Preface

Monetary economics has no founding equation and no organizing theorem. It orbits a question, and the question is the neutrality of money: when does money matter for real outcomes, and through what channel?

What makes the field unusual, and what organizes this history, is that the answer flips sign roughly every twenty years, and each flip is purchased by an inflation or a deflation that the reigning consensus failed to predict. The Depression made money look impotent. The Great Inflation made it look omnipotent in the short run. The rational expectations purge made even the short run disappear. The Great Moderation rebuilt non-neutrality on sticky prices and declared the problem solved. Then 2008 and 2021 arrived, one decade apart, and humiliated the synthesis twice, from opposite directions. Few fields of economics have been so disciplined by events, and none has had its ideas applied at such scale: the Volcker disinflation, inflation targeting, quantitative easing, and forward guidance are all cases of central banks acting on the papers this history covers, which means the debates here were settled partly in seminar rooms and partly in the unemployment statistics.

The load-bearing device, then, is the pendulum itself. Each era below opens by naming the episode the field was answering for and stating the era’s answer to the neutrality question as a claim a reader could contest. Each closes with the loose thread the next era pulls. The papers appear not as contributions in a parade but as ammunition in fights, because that is how the field actually used them.

Prologue: Quantity Theory Without a Crisis (1752-1935)

The prologue has no crisis, and the absence is the point. Every later era in this history is the field’s answer to an empirical humiliation, a moment when the consensus predicted the wrong thing in public. The prologue is the field before it had ever been humiliated. Its live episode is the slowest inflation on record: the New World silver that raised European prices, in Hume’s accounting, “three, or at most, four times” over two centuries (Hume, 1752). Inflation as secular drift invites a leisurely answer, and the prologue supplied exactly that: the quantity theory as a proposition about the price level in the fullness of time, with no theory of fluctuations attached.

Stated as a claim a reader could contest: the prologue’s answer to the neutrality question is that money is neutral in the long run and non-neutral only during a transient “interval,” and the era treated that interval as a nuisance to be explained away rather than as the main event. The claim is contestable precisely because the era’s most durable insights turned out to be buried in the caveats. The field spent 165 years confidently answering “neutral,” and the next century spent itself discovering that the throwaway clauses about the interval were where the action was.

Hume already had both halves in 1752. Prices are “proportioned to the plenty of money”; the absolute quantity is “of no manner of consequence” to a nation’s happiness. And then, almost as an aside, he breaks his own theorem: new money enters unevenly, reaching first “the coffers of a few persons” who hire workers at old prices, so money stimulates industry in the interval “between the acquisition of money and rise of prices” (Hume, 1752). When the raw proportionality looked off, Hume had reconciling machinery ready: commodities brought to market had expanded too, and money “locked up in chests” affects prices “as if annihilated” (Hume, 1752). Long-run neutrality plus a transitional stimulus, stated in the same essay. Every later figure in this prologue inherits that shape and files the short run under corrections.

The received story treats the next century and a half as refinement of one doctrine. I think that story is wrong: what actually happened between 1752 and 1917 is that the causal engine of the quantity theory migrated, and the neutrality answer flipped sign with each move, establishing the field’s twenty-year oscillation before any crisis existed to force it. Hume put causation in the money stock. Wicksell, in 1898, took it out entirely. His pure-credit thought experiment, an imagined banking system where all payments run through ledgers and physical currency is inessential, was built to escape what Ohlin’s introduction to the English translation called the tyranny of the concept “quantity of money” (Wicksell, 1936). In that economy the operative variable is a gap between two interest rates: the natural rate, the return that would clear the market if real capital were lent in kind without money intervening, and the money rate banks actually charge. A money rate held persistently below the natural rate does not produce a one-time price adjustment; it produces a cumulative process, a progressive rise in prices proceeding “without limit,” with the symmetric spiral in deflation (Wicksell, 1936). This is the deepest flip of the era. Non-neutrality is no longer transient, it is unbounded, and the money stock is no longer even the right thing to watch. Wicksell believed he was rescuing the quantity theory by supplying its missing transmission mechanism; what he actually wrote was the founding document of interest-rate central banking, the book that replaced “control the money stock” with “set the interest rate relative to its natural level” (Wicksell, 1936).

Fisher pulled causation back to the stock and made it exact. The equation of exchange, MV = PT, is an accounting identity given causal direction by holding velocity and trade fixed, and Fisher confirmed it statistically: assembling independent U.S. series for all six magnitudes over 1896-1909, he showed that the directly computed price level and the level implied by the other five factors agree to within roughly 11 percent, a residual inside the probable error of the underlying data (Fisher, 1911). That was the strongest quantitative confirmation the era produced, and Fisher intended it as the triumph. The durable part of the book sits elsewhere, in a chapter defensively titled “transition periods,” where a lag of the money interest rate behind inflation drives a self-reinforcing credit cycle (Fisher, 1911). Fisher answers “neutral” in the main text and “non-neutral” in the caveat, and the field later inverted his ranking: the stable identity became pedagogy, the credit-cycle mechanism became debt-deflation after 1929 wiped out Fisher’s own fortune.

Pigou made the final move in 1917, recasting the identity as a demand: P = kR/M, where k is the share of resources people choose to hold as titles to legal tender, grounded in optimizing “volition” rather than mechanical velocity (Pigou, 1917). He insisted the reformulation was merely “more convenient,” “no truer” than Fisher. It was truer in the sense that mattered: it turned “the quantity of money” into “the demand for money,” the object on which Keynes’s liquidity preference and Friedman’s 1956 restatement would both operate (Pigou, 1917; Friedman, 1956). The prologue’s one genuine dispute is latent here, and it will not surface for forty years: is the quantity theory a theory of the money stock or of money demand? The stock side had Fisher’s statistics and Hume’s arithmetic; the demand side had Pigou’s optimization and Wicksell’s wedge, the demonstration that in a pure-credit economy the stock is inessential, so any stock-centered theory describes an artifact of a coin economy rather than a mechanism. When the dispute finally surfaces, in the fight between Friedman and Patinkin over what the 1956 “restatement” restated, both winning arguments will turn out to be prologue property.

Only one of these four figures survives into the modern canon under his own name, and it is not the two who owned the identity. Hume survives as the first page of every money chapter; Fisher survives as an ancestor monetarism invokes; Pigou survives absorbed and unrecognizable inside liquidity preference. Wicksell survives as Wicksell. Woodford titled the 2003 treatise that codified modern central banking Interest and Prices, in deliberate homage, and the natural-rate-versus-market-rate gap is the organizing object of the entire New Keynesian synthesis (Wicksell, 1936; Woodford, 2003). The cumulative process is the direct ancestor of the requirement that a central bank move its policy rate more than one-for-one against inflation, precisely because a persistent gap detonates a spiral.

The loose thread the rest of this history keeps pulling is Wicksell’s, and it is a measurement problem disguised as a concept. He defined the natural rate verbally, as the rate that would emerge if capital were lent in kind, and left it unobserved, noting that it moves independently as a second factor (Wicksell, 1936). A century later Laubach and Williams operationalized it as r-star, a Kalman-filtered state variable the New York Fed now publishes in near-real time, with standard errors of one to two and a half percentage points, too imprecise to use mechanically (Laubach and Williams, 2003). The prologue bequeathed the field a concept it cannot live without and cannot measure. Everything that follows is, in one register or another, the bill for that bequest.

Era I: The Depression’s Answer (1936-1967)

The episode is the Great Depression, and it humiliated classical economics at the one thing a theory of markets must do: explain why markets clear. Classical doctrine said that if labor were priced right, unemployment would cure itself; the 1930s produced years of mass involuntary idleness that no amount of wage deflation corrected. The era’s answer to the neutrality question, stated so a reader can contest it: money is nearly neutral over the business cycle, not because prices are flexible (the classical route to neutrality) but because monetary policy is impotent in a slump, so the level of output belongs to fiscal policy. The claim is contestable because the era’s own formalizers proceeded to prove that the impotence is a special case, and that internal betrayal is the drama of the period.

Keynes and the theory the Depression demanded

The General Theory is the field’s answer to a category error. Classical economics was a theory of allocation given full employment; it had no theory of the level of aggregate output, and Keynes built one (Keynes, 1936). The principle of effective demand fixes employment where the aggregate demand function meets aggregate supply, and because Say’s Law is false, that intersection can and typically does sit below full employment, with no market force pushing it higher. Keynes’s claim about generality is in the title: involuntary unemployment is the general case, full employment the special limit (Keynes, 1936).

Money enters the argument twice, and both entries demote it. First, the interest rate is stripped of its classical role as the price that clears saving against investment. Interest is instead the reward for parting with liquidity, set by liquidity preference and the money stock, M = L(r), which means the rate can lodge above the level consistent with full-employment investment and stay there (Keynes, 1936). Second, even where the rate might move, investment demand is governed by “animal spirits,” a spontaneous urge to act rather than a mathematical expectation, with professional investing resembling a newspaper beauty contest in which one picks the faces others will pick (Keynes, 1936). The lever classical theory trusted, cheaper money calling forth investment, is disconnected at both ends. Chapter 19 closes the escape route: cutting money wages cannot reliably restore employment, because employment depends on effective demand, not the wage level, and falling wages act at best like an increase in the money supply while carrying destabilizing expectational and debt-deflation effects (Keynes, 1936). Note what Keynes is doing to the neutrality question. He is not arguing that money is neutral in the classical sense; he is arguing something stranger, that money is so entangled with expectations and liquidity that the monetary authority cannot use it. Impotence, not neutrality. The distinction will matter when Friedman and Schwartz reopen the case.

The formalizers’ verdict, which cut against their author

The first dispute of the era is Keynes against the classics: does the interest rate clear saving and investment, or does income? The fight is over an unobserved counterfactual, since both mechanisms produce the same observed equilibrium, and it was settled not by data but by arbitration, and the arbitrators’ verdict cut against the man they canonized.

