Employment: Keynes through New Keynesian
The popular story says Keynes was overthrown, then came back. The real story is stranger: the theory of involuntary unemployment survived the rational-expectations revolution by absorbing it — and the model every central bank now runs on has Keynes’s claim at its heart.
以辩论图谱查看Here is the whole thread the way the intellectual-lineage record holds it — nine thinkers across four eras, one continuous argument about what determines employment and what stabilization can do about it. The era-organized textbooks cover these figures in different chapters; the thread is what runs through them. Click any node to peek at where it sits.
Open the full network in the intellectual-lineage graph — the closest named lineage is expectations, which shares this thread’s natural-rate and rational-expectations rungs before diverging toward belief-formation.
The Keynesian revolution
“The difficulty lies, not in the new ideas, but in escaping from the old ones, which ramify, for those brought up as most of us have been, into every corner of our minds.”
— John Maynard Keynes, The General Theory of Employment, Interest and Money, 1936
Keynes wrote that in 1936, with the Depression still raging and roughly a quarter of American workers out of a job. The old idea he meant to escape was not a fringe position. It was the entire respectable apparatus of economics, and it said something startling: these people could not exist. A labor market clears. Persistent mass unemployment is, as a matter of theory, impossible. To make involuntary unemployment a real category, Keynes had to break a theory that was internally coherent and a century deep.
What could the classical apparatus not explain, and what did Keynes add? The classical labor market clears through the wage: if workers are unemployed, wages fall until employers want to hire them all again. Behind it stood Say’s Law — supply creates its own demand, so a general glut, a shortfall of demand across the whole economy, cannot happen. Keynes’s move was to reverse the direction of causation. Output is set by effective demand: firms produce what they expect to sell. If demand is deficient, they produce less, hire fewer workers, and the labor market settles at an equilibrium with mass unemployment that no wage cut reliably fixes — because cutting wages also cuts the incomes that demand the output.
That insight needed a tractable form, and the synthesis supplied it. In 1937 John Hicks compressed the General Theory into the IS-LM diagram — goods-market and money-market equilibrium on one pair of axes — and Paul Samuelson built the textbook around it. The payoff was the Samuelson-Solow Phillips curve: a stable, exploitable menu trading a little more inflation for a little less unemployment. Demand management became something a finance minister could actually do.
The engine is the spending multiplier. An autonomous rise in demand $\Delta A$ raises output by a multiple of itself, because each round of spending becomes someone’s income and is partly re-spent:
$$\Delta Y = \frac{1}{1 - c}\,\Delta A$$where $c$ is the marginal propensity to consume. With $c = 0.8$, a dollar of demand becomes five dollars of output — and idle workers get hired to produce it.
Picture a factory town where the plant cuts orders. Laid-off workers stop spending at the diner; the diner lays off a cook; the cook stops buying from the hardware store. The town can settle into a low-employment rut where everyone wants to work but nobody is hiring, because nobody is spending, because nobody has income. Each business is waiting for the others. An outside push — a government order, a tax cut — breaks the standoff: the first round of spending becomes someone’s wage, which becomes the next round.
That is the apparatus, compressed. The formal home of the Keynesian cross and IS-LM is Economics Ch.8 (Intro Macro Models); the AD-AS and Phillips machinery the synthesis bolted on is Ch.9 (Intermediate Macro). For the lineage — what Keynes inherited, what the General Theory actually argued, and what Hicks’s diagram quietly dropped — see History of Economic Thought Ch.8 (The Keynesian revolution) and the synthesis chapter Ch.9 (The postwar synthesis).
“Men are involuntarily unemployed if, in the event of a small rise in the price of wage-goods relatively to the money-wage, both the aggregate supply of and the aggregate demand for labour at the current money-wage would be greater than the existing volume of employment.”
— John Maynard Keynes, The General Theory, 1936, ch. 2
Can the labor market fail to clear?
Classical theory said no: wages fall, hiring resumes, mass unemployment cannot persist. Keynes said the Depression was proof it could. The whole thread starts with this one claim — that deficient demand can leave willing workers idle with no wage cut to save them.
The orthodoxy and the man who broke it
“The State should not encourage works of doubtful utility to provide employment… Whatever might be the political and social advantages, very little additional employment and no permanent additional employment can in fact, and as a general rule, be created by State borrowing and State expenditure.”
