle, the QE toolkit, the standing invitation to treat the price of credit as an instrument of macroeconomic mood management. The institution's mandate collapses from an unstable dual objective to a single one: keep the money supply aligned with the productive base, and keep the pipes from bursting.

The failure-mode accounting, which is the closest thing this essay has to a closing argument, runs one-for-one:

One architectural change addresses every row. That is not a list of features stapled together; it is the signature of having found the actual joint.

14Honest limits

For the record, in one place, the things this essay does not claim:

• The physical anchor carried 59% of the historical tracking; the i-term carried 41% ✦ (on the rule's ratcheted meter; 56/44 on the flow meter). In the decoupling era the "physics" is barely more than half the story. The defense is the destination regime (coupling tightens, residual falls), not the rearview window.
• The 1.12% is a hindsight fit. Ex ante, a 1975 board could have missed it by half a point in either direction; the decadal review is the correction mechanism, and ±0.5% errors produce drift measured in single-digit percents per decade, not regime failure.
• Velocity is not constant — and the drift has a sign. Velocity fell ~0.5%/yr over the window, and under a supply rule falling velocity is deflationary: money increasingly parked as savings is money not chasing output. The operational consequence belongs in the design, not the footnotes: the zero-inflation setting of i is the physical residual plus observed drift — nearer 1.6% than 1.1% on the historical window — and the review cycle re-benchmarks the drift term against measured velocity exactly as it re-benchmarks the residual against the meters. The rule guarantees supply discipline, not price-level exactness; the i-review is where the slippage gets paid.
• The credit superstructure can still lever a mania. Fractional banking on top of the base means 2008-class dynamics are structurally starved, not structurally impossible. Regulatory vigilance over collateral chains does not retire.
• The secular-stagnation tail. If the natural real rate went persistently negative — the creation-cap branch with no take-off — the cash floor binds and the regime has no negative-rate tool. The per-capita credit is the demand-side answer; it is untested at that duty.
• The transition is the risk concentrate. Reserve-role unwind, re-invoicing network effects, legacy-debt psychology — decades of exposure, manageable only with the emergency valve intact and the phase-in (Part 4's three phases) honored.

15Close

Strip the essay to its spine. A mechanical rule anchored to people and joules, set once in 1975 and never touched, would have supplied this economy's money within nine percent of its real output for fifty years — growing it every single year, through six recessions — no meetings, no forecasts, no committee. What the committee added was a 4.5× price level whose incidence fell on wages, an entry door that ran every new dollar through asset markets first, and — in its most consequential exercise of flexibility — a response to an equity bust that overshot, in depth and duration, into financing the bank-credit bubble that followed. The defense of discretion has always been that economies need judgment. Fifty years of the meter says the judgment was the volatility.

Part 2 argued the economy's true unit is energy. This essay put the currency on that unit and found, in the data, that the anchor holds. Part 4 builds the tax and distribution stack on top of it — because a currency that enters at every citizen equally is necessary but not sufficient for the regime the Limit Case demands. The engineering continues.