Keynes’s side deserves its full weight. If the rate is pinned by liquidity preference, it can sit above the full-employment level indefinitely, leaving income to absorb the saving-investment gap, which is exactly what a decade of depression looked like: falling wages, slack demand, no self-correction (Keynes, 1936). The classical side answered by absorption rather than rebuttal. Hicks wrote both theories as one system of simultaneous equations and showed the entire dispute collapses into the choice of money-demand equation, the classics closing with M = kY and Keynes with M = L(r); whether income or the rate adjusts is nothing but the slope of the LL curve, flat in the liquidity trap, steep in the classical region (Hicks, 1937). Modigliani sharpened the verdict into the era’s most cutting internal result: liquidity preference is neither necessary nor sufficient for underemployment equilibrium. Strip out rigid wages and, outside the single limiting “Keynesian case” where the full-employment rate falls below the liquidity-trap floor, the classical dichotomy and the loanable-funds rate survive intact (Modigliani, 1944). On the literal question, the answer is “it depends on the wage assumption, not on liquidity preference,” a classical-leaning result.

What got canonized is a different matter, and here the scoreboard splits. Keynes lost the theorem: the synthesis re-cast involuntary unemployment as the special case requiring wage rigidity, exactly inverting his claim about which case is general (Keynes, 1936; Hicks, 1937; Modigliani, 1944). But the synthesis won the field by building the machine, IS-LM and the neoclassical synthesis, that ran postwar macroeconomics, and the machine embalmed a caricature. Hicks himself later disowned IS-LM as a classroom gadget that flattened Keynes’s expectations, time, and genuine uncertainty into a static snapshot (Hicks, 1937), and the Chapter 12 core the synthesis quarantined, animal spirits and radical uncertainty, kept being rediscovered in crises, by post-Keynesians, and after 2008 by behavioral macro (Keynes, 1936). The honest summary is not a draw: the synthesis owns the textbooks, Keynes owns the slumps.

The era’s single load-bearing mechanism appears three times under three names, and noticing this is the fastest way to understand what “money barely matters” meant operationally. It is Keynes’s interest rate stuck too high (Keynes, 1936), Hicks’s flat LL segment where additional money is absorbed into idle balances without moving the rate (Hicks, 1937), and Modigliani’s overdetermined Keynesian case (Modigliani, 1944). All three say the same thing: at the floor, money is impotent and only fiscal action moves output. Modigliani’s proof also grounds the claim from the other side, since under flexible wages the classical dichotomy returns and money is neutral in the old sense (Modigliani, 1944). Money barely matters either way: neutral when wages flex, impotent when they do not.

One repair job completes the demand side. Keynes’s liquidity-preference schedule rested on implausibly inelastic interest-rate expectations, and Tobin rebuilt the downward-sloping demand for money from risk aversion in a two-asset portfolio: all risk-averters are diversifiers, plungers do not exist (Tobin, 1958). The rescue conceded more than it defended. By moving money demand from a macro reflex to a portfolio choice, Tobin took the first step toward treating money as one asset among many rather than the special store of value Keynes had made it, and the separation theorem buried in the paper became modern portfolio theory, a larger legacy than the macro wrapper it was built to serve (Tobin, 1958).

The graft: an inflation block with no theory of expectations

The synthesis had a theory of output and no theory of inflation, and the gap was filled by an empirical regularity rather than a mechanism. Phillips fitted a stable, explicitly nonlinear, downward-sloping relation between unemployment and wage inflation on nearly a century of United Kingdom data, 1861-1957, with the fitted curve y + 0.900 = 9.638x^(-1.394), grounded in asymmetric wage adjustment: employers bid wages up quickly when labor is scarce, workers resist cuts when it is slack (Phillips, 1958). The relation tracked out of sample; the 1948-57 points fall nearly on the prewar curve once unemployment is lagged (Phillips, 1958). Two years later Samuelson and Solow translated the relation into price-inflation space for the United States and read off a “menu of choice”: roughly 5.5 percent unemployment for price stability, or 4 to 4.5 percent inflation to hold unemployment near 3 percent (Samuelson and Solow, 1960). The synthesis now had a supply block bolted onto its demand-managed core, and the residual policy problem became choosing a point on a curve.

This is the opening of the Phillips-curve stability dispute, the longest-running fight in this history, and the era’s side of it deserves a fair steelman before the 1970s demolish it. The menu’s best evidence was durability: fitted on 1861-1913, the relation predicted the 1925-29 gold-standard deflation almost exactly and caught the 1948-57 points on the prewar curve, out-of-sample checks spanning radically different regimes across nearly a century (Phillips, 1958). And the historiographical record demands a correction that this history will insist on when the demolition comes: the menu’s authors hedged. Samuelson and Solow warned in the text that the menu “would not keep its shape” over the longer run and might shift upward if a slack economy bred structural unemployment (Samuelson and Solow, 1960). Phillips asserted no policy menu at all; his paper was a cautious, atheoretical accounting of demand-pull wage adjustment, and the exploitable-menu reading was imposed on him by his translators (Phillips, 1958). The stable exploitable trade-off that the 1970s discredited was partly a myth read back onto a caveat-laden two-point sketch. The natural-rate side would win the economics and caricature its opponents in the same motion.

The deeper problem with the graft was not the hedging but the hole underneath it. Nothing in the synthesis pinned down expected inflation. Phillips’s cost-of-living term captured expectations only partially and asymmetrically (Phillips, 1958); Samuelson and Solow’s warning about shape was an intuition without a mechanism (Samuelson and Solow, 1960). The era had welded an inflation-unemployment trade-off onto a body of theory that contained no account of how the public forms beliefs about inflation, which meant the trade-off’s stability was silently conditional on a stable expectations regime nobody had modeled.

The loose thread

Two threads run out of Era I, and both were visible at the time to anyone who looked. The first is the expectations gap just named: a consensus built on a statistical Phillips curve with no theory of expectations underneath. Friedman and Phelps will grab exactly this loose end, make expected inflation an explicit shifter, and watch the long-run curve go vertical. The second is the Depression itself, the era’s founding evidence. Era I read 1929-33 as proof that monetary policy had been tried and found impotent, that one cannot push on a string. Friedman and Schwartz will turn the same episode inside out: the money stock fell by over a third, and feasible actions by the monetary authorities could have prevented it (Friedman and Schwartz, 1963). The catastrophe that launched “money barely matters” will be re-entered into evidence as proof of money’s neglected power, and the neutrality answer will flip its first postwar sign.

Era II: The Counterrevolution (1956-1977)

This era is bracketed by two humiliations and authored by neither. It opens in the shadow of the Depression, the episode that had convinced a generation that money was passive, and it closes as the Great Inflation arrives to humiliate the other side, the belief that a government could choose its preferred point on a stable menu of inflation and unemployment. The counterrevolution answered the first humiliation and was vindicated by the second, a timing so convenient that the field has never fully separated the vindication from the argument.

The era’s answer to the neutrality question, stated contestably: money matters enormously in the short run and not at all in the long run, and the hinge between the two is whether the inflation was expected. Monetary disturbances dominate nominal income over any horizon a policymaker cares about; yet no lasting real trade-off can be bought with inflation, because the long-run Phillips curve is vertical at a natural rate that real forces determine. The claim is contestable because it rests on a specific and, as the next era proved, fragile account of how expectations are formed. The counterrevolution’s conclusion survived; its mechanism did not.

The charter and its contested paternity

The opening document is Friedman’s 1956 “Restatement,” which reframes the quantity theory as, in the first instance, a theory of the demand for money: a portfolio function M/P = f(y, w; R_m, R_b, R_e, g_p; u), with the operative claim, asserted almost in passing, that this function is stable, stable enough that changes in the money stock map into changes in nominal income (Friedman, 1956). Everything the monetarists did for the next two decades is that one conjecture under repeated load.

Whether the restatement restated anything is itself one of the field’s disputes, and it should be scored honestly rather than smoothed over. Patinkin’s charge was that Friedman’s function is, in substance, modern post-Keynesian portfolio and liquidity-preference theory, the Cambridge cash-balance approach that Pigou founded and Keynes extended, “misleadingly entitled” a quantity theory and dressed in a Chicago oral tradition that functioned as a marketing label (Friedman, 1956; Pigou, 1917; Keynes, 1936). On the substance, the historiographical consensus sides with Patinkin: the demand-for-money object is Pigou’s, the asset-theoretic sophistication is Keynesian, and the strongest historiographical reading is that Friedman relabeled sophisticated post-Keynesian asset theory as quantity theory (Friedman, 1956). On the consequences, Friedman won everything that mattered. The relabel reoriented the discipline toward money, and the stability claim became the empirical hinge of two decades of monetarism (Friedman, 1956). Patinkin was right that it was Keynesian portfolio theory in quantity-theory clothing; Friedman was right that the clothing was the point.

In the same 1956 volume, Cagan supplied the empirical muscle and, unknowingly, the era’s single point of failure. Across seven hyperinflations, a single semi-log money-demand function fits strikingly well with a common semi-elasticity of roughly 3 to 6, and it fits by pairing money demand with adaptive expectations, expected inflation as an exponentially weighted average of past inflation (Cagan, 1956). Two devices were born together in that volume: a stable money demand, which would do the monetarist work, and adaptive expectations, which would later be revealed to have been doing all the natural-rate work. Keep the second device in view; the whole era ends on it.

The Depression, retried

The stability conjecture’s first great field test was the founding evidence of Era I. A Monetary History’s Great Contraction chapter reconstructed the money stock month by month and found it fell by more than a third between the cyclical peak of August 1929 and the trough of March 1933, more than triple the largest prior decline in the authors’ near-century series, while over a fifth of commercial banks suspended operations (Friedman and Schwartz, 1963). The load-bearing sentence is a counterfactual: different, and feasible, actions by the monetary authorities could have prevented the decline in the stock of money (Friedman and Schwartz, 1963). The Depression was re-read from proof of impotence to proof of negligence. The Fed did not fail because monetary policy is weak; it failed because it did not use monetary policy.

What actually caused the Depression is a dispute in its own right, and Friedman-Schwartz won the profession’s reading without ending the argument. The credit side, built out later by Bernanke and Gertler, holds that collapsing asset values shredded borrower net worth, raised agency costs and the external finance premium, and choked investment in a self-feeding balance-sheet spiral that a stable money stock alone would not have stopped (Bernanke and Gertler, 1989). The gold-standard revisionists, Temin and Eichengreen, shifted causal weight toward the gold peg as the transmission channel of deflation. Per the record, both amendments refined rather than overturned: money remains the anchor cause in the profession’s telling, with credit and gold as the accepted amplifiers, and the counterfactual hardened into doctrine, all the way to Bernanke’s 2002 birthday concession to the authors: you’re right, we did it, we won’t do it again (Friedman and Schwartz, 1963). The 2008 response would operationalize all three readings at once.