— The “Treasury View,” HM Treasury memorandum, 1929
Read at full strength, the classical position is not a strawman. Its logic was airtight given its premises: with full employment of resources and a fixed pool of saving, every pound the government borrows to spend is a pound that would otherwise have funded private investment — so public works merely crowd out private ones and shuffle employment around. Markets clear; saving equals investment at the market interest rate; persistent demand shortfalls are theoretically impossible. This is why it took a catastrophe to dislodge it. The theory was coherent. It was the world that refused to cooperate.
“The long run is a misleading guide to current affairs. In the long run we are all dead. Economists set themselves too easy, too useless a task if in tempestuous seasons they can only tell us that when the storm is long past the ocean is flat again.”
— John Maynard Keynes, A Tract on Monetary Reform, 1923
Keynes’s counter does not deny that markets clear eventually. It denies that “eventually” is the relevant time frame when a quarter of the workforce is idle now. His premise — that the pool of saving is not fixed but expands and contracts with income — dissolves the crowding-out argument: in a slump, government borrowing draws on saving that would otherwise sit idle, not on saving that would have been invested. The duel between these two readings of the 1930s is held side by side, with the Austrian counter-case argued at its strongest, in the framework-pair walkthrough Hayek vs. Keynes: whose framework did the 1930s vindicate? This thread does not re-stage that fight; it traces forward from Keynes’s victory.
Where this leaves us
Keynes won the founding argument: involuntary unemployment is real and demand management is a genuine tool. But notice what the synthesis did to him on the way to the textbook. The IS-LM diagram and the stable Phillips menu turned the General Theory’s unsettling claims into a tidy fine-tuning exercise — pick a point on the menu, trade a little inflation for a little less unemployment, adjust as needed. The radical edges were sanded off. That domesticated version, with its promise of a stable trade-off you could exploit indefinitely, is exactly the thing the next rung will break. The synthesis built its house on the assumption that the menu would hold still.
For a generation the synthesis ran the show, and the menu seemed to hold. Then in 1968 one man stood up at the American Economic Association meeting and said the menu was an illusion — that the moment policymakers tried to exploit it, it would self-destruct. He said this before the 1970s. The decade that followed would prove him right.
The monetarist and New Classical challenge
“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 per se, but from unanticipated inflation, which generally means, from a rising rate of inflation.”
— Milton Friedman, “The Role of Monetary Policy,” AEA Presidential Address, American Economic Review, 1968
Friedman said this in 1968 — before the stagflation of the 1970s, not after. It is a prediction, not a post-mortem. The synthesis had promised a stable menu; Friedman told the profession that the menu would self-destruct the moment anyone tried to live on it, because workers would learn. Within five years inflation and unemployment were climbing together, the one combination the stable Phillips curve said could not happen.
What could the synthesis Phillips curve not explain, and what did this rung add? The synthesis read the inflation-unemployment relation as a structural menu: accept 4% inflation, get 4% unemployment, and you can stay there. Friedman and Edmund Phelps argued there was no such permanent menu. Workers care about real wages; once they come to expect 4% inflation, they demand it in their wage bargains, and the trade-off vanishes. To keep unemployment below its natural rate, policymakers must deliver inflation that is forever accelerating — always a step ahead of expectations. The long-run Phillips curve is vertical.
The expectations-augmented Phillips curve makes the point precise. Unemployment falls below its natural rate $u^{*}$ only to the extent that actual inflation $\pi$ exceeds expected inflation $\pi^{e}$:
$$\pi = \pi^{e} - \beta\,(u - u^{*})$$In the long run expectations catch up, $\pi^{e} = \pi$, and the equation forces $u = u^{*}$ at any inflation rate. The trade-off is real only while expectations lag; it cannot be exploited once they adjust.
You can fool workers into supplying more labor by letting inflation erode their real wage before they notice — once. The next time, they build the expected inflation into their wage demands, and you get the inflation without the extra employment. To buy the same boost again you need a bigger surprise. Try to keep unemployment permanently low this way and you are committed to inflation that climbs without limit.
Robert Lucas then deepened the assault from ad hoc to airtight. Friedman’s workers learned slowly, with backward-looking “adaptive” expectations — a convenient assumption Lucas found indefensible. Make expectations rational instead: agents use all available information, including their knowledge of how policy works. Two devastating results follow. First, the policy-ineffectiveness proposition (Sargent and Wallace, 1975): if agents see a systematic monetary rule coming, they neutralize it, and predictable policy cannot move output at all. Second, and more corrosive, the Lucas critique: the parameters of any Keynesian econometric model are not stable across policy regimes, because they bundle in people’s expectations of the old policy — so you cannot use the model to evaluate a change in policy. The entire demand-management apparatus, evaluated on its own econometric terms, became untrustworthy.