The horse race, lost and won

The stability conjecture’s second test was a provocation. Friedman and Meiselman raced velocity against the multiplier over 1897-1958 and reported a rout: consumption correlates with money at R-squared near .97 against .57 for autonomous expenditure, the coefficient on autonomous spending comes out wrong-signed in all thirteen multiple regressions, and the one period where the multiplier looks stabler is the early 1930s, precisely the three or four years from which Keynesians had over-generalized (Friedman and Meiselman, 1963). The simple income-expenditure model, they concluded, was almost completely useless.

Ando and Modigliani took the race apart on three counts, and contemporaries largely conceded each. The “autonomous expenditure” variable was a grievous mis-derivation of the consumption function that silently stuffed retained earnings and the statistical discrepancy into disposable income; the result hinged on the unstable war years 1942-46, whose omission makes an overwhelming difference; and the money-consumption correlation was partly spurious common trend, since netting out a trend puts the income-expenditure model ahead on residual variance (Ando and Modigliani, 1965). Their killing blow was methodological: the correlation coefficient can be radically changed by a mere transformation of variables, so it was the wrong yardstick from the start (Ando and Modigliani, 1965).

Score the rounds separately, because they went to different sides. Ando-Modigliani won the technical round decisively. Friedman-Meiselman won the war. Even the victors’ own re-estimation conceded that money’s net contribution “remains significant,” and that concession moved the field’s question from “does money matter?” to “how much, and through what channel?”, which is the monetarist question (Ando and Modigliani, 1965). Friedman-Meiselman was empirically indefensible and historically decisive, seeding the St. Louis equation and, by Sims’s own later accounting, the reduced-form time-series tradition that became the VAR (Friedman and Meiselman, 1963). This pattern, the technical loser winning the war of ideas, recurs so often in this history that it is close to a law of the field’s motion.

The address

Friedman’s 1968 presidential address is the most consequential seventeen pages in this history, and its consequence is inseparable from its modesty. It contains no data and no regressions. It is framed as a demarcation exercise, separating what monetary policy cannot do from what it can, deflating the postwar hubris of fine tuning (Friedman, 1968). Its analytical equipment is two borrowings: Wicksell’s natural rate of interest, extended with Fisher’s nominal-real distinction, and, by explicit analogy, a natural rate of unemployment, the level ground out by the Walrasian system once actual labor-market frictions are embedded in it (Friedman, 1968). The natural-rate analogy is the prologue collecting a debt: Friedman reached back past Keynes to Wicksell to find the fulcrum for dislodging the Keynesian consensus.

The central proposition arrives on page eleven: there is always a temporary trade-off between inflation and unemployment; there is no permanent trade-off. The temporary trade-off comes not from inflation but from unanticipated inflation, which means holding unemployment below its natural rate requires inflation at an accelerating pace, exactly as pegging the market interest rate below Wicksell’s natural rate requires accelerating money growth (Friedman, 1968). The Phillips curve of Era I is thereby reclassified. It is not a menu; it is a temporary misperception, conditional on a fixed expected inflation rate, and it shifts up one-for-one as the public learns. Any menu the policymaker exploits is a menu the public withdraws.

Phelps arrived at the same structure independently and dressed it differently, deriving the vertical long-run curve as a maintained assumption, f(y*) = 0, inside an optimal-control problem the field promptly discarded while keeping the curve (Phelps, 1967). The priority question should be acknowledged rather than adjudicated: Phelps published in 1967 from a 1966 manuscript, Friedman delivered in December 1967 and printed in 1968, and the record supports independent and roughly simultaneous discovery from two directions, Phelps from optimal-growth theory, Friedman from the Wicksell-Fisher analogy (Phelps, 1967; Friedman, 1968). This history awards no first mover.

Both statements carried the same mechanical dependency: the curve shifts because expectations adapt, and the adaptation is backward-looking. Phelps was candid about the machinery, taking his mechanism directly from Cagan, an exponentially weighted average of past inflation; Friedman’s version ran through workers’ slow-adjusting failure to distinguish nominal from real wages, the same extrapolative logic without the formula (Cagan, 1956; Phelps, 1967; Friedman, 1968). The era’s crown jewel was built on a device introduced as an estimation convenience.

The address also carried a policy prescription that history treats less kindly: a publicly announced steady growth rate of a monetary total, roughly 3 to 5 percent a year, the k-percent rule (Friedman, 1968). The rule died in the 1980s when money demand destabilized and monetarist targeting failed operationally. The demarcation survived. No long-run trade-off, so anchor policy to a nominal target: that sentence underwrites inflation targeting to this day, which means the address’s deepest legacy is an institution, not a rule.

Scoring the menu fight

The Phillips-curve dispute now had both sides fully armed, and the 1970s adjudicated it. Round one, the theory, went to Friedman and Phelps before any data arrived; the logic that a trade-off built on fooling people cannot survive the people’s learning needed no sample. Round two went to them on the evidence: stagflation, rising inflation and rising unemployment together, the one configuration a stable menu cannot produce and the accelerationist hypothesis predicts, the prediction that became the single most consequential macro forecast of the century (Friedman, 1968). Round three belongs to the historians, and it cuts the other way: the menu side was partly libeled, since Samuelson and Solow’s text was hedged and short-run and Phillips asserted no menu at all (Samuelson and Solow, 1960; Phillips, 1958). The natural-rate side won the economics and flattened its opponents into a strawman in the same motion. Both facts are true, and a history that reports only one of them is doing public relations.

The loose thread

Adaptive expectations was doing all the work. The era’s terminal achievement, the vertical long-run curve and the accelerationist prediction, rests entirely on how the public forms expectations, and the mechanism in use was purely backward-looking (Cagan, 1956). That is defensible while the policy regime is stable, but it contains a concession fatal to the era’s own logic: if agents extrapolate only from the past, a sufficiently clever policymaker can stay one step ahead of them forever, and the natural-rate result degrades into a claim about learning speed. The next era pulls exactly this thread, replaces the backward-looking average with agents who forecast the policy rule itself, and reaches a conclusion the counterrevolution never wanted: not just no long-run trade-off, but no exploitable short-run trade-off either, and any systematic policy neutral by construction. Cagan’s own prize episodes would be turned against him, re-read by Sargent as evidence that the hyperinflations ended abruptly when fiscal regimes changed, faster than adaptive learning permits (Cagan, 1956). The counterrevolution’s answer was right; its microfoundation was the weak link, and the demolition crew was already at work.

Era III: The Rational Expectations Purge (1972-1985)

The episode is the Great Inflation itself, now in full flower. Through the 1960s the profession had sold policymakers an exploitable Phillips curve; the 1970s delivered unemployment and inflation both above 7 percent by 1975, the configuration the menu said could not happen. The failure was read not as a puzzle but as a verdict, and the verdict fell on an entire way of doing economics: the large Keynesian macroeconometric models had failed at the one task they were built for, forecasting the consequences of policy, and the failure was structural rather than a mis-estimated coefficient. Era III treats stagflation as double vindication, of Friedman and Phelps on substance and of a new methodological program on method.

The era’s answer to the neutrality question, stated at full contestable strength: systematic, anticipated money is neutral even in the short run, so it does not matter at all; only monetary surprises move output; and the era’s limiting case pushes further still, building business cycles with no money in them whatsoever. This is the hardest neutrality answer in the field’s history, harder than the prologue’s, and the claim is contestable in a way the era itself half-knew: it is a theorem about flexible-price economies, and the era bet everything on the antecedent. The theorem was true. The models still lost, and where they lost matters for everything that follows.

One idea, four detonations

The era is best read not as four discoveries but as one logical move, “agents see through the policy,” applied to four objects, each application becoming a canonical result.

The possibility proof came first. Lucas built an explicit general-equilibrium economy, overlapping generations trading across two physically separated island markets, in which a Phillips curve emerges from optimizing, market-clearing agents with rational expectations, and no money illusion or ad hoc stickiness anywhere in sight (Lucas, 1972). An agent observes only his local price, which confounds two movements he cares about differently: a relative-demand shift, to which he should respond by producing more, and an aggregate monetary shift, to which he should not respond at all. The signal-extraction problem delivers a supply curve in which output responds only to the surprise component of prices, with a response coefficient equal to the fraction of local price variance that is genuinely relative: the noisier the monetary regime, the flatter the curve (Lucas, 1972). The trade-off exists and is not exploitable, because any systematic policy is forecast and undone. Lucas presented this as a modest existence proof in the Journal of Economic Theory, a demonstration that the Friedman-Phelps intuition could survive rigorous general equilibrium. The field read it as a mandate, and the gap between what the paper proved and what the profession did with it is one of the great overreadings in the history of economics (Lucas, 1972).

The second detonation aimed the logic at the money-supply rule. Sargent and Wallace dropped rational expectations and a Lucas supply curve into an otherwise conventional Keynesian structure and derived the policy-ineffectiveness proposition: the distribution of output is independent of the deterministic money rule, so activist demand management is futile by construction and the best the central bank can do is stabilize prices (Sargent and Wallace, 1975). The third aimed it at government debt. Barro showed that with an operative chain of intergenerational bequests, households act as though infinite-lived, capitalize the future taxes implied by today’s deficit, and save the tax cut, so government bonds are not net wealth and deficits do not stimulate demand (Barro, 1974). The fourth, which deserves its own section below, aimed it at the policymaker’s own future self.

Ricardian equivalence is worth pausing on, because the profession’s memory of it inverts the paper. Barro framed the result narrowly and defensively, as a theoretical audit of an assumption, and devoted most of the paper to the four channels that break it: finite horizons without operative bequests, imperfect private capital markets, a government monopoly in supplying bond liquidity services, and uncertainty about future tax liabilities, hedging explicitly that a negative effect of debt was a priori as plausible as a positive one (Barro, 1974). The field remembers the neutrality benchmark and not the ambiguity. The follow-up literature, Kotlikoff, Bernheim, Feldstein, Hubbard, decided against strict equivalence on exactly the frictions Barro flagged, and decades later the heterogeneous-agent literature gave the rejection teeth, showing that high-MPC hand-to-mouth households spend transfers rather than saving them against future taxes (Barro, 1974; Kaplan et al., 2018). Equivalence stands as the frictionless null, pedagogically permanent, descriptively rejected, and rejected on its own author’s terms.

The critique

Lucas’s 1976 essay is the era’s methodological death certificate, and it earns multi-page treatment because it is the rare paper whose one syllogism reorganized what counts as knowledge in a field. The syllogism, stated on his page 41: the structure of an econometric model consists of the optimal decision rules of economic agents; optimal decision rules vary systematically with the structure of the series agents are trying to forecast; therefore any change in policy systematically alters the structure of the estimated model (Lucas, 1976, p. 41). The corollary is total: simulations of alternative policies using estimated models are invalid regardless of the performance of those models over the sample period or in short-term forecasting (Lucas, 1976, p. 41). Good fit is no defense. The models being condemned, Wharton, FRB-MIT-Penn, Brookings, were the operational core of 1970s policymaking, and the essay told their operators that the machines could forecast but could never answer a counterfactual.

The critique’s power came from its worked examples, and the sharpest is fiscal rather than monetary. Hall-Jorgenson investment analysis treated a tax credit as permanent when the credit was in fact switched on and off stochastically; correcting for the policy process, the true stimulus of a transitory credit exceeds the permanent-credit estimate by a ratio of (2 + r − δ)/(r + δ), roughly 7.3 at Lucas’s parameter values, and about 4.5 for a credit off five years and on one (Lucas, 1976). A reduced-form coefficient is a function of the policy regime; change the regime and the coefficient walks. Applied to the Phillips curve, the point diagnosed the decade: over a sample where the inflation process is stable, the estimated equation traces a stable trade-off, and a sustained rise in inflation then shifts the whole relation adversely, which is why the models justified inflationary policy right up until the policy destroyed the correlation that justified it (Lucas, 1976).

Two ironies attend the reception. Lucas pitched the essay as intellectual housekeeping, disclaiming novelty, saying there was little in it not implicit in Friedman, Muth, and still earlier in Knight (Lucas, 1976, p. 20); the profession received it as a foundational rupture and named a research program after him. And the essay’s practical demand, that macro be rebuilt from deep taste-and-technology parameters invariant to policy, licensed a modeling ambition far exceeding anything the syllogism strictly required. The critique proved that reduced forms cannot answer counterfactuals. It did not prove that anyone knew the deep parameters either, a gap the next era’s calibrators would step over rather than close.

Sims’s 1980 “Macroeconomics and Reality” is conventionally shelved with the critique as a second demolition of the big models, and it did agree that their claims to identification “cannot be taken seriously” (Sims, 1980, p. 1). But Sims was a genuine third position, not an ally, and the fight among the three positions is the era’s second great dispute. His target was broader than Lucas’s: the “incredible” exclusion and exogeneity restrictions of the Cowles tradition were imposed for statistical convenience rather than derived from theory, and, he insisted, the identifying restrictions of Lucas’s own program were no better grounded (Sims, 1980). His evidence was formal: the joint exogeneity of unemployment and real GNP in a conventional wage equation is rejected for the United States at chi-squared(32) = 58.26 (Sims, 1980, p. 32). His constructive alternative was the vector autoregression, all variables treated symmetrically as endogenous, dynamics summarized by impulse responses and innovation accounting.

Score the three-way fight as the record does. Lucas won the negative claim overwhelmingly; reduced-form policy simulation died. Sims won a research program, the VAR became the empirical workhorse of macroeconomics and of every central bank research department, but he lost the philosophy behind it: “atheoretical” VARs turned out to need identifying assumptions to answer any causal question, and the structural-VAR literature reintroduced exactly the identification Sims had questioned, some of it in his own later work (Sims, 1980). Neither purist position survived contact with practice. And buried in Sims’s own illustrative six-variable system is the era’s time bomb: a money innovation is followed by a rise in output and a fall in unemployment (Sims, 1980). The era’s theorems said anticipated money is neutral; the era’s own empirical method showed monetary surprises carrying large, persistent real dynamics. The two results coexisted, unreconciled, in the same journals in the same years.

Why the inflation happened: the government as the agent seen through

Kydland and Prescott’s 1977 paper turned the see-through logic reflexive, aiming it at the policymaker’s own future reoptimization, and produced the era’s answer to why the Great Inflation happened at all. Even a benevolent policymaker with a fixed objective and perfect control produces the wrong outcome, because the optimal plan is time-inconsistent: each period the authority is tempted to exploit the Phillips curve it promised not to exploit, the public knows this, and the discretionary equilibrium delivers the same unemployment at strictly higher inflation, point C on a worse indifference curve than price stability (Kydland and Prescott, 1977). Barro and Gordon then converted the example into the workhorse: in a stochastic natural-rate model the discretionary regime’s equilibrium inflation is a formula, rising with the natural rate and with the employment gap the authority covets, positive inflation purchased for zero employment gain, and the model became the standard positive explanation of the Great Inflation (Barro and Gordon, 1983).

Both papers were canonized against their authors’ framing, an identical reversal the record flags in each: Kydland-Prescott began as a warning to optimal-control modelers, Barro-Gordon as a positive description of Fed behavior, and the field read both as normative charters for binding commitment (Kydland and Prescott, 1977; Barro and Gordon, 1983). The rules side of the rules-versus-discretion dispute thereby won the theory round outright. The rebuttal from qualified discretion, that rigid rules discard stabilization and the optimum is an institution rather than a formula, was live and serious, but its resolution belongs to the next era, where Rogoff and Taylor supply it.

Sargent and Wallace’s 1981 “unpleasant arithmetic” closed the era’s policy logic with its most subversive result: under fiscal dominance, with the real rate above the growth rate, the central bank may not control inflation at all. A monetary tightening today means more bonds outstanding, whose interest burden grows faster than the tax base and must eventually be monetized, so tighter money now can mean higher inflation now, through forward-looking expectations (Sargent and Wallace, 1981). As pitched, this was a narrow, conditional debating point against monetarism, published in the Minneapolis Fed’s own journal at the exact moment Volcker was testing whether resolute money could conquer inflation while deficits widened (Sargent and Wallace, 1981). The caveat outgrew the doctrine it qualified. The paper became the founding text of fiscal dominance and, eventually, the intellectual charter for the view that the central bank is not, ultimately, in charge of inflation, a claim far larger than the original arithmetic and one that would be re-mobilized, by name, for 2021 (Sargent and Wallace, 1981). Monetarism’s defense held on its own terms: the result bites only under specific and avoidable regime assumptions, and Volcker’s success became its standing counterexample. Both readings persist, which is why this dispute reopens twice more in this history.

The limiting case

Kydland and Prescott’s second paper completed the purge by changing the subject. Time to Build constructs a frictionless competitive growth model driven only by technology shocks, no money, no nominal rigidity, no demand disturbance, calibrates its roughly seven free parameters from national-income accounting and micro evidence rather than estimating them, and matches the model’s filtered second moments against the postwar data’s: model output standard deviation 1.80 percent against 1.8 in the data, investment three times as volatile as output, consumption far smoother, both strongly procyclical (Kydland and Prescott, 1982). Two ingredients are shown essential, multi-period time-to-build (adjustment costs demonstrably cannot substitute for it) and non-time-separable leisure, needed to make hours fluctuate about 18 percent more than productivity, which the separable case gets backwards (Kydland and Prescott, 1982). The paper’s own tone is technical and modest. The reading the profession attached to it was not: business cycles as the Pareto-efficient equilibrium response to real shocks, requiring no stabilization and no monetary theory at all.

I read the RBC episode as a boundary case of the purge rather than a separate research program, because its force in monetary economics was always the force of a limit: it showed how much of the cycle could be explained with money set exactly to zero, and thereby converted “does money matter?” from a rhetorical question into a quantitative one with a demanding null. A paper that never mentions money became foundational to monetary economics by negation (Kydland and Prescott, 1982). Its durable contribution, the record is blunt on this, was the calibration-and-moment-matching protocol rather than the time-to-build mechanism, which later RBC work quietly dropped; the protocol became the modeling language even of the program built to refute it (Kydland and Prescott, 1982).

The era also settled, in passing, the expectations dispute inherited from Era II. Rational expectations won the methodological revolution, and the adjudicating evidence was taken from the adaptive side’s own exhibit: Sargent’s re-reading of the great hyperinflations found they stopped abruptly when fiscal regimes changed, faster than any backward-looking learning rule permits, turning Cagan’s seven episodes into witnesses for the prosecution (Cagan, 1956; Lucas, 1972; Sargent and Wallace, 1975). The caveat the record insists on: the island mechanism that generated Lucas’s results was itself quietly abandoned as implausible once everyone conceded that aggregate price data are published with short lags, so the profession kept the method and the conclusions while discarding the story that produced them (Lucas, 1972).

The loose thread

The purge proved a conditional theorem, anticipated money is neutral when prices clear continuously, and then bet the field on the antecedent. The bet failed in two places the flexible-price models could not reach. First, output effects: the money-free model needed specific real machinery to match amplitudes and persistence (Kydland and Prescott, 1982), while the era’s own empirical tool showed monetary surprises producing sizeable, persistent real responses that a money-free economy has no way to generate (Sims, 1980), and the identified-shock literature about to arrive would make those responses impossible to dismiss. Second, the mechanism: with the island story dead, the purge had conclusions without a channel (Lucas, 1972). The opening this left was precise. If money is neutral only under continuous market clearing, and the data reject a money-free cycle, then the productive response is not to abandon rational expectations, the purge’s permanent achievement, but to graft a nominal friction onto the purge’s own dynamic-equilibrium scaffolding and re-litigate a single assumption: flexible prices. That is exactly what happened next, and the neutrality answer flipped sign on schedule.