Push rational expectations together with continuously clearing markets and you arrive at real business cycle theory (Kydland and Prescott): if money is neutral and markets always clear, then observed booms and slumps are not failures at all but the economy’s efficient response to real shocks — productivity, technology, taste. There is nothing for stabilization policy to stabilize. The natural-rate and Lucas-critique apparatus lives in Economics Ch.9 (Intermediate Macro) and Ch.14 (Real Business Cycles); the counter-revolution’s lineage — Friedman, Phelps, Lucas, Sargent — runs through History of Economic Thought Ch.10 (The counter-revolution).
The falsifying fact was not theoretical. Through the 1970s the G7 economies posted rising inflation and rising unemployment together — the misery index doubling — the exact combination the stable Phillips curve ruled out. That break is documented in Economic History Ch.16 (Stagflation and the neoliberal turn). The decade as a lived case — Nixon to OPEC to Volcker, and the apparatus it forced into being — is the load of the case-up walkthrough The 1970s: how stagflation built a new macroeconomics; here it is one fact, the structural break that ended the synthesis menu.
“There is no permanent trade-off. The temporary trade-off comes not from inflation per se, but from unanticipated inflation.”
— Milton Friedman, AEA Presidential Address, 1968
Does the Phillips menu self-destruct when you use it?
The synthesis sold a stable inflation-unemployment trade-off you could pick a point on. Friedman and Phelps said the act of exploiting it would destroy it — and they said so before the 1970s confirmed it. This is the thread’s real defeat, and conceding it is what makes the survival story honest.
The critique at full strength, and the Keynesian who conceded
“Given that the structure of an econometric model consists of optimal decision rules of economic agents, and that optimal decision rules vary systematically with changes in the structure of series relevant to the decision maker, it follows that any change in policy will systematically alter the structure of econometric models.”
— Robert Lucas, “Econometric Policy Evaluation: A Critique,” 1976
This is the New Classical critique at its most lethal, and it must be granted its full force. Lucas is not quibbling. He is showing that the large Keynesian models used to forecast the effect of policy were built on relationships that would themselves change the moment policy changed — because those relationships encoded people’s expectations of the old regime. The critique is permanently correct. It did not get refuted; it got absorbed into how everyone, Keynesian or not, now builds macroeconomic models. To answer it you could not patch the old apparatus. You had to rebuild on microfoundations and rational expectations — on the New Classicals’ own terms.
“The trade-off has not so much shifted as it has become much steeper, or perhaps even vertical… We are all monetarists now, in the sense that we accept the natural-rate hypothesis as a long-run proposition.”
— the broad Keynesian concession of the late 1970s, reflecting James Tobin and Franco Modigliani’s acceptance of the long-run natural rate
Crucially, the leading Keynesians did not dig in. Tobin and Modigliani — no monetarists — conceded the long-run natural rate and the absence of a permanent trade-off. So this rung is not a one-sided counter-revolutionary victory lap; it is a genuine intellectual defeat that honest Keynesians acknowledged. The engine driving the whole transition is the move from adaptive to rational expectations — the same machinery whose own lineage, from animal spirits to prospect theory, is traced in the adjacent expectations thread. What the Keynesians conceded at the frame level, they would have to win back at the level of one assumption the New Classicals had waved away.
Where this leaves us
Concede it fully, because the thread depends on the honesty: the naive Phillips curve genuinely failed, Friedman and Phelps genuinely called it in advance, and the rational-expectations critique was genuinely devastating. If money is neutral whenever policy is predictable, and the old econometrics cannot be trusted to evaluate any policy change, then what, exactly, is left for demand management to do? At this rung the answer looked like “nothing,” and serious people wrote the obituary for Keynesian stabilization. The thread does not minimize that defeat to protect the story. The survival in the next stage is worth more because the defeat was real.
Rational expectations looked like the end of Keynesian stabilization. If markets clear and people see policy coming, government cannot move output. The obituaries were written. And then a handful of economists noticed one thing the New Classicals had simply assumed away — and rebuilt the entire Keynesian case on top of the enemy’s own foundations.
The New Keynesian synthesis
“Small menu costs can be the source of large fluctuations in output and welfare… a privately small cost of price adjustment can have a socially large consequence.”