Era IV: The Reconstruction (1983-2007)

The episode this era answers for is the Great Inflation and its conquest, cashed out over two decades as the triumphalism of the Great Moderation. The era’s answer to the neutrality question, stated as a claim to contest: money is non-neutral in the short run because prices are sticky, and the neutrality problem is solved as an engineering matter, since a central bank following a Taylor rule with an inflation coefficient above one, insulated from politics by independence and a conservative mandate, delivers a determinate, inflation-stabilizing equilibrium. Solved. That is the word to contest, and the era’s own architecture shows where: the solution assumed a representative agent whose Euler equation carried all of transmission, a frictionless financial sector, and a policy rate that could always be cut. All three assumptions failed in the same year.

Volcker first: the experiment that preceded the theory

Before the reconstruction had its models, it had its experiment, and the sequence matters for how one reads everything after. In October 1979, one of the six months Romer and Romer would later identify as deliberate anti-inflationary policy shocks, the Federal Reserve under Volcker chose to induce a recession to lower inflation (Romer and Romer, 1989). What such a decision does to an economy is now measured: on the Romers’ estimates, an anti-inflationary shift produces an ultimate fall in industrial production of about 12 percent and an ultimate rise in unemployment of about 2 percentage points, with the maximum effect arriving after roughly 33 months (Romer and Romer, 1989). The Volcker episode ran the full course: inflation above 13 percent in 1980 fell to around 4 percent by 1983, and unemployment passed 10 percent in 1982, the highest since the Depression, before the disinflation held. This is the sense in which monetary economics differs from its neighboring fields: the natural-rate hypothesis, the credibility literature, and the primacy of nominal anchors were not academic positions awaiting citation counts. They were acted on, at a cost of millions of jobs, and the disinflation’s success became a data point that disciplined every subsequent theory. A field whose ideas are applied at that scale does not get to treat its disputes as parlor games, and mostly it has not.

The regime change also left a diagnostic fingerprint the era’s theory would later formalize. Clarida, Galí, and Gertler estimated the Fed’s inflation response at 0.83 before Volcker and 2.15 after (Clarida et al., 1999). Below one, and the real rate falls as inflation rises, feeding the spiral; above one, and the system anchors. The Great Inflation, on this reading, was not bad luck. It was a violated stability condition, and the condition had not yet been written down when it was violated.

Making stickiness respectable

The reconstruction’s first intellectual task was to answer Lucas on his own terms. The purge’s standing objection to sticky prices was the “$500 bills on the sidewalk” test: if adjusting a price is profitable, rational firms adjust, so nominal rigidity assumes irrationality. The answer came in two layers, an engine and a legitimacy proof, and the field needed both.

The engine came first. Taylor showed that staggered, fixed-length nominal contracts generate a contract multiplier, persistence in real effects extending well beyond any single contract’s length, under fully rational expectations (Taylor, 1980). Calvo replaced the fixed calendar with a memoryless constant-hazard reset, the device every graduate student now meets as the Calvo fairy, purchasing the aggregation tractability that delivers the clean forward-looking New Keynesian Phillips curve (Calvo, 1983). Which friction was right became its own dispute, and its scoreboard is instructive: Calvo won adoption on tractability alone, his convenience assumption becoming the supply block of essentially every central-bank DSGE model, while Taylor’s staggering remained the microfoundation blueprint taught through Calvo’s cleaner algebra (Taylor, 1980; Calvo, 1983). The winning device was also the vulnerable one. Calvo pricing implies that the timing of a price change is independent of its size, a counterfactual that drew heavy fire after 2008’s missing disinflation and again after 2021, so the era’s most-used equation rested on its least defensible assumption (Calvo, 1983).

Neither Taylor nor Calvo answered why a rational firm tolerates a stale price, and that is what the legitimacy layer supplied. Mankiw’s menu-cost argument turned on a wedge: the firm’s private loss from a stale price is a second-order Harberger triangle while the social loss is first-order, with the social-to-private ratio around 23 for a 1 percent shock and above 200 at low elasticities, so a menu cost far below the social loss it permits still exceeds the private gain from adjusting (Mankiw, 1985). Akerlof and Yellen proved the same structure by the envelope theorem: at the pre-shock optimum, the derivative of profits with respect to the price is zero, so inertia costs the firm only second-order losses while moving employment first-order (Akerlof and Yellen, 1985). Small frictions, large cycles, no irrationality. The sidewalk objection was answered on its own terms: the $500 bill was lying on society’s sidewalk, not the firm’s.

Then came the discriminating test, the piece of evidence that separated the sticky-price account from Lucas’s imperfect-information account rather than merely coexisting with it. In the menu-cost world, higher mean inflation makes firms adjust more often, so the output-inflation trade-off should shrink as trend inflation rises; in the signal-extraction world, mean inflation is irrelevant. Ball, Mankiw, and Romer took the prediction to 43 countries and confirmed it, the trade-off roughly twice as large at 5 percent inflation as at 10 (Ball et al., 1988). This is the New Keynesian side of the era’s most rhetorically hostile fight, the exchange with real business cycle theory, and the final scoreboard should be stated without diplomacy. RBC won the tools; the New Keynesians conceded the modeling language and built their rigidities on exactly the Kydland-Prescott dynamic-equilibrium scaffolding, which is what produced DSGE (Kydland and Prescott, 1982). The New Keynesians won the substance: money matters, and the identification evidence sealed it, with the Romers’ narrative dates yielding the 12 percent industrial-production decline already cited Romer and Romer (1989) and their later forecast-purged measure a 4.3 percent output fall (Romer and Romer, 2004). Each side kept the other’s concession. Modern macro is RBC machinery running New Keynesian frictions, a truce so complete that students no longer notice it was ever a war.

The Romer-Romer paper also opened the era’s identification dispute, VARs against the narrative approach, and its outcome was a merger dressed as a fight. The narrative side’s claim was that statistical orderings cannot separate policy from its anticipation: money can lead output for non-causal reasons, and an actively stabilizing Fed can erase the money-output correlation entirely even when money matters greatly (Romer and Romer, 1989). Their remedy was to read FOMC minutes and isolate months of deliberate disinflationary intent, the six Romer dates. The reversal is that the six dates were promptly folded into VARs as an off-the-shelf shock series, the narrative method becoming the input the statistical method lacked (Romer and Romer, 1989). A paper skeptical that statistical methods could identify monetary causation ended up supplying the identifying variation those methods ran on for a generation.

The institutional fix

The rules-versus-discretion dispute, inherited from Kydland-Prescott and Barro-Gordon with the theory round already scored for rules, received its institutional resolution in this era, and the resolution was subtler than the slogan that survived it. Rogoff proved that society gains by appointing a central banker more inflation-averse than itself, which reads like a rules victory and is not quite one: the optimal conservatism is strictly interior, finite, because as the banker’s inflation-aversion rises the marginal inflation gains vanish faster than the stabilization losses recede, so the residual stabilization cost eventually dominates and it is never optimal to appoint someone who cares nothing for unemployment (Rogoff, 1985). Taylor supplied the operational form, the reaction function r = p + 0.5y + 0.5(p − 2) + 2 with its inflation coefficient of 1.5, and showed it tracked actual 1987-92 Fed policy remarkably well (Taylor, 1993). Both papers were canonized past their own hedges. Rogoff’s finiteness caveat was dropped in the retelling; Taylor labeled his equation hypothetical and representative, insisted judgment could not be replaced, and demonstrated that the rule should be overridden for the 1990 oil shock when futures prices signaled the disturbance was temporary (Rogoff, 1985; Taylor, 1993). What won, then, was a qualified-rules synthesis, commitment embodied in institutions rather than formulas, and the qualifier proved load-bearing: after 2008, Rogoff’s discarded warning returned as the accusation that rigidly inflation-focused mandates had tolerated too much unemployment (Rogoff, 1985).

The institutions followed the theorems with unusual speed. New Zealand adopted the first formal inflation target in 1990, and within a decade explicit targets, independent central banks, and hawkish appointments were the international default, the Rogoff paper functioning as the intellectual charter for the wave (Rogoff, 1985). Volcker and the Bundesbank were the existence proofs; the theory explained why they worked and told everyone else to copy them. Inflation in the advanced economies fell and stayed down, output volatility fell with it, and the period acquired its name. The Great Moderation was, on the era’s self-reading, the theorems operating at scale.

Woodford: the synthesis as treatise

Every consensus eventually writes its summa, and this one’s is Woodford’s Interest and Prices, the 2003 treatise whose title repays Wicksell across a century (Woodford, 2003; Wicksell, 1936). Its ambition deserves to be stated at full size, because both the triumph and the failure live there. Modern central banks steer with a short-term interest rate and increasingly ignore monetary aggregates, yet the inherited theory of the price level was a money-supply-and-demand theory. Woodford asked whether the practice could be given rigorous foundations with no essential reference to money at all, and answered yes, twice over (Woodford, 2003).

The first result is price-level determinacy without money: in a fully cashless economy, where money is only a unit of account, an interest-rate rule pins down a unique bounded equilibrium for inflation, so there is no logical inconsistency in a nonmonetarist theory of the general price level (Woodford, 2003). The second is the Taylor principle as theorem: determinacy requires the central bank to move its rate more than one-for-one with inflation, and a rule violating this admits an enormous set of bounded equilibria, sunspot fluctuations driven by self-fulfilling expectations unrelated to fundamentals (Woodford, 2003). The empirical coefficient Taylor had fit to 1987-92 data and the estimated regime break at 1979 thereby acquired the status of a stability law: one number, 1.5 against 0.83, was simultaneously the era’s policy prescription, its determinacy condition, and its historical diagnosis of the Great Inflation (Taylor, 1993; Clarida et al., 1999; Woodford, 2003). Around the determinacy core, Chapter 4 builds the neo-Wicksellian engine, Calvo pricing on optimizing foundations, a forward-looking IS curve, and the gap between the actual real rate and a time-varying natural rate as the single state variable through which policy has any effect at all (Woodford, 2003). Central banking, on this account, is the management of expectations under commitment. Clarida, Galí, and Gertler had already assembled the same three equations under the confident title “The Science of Monetary Policy” (Clarida et al., 1999); Galí standardized the apparatus into the textbook every central-bank research department taught from (Galí, 2008). Rogoff’s institutional fix reappears inside the model as a derived result, a low-weight central banker shrinking the discretionary bias, so the institutional-design literature and the DSGE core were one program in two registers, which is precisely why the era could believe policy was solved (Rogoff, 1985; Clarida et al., 1999).

Read now, the treatise is the high-water mark of a consensus and the map of its blind spots, and the two are the same pages. The cashlessness that was the book’s rhetorical coup left it thin on financial intermediation, balance sheets, and liquidity (Woodford, 2003). The interest-gap engine presupposed the rate could always be set freely. And transmission ran entirely through one representative household’s Euler equation (Woodford, 2003; Galí, 2008). The treatise did not fail to anticipate 2008 by carelessness; it failed by construction, because everything 2008 required had been abstracted away as inessential to the questions the Great Moderation was asking.

The dissents filed as footnotes

Two lines of dissent ran under the consensus for its entire life, and the era’s real indictment is not that it lacked the ideas 2008 needed but that it had them and shelved them.

The credit view was fully built. Bernanke and Gertler made borrower net worth a state variable through costly state verification: falling asset values shred net worth, agency costs and the external finance premium rise, credit and investment collapse, and the collapse feeds back into asset values, Fisher’s debt-deflation given rigorous microfoundations (Bernanke and Gertler, 1989). Bernanke, Gertler, and Gilchrist then embedded the financial accelerator inside a Calvo-sticky-price DSGE with a Taylor rule, a crisis model living inside the consensus machinery (Bernanke et al., 1999). Both papers framed the mechanism as amplification of ordinary business-cycle shocks, a refinement for normal times rather than a theory of catastrophe, and the consensus filed it accordingly (Bernanke and Gertler, 1989; Bernanke et al., 1999). The template sat on the shelf until the exact person who built it was running the Federal Reserve when the mechanism detonated.

The fiscal view attacked the consensus at its foundation, the nominal anchor itself. While Woodford proved the interest-rate rule anchors the price level, Leeper’s four-region taxonomy showed the anchor is conditional on a regime pairing: a unique bounded equilibrium requires exactly one active and one passive authority, and the region with passive money and active fiscal policy, where the government’s intertemporal budget constraint becomes the equation that pins the price level, is exactly as coherent as the monetarist case (Leeper, 1991). Sims’s nonlinear analysis pushed the inversion further: fixing the money stock can leave the price level indeterminate while pegging the nominal rate can deliver uniqueness, the opposite of monetarist intuition, leading to his formulation that inflation is a fiscal phenomenon even more fundamentally than a monetary one (Sims, 1994). The paternity of the fiscal theory is itself disputed, and this history follows its rule of acknowledging rather than adjudicating: Leeper’s taxonomy, Sims’s model, and Woodford’s contemporaneous papers each ground a claim to origination, and all of them descend recognizably from Sargent-Wallace’s unpleasant arithmetic (Leeper, 1991; Sims, 1994; Sargent and Wallace, 1981).

Buiter attempted the knockout, and the exchange is the era’s best example of a dispute that ended in a genuine stalemate rather than diplomatic mush. His charge: the government’s intertemporal budget constraint is an identity that must hold for all admissible price paths, not an equilibrium condition that selects among them, and treating it as the latter over-determines the system and generates absurdities, a theory that can price “phlogiston” money nobody holds, and a symmetric “household theory of the price level” that fixes consumption arbitrarily (Buiter, 2002). Buiter believed he had shown the FTPL emperor wore no clothes. The field did not accept the knockout. The critique forced the fiscal theorists to state precisely what object the budget constraint is; they did, the theory survived and was later mainstreamed, and the core question, identity or equilibrium condition, was never resolved by fiat (Buiter, 2002). Buiter’s paper stands as the strongest brief for the prosecution in a case that stayed open, and the case’s reopening is the next era’s business.

The loose thread, in triplicate

The reconstruction closed declaring victory, and its own documents identify the three omissions that came due simultaneously. No financial sector: the cashless coup left the workhorse without intermediaries, while the accelerator waited on the shelf (Woodford, 2003; Bernanke and Gertler, 1989; Bernanke et al., 1999). No zero bound: the apparatus presupposed a freely adjustable policy rate, and the concluding discussion of the bound in the era’s own survey, motivated by Japan, was nearly an aside that turned out to be prophetic (Clarida et al., 1999). One representative agent: transmission was an Euler equation, undisciplined by any microdata on which households actually respond to rates (Woodford, 2003; Galí, 2008). In 2008 the financial sector failed, the policy rate hit zero, and the fiscal response made household heterogeneity first-order, all within roughly twelve months. The era had written its successor’s toolkit, the accelerator, the ZLB analysis, the fiscal theory, and filed each as a footnote to a solved problem. The neutrality answer was about to flip again, this time in the middle of a burning building.

Era V: Two Humiliations (2008-present)

This era answers for two episodes, and their doubleness is its structure: the Great Financial Crisis with the zero-rate decade it opened, and the 2021 inflation, the first consensus-breaking inflation since the 1970s. The era’s answer to the neutrality question, stated as one contestable claim spanning both: money matters, but not where the synthesis had put it, not through the short rate and the representative household’s Euler equation but through expected policy paths, heterogeneous balance sheets, and a supply constraint whose shape the models had assumed away. The claim is contestable in each half separately, which is why each humiliation below states its own version and names its own opposition. What is not contestable is that the two episodes struck the same synthesis from opposite directions, one exposing what the workhorse had abstracted away, the other what it had mis-specified.

The ZLB decade

The first humiliation was the Great Financial Crisis and the decade of zero rates that followed. The workhorse the profession had spent thirty years building could not represent a crisis that originated inside financial intermediation, because it had no intermediaries (Gertler and Kiyotaki, 2010); it had nothing to say once the funds rate hit zero in December 2008, because it assumed the rate could always be cut (Eggertsson and Woodford, 2003; Gertler and Karadi, 2015); and it mislocated the transmission mechanism the moment fiscal transfers became the live instrument, because it had one representative household (Kaplan et al., 2018; McKay et al., 2016). Three abstractions, all traceable to the cashless coup, all biting in the same year.

The decade’s answer to the neutrality question, stated as a claim to contest: when the short rate is pinned at zero, conventional policy stops mattering through its textbook channel, and what still moves the economy is not the balance sheet as such but the expected future path of policy; and even that path reaches households far more weakly than the representative-agent model claims, because most of monetary transmission was never intertemporal substitution in the first place but general-equilibrium income flowing to a heterogeneous, high-MPC population. The claim is contestable from two standing positions the decade never closed out: the Fed’s own portfolio-balance account of QE, and the fiscal theorists’ insistence that the price level was never a monetary object at all (Krishnamurthy and Vissing-Jorgensen, 2011; Sims, 2013; Cochrane, 2023). What is not contestable is the double demotion the decade executed: the balance sheet demoted in favor of expectations, and the Euler equation demoted in favor of income.

The strangest feature of the decade is that its playbook was written before the crisis, for someone else’s crisis. Eggertsson and Woodford, staring at Japan in 2003, derived the two results that governed the next fifteen years. First, an irrelevance proposition: the equilibrium paths of prices, output, and asset prices are unaffected by the size or composition of the central bank’s balance sheet, provided the operations change no expectations about future policy, so quantitative easing, as such, does nothing (Eggertsson and Woodford, 2003). Second, the constructive counterpart: the zero bound is a problem of expectations management, and optimal commitment is history-dependent price-level targeting, a target that ratchets up whenever the bound forces an undershoot, promising to bring prices back above the path they would have followed, which raises expected inflation and cuts the real rate exactly when the economy needs it (Eggertsson and Woodford, 2003). Their calibration made the stakes concrete: a natural rate stuck at minus 2 percent for an average of ten quarters produces, under a fixed zero-inflation target, an output gap around 12 percent and deflation near 9 percent a year, losses the history-dependent commitment mostly eliminates (Eggertsson and Woodford, 2003). Written as a warning that balance-sheet policy is impotent, the paper was received after 2008 as the instruction manual for the one thing that works: credible commitment became forward guidance, and the ratcheting price-level target became the direct ancestor of the Fed’s 2020 average-inflation-targeting framework (Eggertsson and Woodford, 2003). Krugman’s black-hole pessimism and the QE enthusiasts’ balance-sheet optimism were both rejected in the same document, five years early.

QE happened anyway, at trillion-dollar scale, and the empirical decomposition of what it did is the decade’s first live dispute. Krishnamurthy and Vissing-Jorgensen took the announcement-day evidence apart channel by channel: QE1’s two-day moves reached minus 107 basis points on the ten-year Treasury and minus 200 on ten-year agencies, with an inflation channel (expected inflation up 96 to 146 basis points) and a signaling channel (about 40 basis points off the 24-month funds-futures contract), while QE2’s effects were an order of magnitude smaller and ran almost entirely through signaling (Krishnamurthy and Vissing-Jorgensen, 2011). Two negatives carried the argument. They found no support for the duration-risk channel, the classic portfolio-balance story and the Fed’s own preferred framing, leaving a genuine unresolved dispute with the Gagnon-lineage view; and since QE2 worked mainly by signaling, they asked in print whether a rate-path commitment could have achieved the same without putting the balance sheet at risk (Krishnamurthy and Vissing-Jorgensen, 2011). Theory and event study thus converged from independent directions on the same demotion: Eggertsson-Woodford’s irrelevance proposition became the foil QE advocates argued around, and KVJ supplied the matching evidence (Eggertsson and Woodford, 2003; Krishnamurthy and Vissing-Jorgensen, 2011). What survived of QE was asset-specific: a safety channel reflecting a clientele for extremely safe long assets, and the MBS-specific channels of QE1, which is why buying Treasuries never moved mortgage rates the way buying mortgages did (Krishnamurthy and Vissing-Jorgensen, 2011). The safety channel, offered as one of six, outgrew the paper and seeded the safe-asset and convenience-yield subfield. Meanwhile the model-builders re-installed what cashlessness had removed: Gertler and Kiyotaki built the intermediary balance-sheet constraint that rationalized the credit-policy interventions (Gertler and Kiyotaki, 2010), and Cúrdia and Woodford, grafting borrowers, savers, and an endogenous spread onto the NK skeleton, showed that “lean against spreads” is not a robust rule, the optimal response swinging from near one for a financial shock to negative for persistent real shocks (Cúrdia and Woodford, 2010).