— N. Gregory Mankiw, “Small Menu Costs and Large Business Cycles,” Quarterly Journal of Economics, 1985
A trivially small cost — the bother of reprinting a menu, relabeling a shelf, renegotiating a contract — sounds like the kind of thing a serious theory rounds to zero. Mankiw’s 1985 paper showed it was the loose thread that could unravel the whole New Classical conclusion. The New Keynesians did not reject rational expectations. They accepted it, accepted microfoundations, and added one realistic friction the New Classicals had assumed away — and on that single concession, Keynes walked back out of the grave.
What did the New Classical rung leave unexplained that one friction fixes? Its whole anti-Keynesian punch came from continuous market-clearing — prices adjust instantly, so money is neutral and demand cannot move real output. But continuous price adjustment is an assumption, and it is the one the data most plainly reject: prices are demonstrably sticky, changed every few quarters, not every instant. The New Keynesians took the New Classicals’ own machinery — optimizing agents, rational expectations, general equilibrium — and changed exactly one thing. Firms face a small cost to changing prices (Mankiw’s menu cost), or they can only re-set prices at random intervals (Calvo’s staggered pricing). That is enough.
With sticky prices, money is non-neutral even under rational expectations: a demand shock that would be absorbed by an instant price change instead moves real output, because prices cannot all jump at once. Output gaps reopen — and because Michael Woodford grounded the framework in a welfare-theoretic loss function, those gaps are genuine welfare losses, not accounting artifacts. So stabilization policy is warranted again, on impeccably New Classical terms. The operational form is the three-equation model: a dynamic IS curve, the New Keynesian Phillips curve, and a Taylor rule for the central bank.
The New Keynesian Phillips curve is the synthesis in one line. Current inflation depends on expected future inflation and the current output gap $x_{t}$:
$$\pi_{t} = \beta\,\mathbb{E}_{t}\,\pi_{t+1} + \kappa\,x_{t}$$It is forward-looking (rational expectations, $\mathbb{E}_{t}\pi_{t+1}$, conceding the New Classical point) yet the slope $\kappa$ — which comes from the price stickiness — is what makes demand matter (the Keynesian point). Both inheritances sit in a single equation.
Imagine every shop could only change its price tags on a random day each year. When demand suddenly drops, most shops are stuck with last year’s prices — too high — so they sell less and cut staff instead of cutting prices. The economy feels the demand shock as lost jobs, exactly as Keynes said, even though everyone is perfectly rational and forward-looking. The single assumption that prices cannot all move at once is enough to bring back the entire Keynesian world the New Classicals had argued away.
The Calvo-pricing and three-equation apparatus, and the Taylor rule that operationalizes it, live in Economics Ch.15 (New Keynesian Economics) — the New Keynesian Phillips curve at §15.3, the Taylor rule at §15.5, the full three-equation model at §15.6. The lineage — Mankiw, David Romer, Woodford, Galí, and the synthesis built on the victor’s foundations — runs through History of Economic Thought Ch.12 (New Keynesian and the modern monetary-policy consensus).
“A privately small cost of price adjustment can have a socially large consequence.”
— N. Gregory Mankiw, Quarterly Journal of Economics, 1985
Did one friction bring Keynes back from the dead?
New Classical economics had seemingly retired demand management. The New Keynesian reply was not to fight rational expectations but to accept it — and add a single realistic assumption, sticky prices, that the other side had waved away. On that one concession, money becomes non-neutral and stabilization comes back.
Meeting the bar, not dodging it
“The Marshallian model of the cycle is dead… Economic fluctuations are the optimal response of the economy to real shocks. There is no need to invoke market failure or sticky prices to explain them.”
— the real business cycle position, after Edward Prescott’s “Theory Ahead of Business Cycle Measurement,” 1986
Held at full strength, the RBC challenge is formidable, and the New Keynesians had to answer it on its own ground. Prescott’s claim is that a model with no money, no frictions, and no role for policy can reproduce the major business-cycle statistics — so why add epicycles? The methodological demand behind it is entirely legitimate: any serious macro model must be built from optimizing agents with rational expectations, not from aggregate relationships pulled out of the data. That bar is not pedantry. It is the standard the whole profession adopted, and the New Keynesians accepted it without complaint.
“New Keynesian economics… takes as its starting point the rational-expectations revolution… The Keynesian conclusion — that the economy can suffer from a deficiency of aggregate demand — can be derived from, rather than assumed in contradiction to, the assumption of rational, maximizing agents.”