The identification war matured in parallel, and its trajectory is a lesson in how technical victories reopen deeper problems. Kuttner had shown in 2001 that splitting a rate change into anticipated and surprise components using funds futures roughly doubles the measured yield response, 24 basis points becoming 50 at the three-month maturity, and rescues significance across the curve (Kuttner, 2001). Romer and Romer showed that the raw funds rate produces a strong price puzzle, prices rising after tightening for four straight years, while their forecast-purged series delivers clean and larger effects, a 4.3 percent output fall and a 6 percent price-level decline, direct evidence that anticipation and endogeneity, not weak policy, had produced the earlier nulls (Romer and Romer, 2004). Gertler and Karadi fused the two traditions, high-frequency surprises as external instruments in a VAR, with a first-stage F around 17 to 21, and showed transmission running through term premia and credit spreads, with forward-guidance content doing much of the work (Gertler and Karadi, 2015). High-frequency identification won the toolkit war outright; the Kuttner surprise and the GK instrument became the off-the-shelf shocks of hundreds of papers (Kuttner, 2001; Gertler and Karadi, 2015).

Then the winning method disgorged an anomaly that reopened everything. Nakamura and Steinsson, measuring non-neutrality in thirty-minute windows around FOMC announcements, found that after a contractionary surprise, survey forecasts of output growth rise, the wrong sign for a pure policy shock, robust across every sample cut, implying that around two-thirds of the real-rate response is news about the natural rate rather than tightening relative to it (Nakamura and Steinsson, 2018). The Fed information effect, treated in the paper as a sideshow to be modeled away, became the main event of the next decade’s identification debate, with Bauer and Swanson mounting the skeptical rebuttal that the anomaly reflects the Fed responding to public news markets also see (Nakamura and Steinsson, 2018). The dispute is genuinely open, and its shape should be appreciated: the high-frequency program was invented to close the endogeneity between policy and the economy, and at its technical peak it discovered the endogeneity again, one level down, between policy and beliefs about the economy. Romer and Romer’s organizing principle, control for the central bank’s own information set, turned out to be the demanding standard all along (Romer and Romer, 2004).

The deepest repair of the decade began as a quantitative embarrassment. Del Negro, Giannoni, and Patterson noticed that the representative-agent Euler equation does not discount far-future promises, so the canonical model implies a rate cut promised five years out moves current inflation many times more than a cut today, and the standard estimated DSGE predicts explosive output at an eight-to-nine-quarter peg, a horizon the FOMC actually reached in early 2012; their published quantification puts the model’s overstatement around 28 times on output and 8 on inflation against the modest effects in the data (Del Negro et al., 2015). The forward guidance puzzle was uncontested as diagnosis. The cure was contested, and the field decided against the diagnosing authors’ own remedy: their perpetual-youth discounting lost to McKay, Nakamura, and Steinsson’s incomplete-markets mechanism, in which uninsurable risk and borrowing constraints make precautionary saving counteract intertemporal substitution, yielding a discounted Euler equation under which guidance five years out retains about 40 percent of its power and ten-year guidance essentially none (Del Negro et al., 2015; McKay et al., 2016). The incomplete-markets fix won because it was not a fix. It was a diagnosis that generalized: if precautionary saving damps the response to future rates, then the direct intertemporal-substitution channel was never the main transmission mechanism, and the puzzle became the leading motivation for rebuilding monetary economics on heterogeneous-agent foundations (McKay et al., 2016).

Kaplan, Moll, and Violante completed the relocation. In a model disciplined by the microdata, the joint distribution of earnings, liquid wealth, and illiquid wealth, and therefore by realistic hand-to-mouth shares and MPCs, roughly 90 percent of the first-quarter consumption response to a rate cut is the indirect general-equilibrium income channel and only about 10 percent is direct intertemporal substitution, the mirror image of the representative-agent decomposition, where the direct share is about 99 percent (Kaplan et al., 2018). The fiscal-transfer term inside the indirect channel is quantitatively significant, so the fiscal reaction function co-determines monetary potency and Ricardian equivalence breaks exactly where Barro’s own hedges said it would (Kaplan et al., 2018; Barro, 1974). Score the RANK-HANK dispute the way the record does, without triumphalism in either direction: RANK still owns price-level determinacy, the nominal anchor, and the pedagogical core (Woodford, 2003; Galí, 2008); HANK won the positive question of how rate changes reach consumption, on microdata the aggregate models had never confronted, and became the organizing framework of a decade (Kaplan et al., 2018). The neutrality question’s channel component, “through what channel?”, was thus re-answered wholesale: not the Euler equation but redistributed income; not the balance sheet but the expected path.

Two more currents ran through the decade and primed the next episode. Sims used his 2013 presidential address to bring the fiscal theory down to earth: once the Fed pays interest on reserves and holds long-maturity assets, a rate hike without fiscal backing passes straight through the consolidated budget as more interest-bearing debt and can raise inflation rather than damp it, and a central bank with negative net worth may be forced into inflation to rebuild its balance sheet; with debt near 100 percent of GDP, rates rising from 2 to 6 percent would push interest expense from under a tenth to roughly 30 percent of federal spending (Sims, 2013). Cochrane’s treatise, finished as the 2021 surge was breaking, pressed the institutional point to its limit: money-demand theories and Taylor-rule theories no longer describe cashless, interest-on-reserves, zero-bound economies, leaving the fiscal theory, on his reading, the only coherent account compatible with current institutions (Cochrane, 2023). The honest caveat cuts both ways: the fiscal theory had developed for thirty years during quiet inflation, a theory in search of an episode (Cochrane, 2023), and it was about to get one.

And Wicksell returned, as promised in the prologue, in the form of a number. Laubach and Williams operationalized the natural rate as a Kalman-filtered state variable tied to trend growth, and their own results carried a loud warning: standard errors of one to two and a half percentage points and real-time measurement error that is essentially permanent (Laubach and Williams, 2003). Holston, Laubach, and Williams internationalized the estimate, added their own caution, an IS slope near zero that leaves r-star barely identified, and found the decade’s load-bearing fact, a large common decline in r-star across four advanced economies, the U.S. rate falling toward zero after 2007 with no sign of reverting (Holston et al., 2017). The estimate won operational adoption, published by the New York Fed and quoted as “the” r-star; the imprecision warning was ignored in downstream use until the 2021 surge and the 2022-23 rate rise forced the reckoning and the authors’ own post-COVID revision (Laubach and Williams, 2003; Holston et al., 2017). Wicksell’s mechanism came back as the operational core of policy, and came back exactly as he left it: indispensable and unmeasured (Wicksell, 1936).

The decade closed with three charges primed. A fiscal theory that had waited thirty years for a fiscal inflation (Sims, 2013; Cochrane, 2023). An r-star treated as precise and permanent against its estimators’ own warnings (Holston et al., 2017). And a HANK apparatus, built for a demand-deficient economy, that made transfers to high-MPC households first-order for aggregate demand (Kaplan et al., 2018), completed just in time for the largest transfer program in American history.

The 2021 inflation

The second humiliation was the first consensus-breaking inflation since the 1970s, and it struck the synthesis on its home ground. The flat Phillips curve was the central equation of the Era IV consensus, the thing thirty years of data had made the profession most confident about, and the workhorse model predicted almost no inflation from observed conditions; core inflation then surged to multi-decade highs while the Fed, the forecasters, and the models were caught flat-footed together (Benigno and Eggertsson, 2023). Unusually in this history, the scholarship ran concurrently with the episode judging it, so the papers read as live testimony rather than settled doctrine, and this section preserves that status: no verdict below is historiographically final.

The era’s answer to the neutrality question, stated as the contestable claim it is: the price level is set neither by the money stock nor by intertemporal substitution but by the collision of aggregate demand with a supply constraint that is regime-dependent, flat and forgiving when the economy is slack, near-vertical and unforgiving once labor is fully employed. On this reading both prior answers, the 1970s’ “expectations do everything” and the Great Moderation’s “a flat anchored curve absorbs demand,” were special cases observed on the flat arm of a curve whose steep arm nobody had seen since the 1940s (Benigno and Eggertsson, 2023). The claim rests on one episode, and a rival account fits the same data, which is exactly why it is stated as contestable rather than settled (Bernanke and Blanchard, 2023).

The defining dispute was a forecast placed in public before the resolving data existed. Summers and Blanchard warned in early 2021 that the American Rescue Plan would overheat the labor market and ignite inflation; Team Transitory, including much of the Fed, held that the surge was supply-chain disruption that would self-correct (Bernanke and Blanchard, 2023). The forensic decomposition arrived in 2023 from a source with standing on both sides: Bernanke and Blanchard, the latter auditing his own forecast. Their deliberately bare-bones four-equation wage-price model, largely estimated on pre-COVID data, delivers a verdict with two edges. Most of the 2021-22 surge came from shocks to prices given wages, energy, food, and supply-chain shortages, with the durables-demand-and-shortage combination dominant in mid-2021; tight labor markets contributed little early and were actually negative in 2020 and early 2021; and the feared wage-price spiral is not in the data, the estimated catch-up effect weak and insignificant, expectations rising modestly without de-anchoring (Bernanke and Blanchard, 2023). That is Team Transitory’s mechanism. The other edge is Team Permanent’s: the labor-market component, small at first, is the persistent one, and by early 2023 tightness still explained about a third of excess wage growth, with the model projecting that returning inflation to target required grinding the vacancy-to-unemployment ratio back to its pre-pandemic value near 1.2, mapping to roughly 2.7 percent CPI inflation (Bernanke and Blanchard, 2023). The authors’ own referee call: the optimists were initially more right than wrong, the overheating critics may claim some vindication over the longer run (Bernanke and Blanchard, 2023). Blanchard’s audit found his warning right in direction and wrong in mechanism, which is a harder thing to publish than being simply right, and the paper deserves its centrality partly for that.

Then the episode ran its own out-of-sample test, and the test favored the optimists. Inflation fell substantially through 2023-24 without the large unemployment rise the persistent-labor-market channel implied, the “immaculate disinflation,” closer to the optimistic Beveridge-curve scenario the decomposition had flagged without endorsing (Bernanke and Blanchard, 2023). Supply-driven-with-anchored-expectations beat entrenched-overheating on the one outcome the models had to predict.

The theoretical repair reached back past the entire Era IV synthesis to the original curve. Benigno and Eggertsson embedded Phillips’s 1958 mechanism, wages bid up fast when labor is scarce, falling slowly when it is slack, in a New Keynesian model with search frictions and downward-rigid wages, and derived an Inverted-L Phillips curve, flat while vacancies fall short of unemployment, steep once they exceed it (Benigno and Eggertsson, 2023; Phillips, 1958). The kink is not a nuance; it is the result. In their regressions the slope on log tightness jumps from an insignificant 0.24 below the threshold to about 3.97 above it in the full sample, and from 0.52 to 5.72 in the post-2008 subsample, significant at the 1 percent level, while below the threshold one cannot reject a completely flat curve (Benigno and Eggertsson, 2023). The economy crossed the threshold in May 2021 and pushed the ratio past 2 by March 2022, the tightest labor market since the Second World War (Benigno and Eggertsson, 2023). The flat curve of the Great Moderation, on this reading, was an artifact of a sample that never contained a fully employed labor market: the profession spent sixty years smoothing away a nonlinearity that Phillips had drawn into his original figure (Phillips, 1958; Benigno and Eggertsson, 2023). Their contrarian corollary, that disinflation from the steep arm would be cheap because the surge was an unwinding labor shortage under anchored expectations rather than the entrenched expectations of the 1970s, was substantially vindicated by the same disinflation that disciplined Bernanke-Blanchard’s central projection (Benigno and Eggertsson, 2023).

What 2021 did to the oldest dispute in this history must be stated precisely, because the temptation to over-claim runs in both directions. The Phillips-curve stability fight reopened, and the menu side won a partial resurrection, in nonlinear form only: a stable-if-kinked structural relation is back on the table (Benigno and Eggertsson, 2023). What did not come back is the exploitable menu, and the reason is Friedman’s: the entire soft-landing argument turns on expectations being anchored, which is precisely the condition Friedman said any short-run trade-off is hostage to, and on the steep arm holding unemployment below its natural level would still require accelerating inflation (Friedman, 1968; Benigno and Eggertsson, 2023). So 2021 revives the shape of the pre-Friedman curve while retaining the natural-rate anchoring logic, a synthesis rather than a repudiation, and it corrects the sixty-year-old misreading in both directions at once: Phillips’s honest nonlinearity is recovered, Samuelson-Solow’s imputed menu stays refused (Phillips, 1958; Samuelson and Solow, 1960).

The era’s open question, stated here and deliberately not resolved: was the Era IV synthesis wrong, or merely incomplete? The incomplete reading has real evidence: Bernanke and Blanchard show a rather traditional wage-price model, estimated mostly on pre-pandemic data, tracks the surge once commodity and shortage terms are included, so nothing needed to be thrown out (Bernanke and Blanchard, 2023). The wrong reading has real evidence too: the profession missed the surge because its central equation assumed a flatness that was a mis-specification, an artifact of the sample, and a model that fails exactly when the constraint binds is not incomplete, it is wrong about the constraint (Benigno and Eggertsson, 2023). The disinflation test cuts against the strongest version of each: the persistent-labor channel was weaker than the traditional model projected, and the surge was sharper than the smooth curve allowed (Bernanke and Blanchard, 2023; Benigno and Eggertsson, 2023). I decline to adjudicate, and not out of politeness; the kink is estimated on a single crossing of the threshold, and the immaculate disinflation is a single realization. One episode cannot distinguish a durable nonlinearity from a lucky draw.

The loose thread the next era inherits is therefore the shape and stability of the supply constraint itself: whether the Inverted-L is structure or artifact, whether the vacancy-unemployment ratio is the right tightness measure or a fragile proxy, and whether anchored expectations were a fact about the 2020s or a fact about institutions that happened, this time, to hold (Benigno and Eggertsson, 2023; Bernanke and Blanchard, 2023). If this history’s rhythm holds, the question will be answered the way every era’s question has been answered: by the next inflation the consensus fails to predict.

Epilogue: The Pendulum and the Balance Sheet

The pendulum

Count the flips. The prologue answered that money is neutral in the long run, with the interval a nuisance (Hume, 1752; Fisher, 1911). Era I flipped to impotence: money barely matters, because at the floor extra money buys idle balances and animal spirits ignore the rate, so fiscal policy owns the level of output (Keynes, 1936; Hicks, 1937; Modigliani, 1944). Era II flipped again: money matters enormously in the short run and not at all in the long, the hinge being whether the inflation was expected (Friedman and Schwartz, 1963; Friedman, 1968; Phelps, 1967). Era III flipped harder: anticipated money matters not at all, even in the short run, with the limiting case a business cycle containing no money whatsoever (Lucas, 1972; Sargent and Wallace, 1975; Kydland and Prescott, 1982). Era IV flipped back: money matters because prices are sticky, and a Taylor rule plus an independent central bank makes the problem an engineering matter (Calvo, 1983; Taylor, 1993; Woodford, 2003). Era V flipped once more, in two movements: at the zero bound money matters through the expected path and the income of heterogeneous households rather than the short rate and the Euler equation (Eggertsson and Woodford, 2003; Kaplan et al., 2018), and in 2021 demand matters conditionally, through a supply constraint that is forgiving until it is not (Benigno and Eggertsson, 2023). Five flips in under a century, roughly one per generation.

Now count what purchased them. The Depression bought Era I; the field did not reason its way out of the quantity theory, it was evicted. The Great Inflation bought Eras II and III twice over, vindicating the counterrevolution it had not yet finished reading and demolishing the econometric practice that had promised to manage it (Friedman, 1968; Lucas, 1976). The conquest of that inflation, at a measured cost of roughly two points of unemployment and 12 percent of industrial production per deliberate disinflationary decision (Romer and Romer, 1989), bought the Great Moderation and its triumphalism. Then 2008 and 2021 bought Era V, each exposing a different abstraction the synthesis had filed as harmless (Woodford, 2003; Benigno and Eggertsson, 2023). Not one of these flips originated inside the seminar room. Every reigning consensus in this history was ended by an inflation or a deflation it failed to predict, which makes monetary economics the field of economics most disciplined by events, and the least able to pretend otherwise. Most fields can carry an anomaly for a generation while the paradigm waits for better data. A monetary consensus that misses an inflation is dismissed within the decade, because its predictions are printed monthly, in the price data, next to its name.

The discipline runs in both directions, and that is the field’s other distinction. Central banks acted on these papers. The natural-rate hypothesis was two years old as doctrine when Volcker bet the American economy on it; Rogoff’s conservative banker staffed a generation of central banks; New Zealand wrote the inflation target into legislation; Eggertsson and Woodford’s commitment logic became forward guidance, and their ratcheting target became the Fed’s 2020 framework (Friedman, 1968; Rogoff, 1985; Eggertsson and Woodford, 2003). The debates in this history were settled partly by argument and partly by institutions staking unemployment on one side of them. It is the rare field whose referee reports arrive as recessions.

What survived every flip is worth naming, because the pendulum is not a counsel of despair. The vertical long-run curve survived; every era since 1968 has retained it, including the one that revived Phillips’s shape (Friedman, 1968; Benigno and Eggertsson, 2023). Rational expectations survived as method even after its founding mechanism was abandoned (Lucas, 1972). The natural rate survived from 1898 to the New York Fed’s website, never once measured to anyone’s satisfaction (Wicksell, 1936; Laubach and Williams, 2003). And the identification problem survived everything, reappearing at each level of technical sophistication, from the horse races of the 1960s to the thirty-minute windows that discovered the Fed information effect (Friedman and Meiselman, 1963; Nakamura and Steinsson, 2018). The field’s constants are a limit theorem, a method, an unmeasurable number, and a difficulty.

The balance sheet

By Era IV, monetary economics had built itself a frictionless benchmark: the cashless representative-agent economy, in which the price level is anchored by an interest-rate rule, transmission is one household’s Euler equation, and the financial system is a veil (Woodford, 2003; Galí, 2008). It was the purest irrelevance proposition in the field’s history, and Era V named what it left out: the balance sheet in the middle.

The credit channel put the borrower’s balance sheet into the transmission mechanism, and then the intermediary’s: net worth as a state variable, the external finance premium moving inversely with it, a spiral in which falling asset values tighten the very constraint that forces the assets to be sold (Bernanke and Gertler, 1989; Bernanke et al., 1999; Gertler and Kiyotaki, 2010). HANK put the household’s balance sheet in: the joint distribution of liquid and illiquid wealth determines who spends a rate cut or a transfer, and with it, whether the cut or the transfer does anything at all (Kaplan et al., 2018). Even the sovereign was re-read as a balance sheet: the fiscal theory prices money off the government’s, and Sims’s warning about a central bank with negative net worth is a solvency argument applied to the institution that was supposed to stand outside solvency (Sims, 2013; Cochrane, 2023). In each case the frictionless benchmark had made the same move, aggregating away the middle of the economy, and in each case the repair consisted of putting the middle back: whose balance sheet absorbs the shock determines what the shock does.

The omission was not a modeling convenience. The neutrality results failed exactly where balance sheets turn relative prices into spending: the household that cannot borrow, the bank that cannot lever, the government whose debt is the economy’s residual claim. The frictionless benchmark was never wrong as a benchmark. It was a map of what must be true for the middle of the economy not to matter, and its value was to make the middle’s mattering measurable.

So this history ends where the field now stands, between humiliations. The pendulum’s position: money matters, conditionally, through balance sheets and expectations, against a supply constraint of uncertain shape. The pendulum’s history says this answer, too, is provisional, held in place by institutions that have anchored expectations for one episode and by a nonlinearity estimated on one crossing (Benigno and Eggertsson, 2023). Wicksell’s rate remains unmeasured; the budget constraint remains an identity or an equation, depending on whom one asks (Laubach and Williams, 2003; Buiter, 2002). The next consensus will be written, like every one before it, by the authors of the next mistake. The reader is entitled to one prediction, and history licenses only this one: when the flip comes, the balance sheet will be somewhere in the middle of it.