— N. Gregory Mankiw & David Romer, New Keynesian Economics, 1991
The New Keynesian reply does not lower the bar — it clears it. Mankiw and David Romer (the New Keynesian Romer, not Paul Romer of growth theory) accept rational maximizing agents and then derive the Keynesian conclusion from them, friction and all. They beat RBC on RBC’s own turf: same optimizing microfoundations, same rational expectations, one extra assumption that happens to be the empirically correct one. The labor-market half of this question — whether today’s unemployment is better read through a labor lens or a monetary-demand lens — is a contemporary reframe in its own right; the thread records only that the natural rate, once a labor concept, came back recast as a monetary one.
Where this leaves us
Keynes survived rational expectations by absorbing it, not by resisting it. The New Keynesians conceded the methodological battle — rational expectations and microfoundations are now non-negotiable — and won the substantive one: with one friction the data demand, money is non-neutral, output gaps are real welfare losses, and stabilization policy has a warranted role, operationalized as the Taylor rule that every inflation-targeting central bank now follows. The involuntary-unemployment claim from Stage 1 is sitting inside this model, formalized and doing real work. This is the thread’s central event: not a comeback, but an absorption.
By 2003 the synthesis looked complete. A Nobel laureate told the profession that the central problem of depression-prevention had been solved. Central banks ran New Keynesian models; the cycle seemed tamed. Five years later, the model that was supposed to explain recessions could not explain the one that was actually happening.
The post-2008 augmentation
“Macroeconomics in this original sense has succeeded: its central problem of depression prevention has been solved, for all practical purposes, and has in fact been solved for many decades.”
— Robert Lucas, AEA Presidential Address, 2003
Lucas said this in 2003, and it was not hubris talking — it was a reasonable reading of the record. Two decades of mild recessions, anchored inflation, and a tamed cycle stood behind him. Five years later, a financial collapse the canonical New Keynesian model had no sector for would put output gaps in the double digits and pin interest rates at zero. The question this stage answers is not whether the confidence was foolish. It is whether 2008 broke the thread — or extended it.
What could the canonical New Keynesian model not explain, and what did the post-2008 rung add? The pre-crisis workhorse had two blind spots. It had no real financial sector — banks, leverage, and collateral were abstracted away, so a credit crunch had nowhere to enter the model. And it had only a weak answer for the zero lower bound: with the policy interest rate stuck at zero, the Taylor rule cannot deliver the negative rate the model wants, and the central bank’s main lever stops working. The crisis hit both gaps at once. The synthesis was not abandoned in response. It was extended — three additions bolted onto the same DSGE chassis.
First, financial frictions. Bernanke, Gertler, and Gilchrist’s financial accelerator (1999, finally central after 2008) lets small shocks amplify through collateral, credit, and debt-deflation channels — a fall in asset prices tightens borrowing constraints, which deepens the fall. Second, heterogeneous agents. HANK models (Kaplan, Moll, and Violante, 2018) replace the single representative consumer with a distribution of households who differ in wealth and liquidity, which changes how the economy responds to shocks and to policy — hand-to-mouth households make fiscal transfers far more potent than the representative-agent model implied. Third, the zero lower bound became a first-class feature rather than an afterthought, reshaping what monetary and fiscal policy can do when rates are pinned.
Each of these is named here as a rung of the thread, not re-derived — the depth belongs elsewhere. The impulse-response comparison of New Keynesian against RBC dynamics, and the zero-lower-bound apparatus, sit in Economics Ch.15 §15.8 (the zero lower bound) and §15.9 (NK vs. RBC impulse responses compared). The window the synthesis seemed to vindicate — the 1984–2007 Great Moderation — is covered in Economic History Ch.18; the stress test that forced the extension is Ch.19 (The 2008 crisis and after). The post-2008 reckoning at the lineage level — what 2008 exposed and what survived — closes History of Economic Thought Ch.12.
The synthesis also has live heterodox edges on this modern rung — Post-Keynesians who say the friction-patching never addressed endogenous financial instability, Modern Monetary Theory on the fiscal side, market-monetarists who would target nominal GDP, an Austrian tradition that rejects the whole stabilization project. This thread locates them rather than engaging them at strength: each gets its proper hearing in its own walkthrough, and steelmanning them here would turn a lineage trace into a contested-question debate. They populate History of Economic Thought Ch.17 (Modern pluralism) — the modern-pluralism chapter, deliberately distinct from the New Keynesian Ch.12 where the thread’s own terminus lives.
“The state of macro is good.”
— Olivier Blanchard, “The State of Macro,” written in August 2008, weeks before the crash
Did 2008 break the thread or extend it?
On the eve of the crisis the profession declared depression-prevention essentially solved. Then the model with no financial sector met a financial crisis. The thread’s reading: the synthesis was extended, not replaced — and the endpoint is Keynes plus rational expectations plus microfoundations plus financial frictions.
Extended or broken?
“The crisis did not invalidate the New Keynesian framework so much as reveal what it had left out. The response — integrating financial intermediation, balance sheets, and the zero lower bound — has been an extension of the program, not a repudiation of it.”
— the mainstream post-crisis reading, reflected across the financial-frictions and HANK literatures
The pre-2008 consensus, argued at full strength, was a real achievement and not naivety. Every inflation-targeting central bank in the world ran on the New Keynesian framework; inflation was anchored, recessions were mild, and the apparatus delivered. When the crisis exposed the missing financial sector, the program did what a healthy program does: it extended. The financial accelerator and HANK are not foreign bodies grafted on — they are the same microfounded, rational-expectations machinery applied to the parts of the economy the canonical model had simplified away. The chassis absorbed the lesson.
“The economics profession went astray because economists, as a group, mistook beauty, clad in impressive-looking mathematics, for truth… they turned a blind eye to the limitations of human rationality… and to the dangers of an unregulated financial system.”
— Paul Krugman, “How Did Economists Get It So Wrong?”, 2009
The broken-macro critique, also at full strength, is that a model which only learns to represent a crisis after the crisis has happened has failed at the one task — prevention — that its confidence rested on. Adding financial frictions post hoc, on this reading, is an admission rather than a triumph, and the deeper Minsky charge stands: instability that the system breeds endogenously is still being smuggled in as an outside shock. Whether this amounts to a thread extended or a thread broken is the discipline-as-a-whole question that the synthesis-sibling walkthrough Did 2008 break macroeconomics? takes up directly; this thread renders only the lineage verdict and hands the reckoning across.
Where this leaves us
The thread is cumulative. Each rung grew out of the previous rung’s failure, and the founding Keynesian insight survived every challenge by absorbing it. The terminus is New Keynesian DSGE with financial frictions — Keynes’s demand-and-stabilization core, inside a rational-expectations and microfounded frame, augmented after 2008. The verdict has layers and they should not be collapsed. At the frame level there is consensus: demand matters, nominal rigidities make money non-neutral, stabilization has a role, and the founding insight is in the modern model — this is settled, not a live split. At the method level there is consensus too: rational expectations, the Lucas critique, and microfoundations are the universal default, the New Classical inheritance the thread names rather than hides. What is open is the numbers level: how large fiscal multipliers are (the load of Does government spending help the economy?), how much financial frictions add, whether HANK heterogeneity changes aggregate dynamics qualitatively, and whether the post-2008 extension is the synthesis working or breaking (the load of Did 2008 break macroeconomics?). That is the honest shape: an open numbers-frontier sitting on top of a settled frame and method — not “Keynes won” and not “Keynes was overthrown,” but the most cumulative success story in macroeconomics. The same apparatus, deployed as a per-episode toolkit for diagnosing recessions, is the load of What causes recessions?; deployed on the question of central-bank power, it is Can central banks control the economy?
One thread, four rungs
Trace the whole arc and the shape is unmistakable. Keynes made involuntary unemployment a real category against a classical orthodoxy that said it could not exist, and the synthesis operationalized him into demand management and a stable Phillips menu. Friedman, Phelps, and Lucas broke that menu — there is no long-run trade-off, expectations are rational, the old econometrics cannot evaluate policy — and for a moment the Keynesian role for stabilization looked finished. The New Keynesians answered not by resisting the rational-expectations revolution but by absorbing it: they conceded the methodological battle, added the one friction the data demand, and recovered money non-neutrality and a stabilization role on the victor’s own foundations. Then 2008 exposed the canonical model’s missing financial sector, and the synthesis was extended — financial frictions, heterogeneous agents, the zero lower bound — rather than replaced.
This is not a sequence of fashions. It is the most cumulative success story in macroeconomics: each step forced the next to be more rigorous, and the modern synthesis keeps Keynes’s demand-and-stabilization core inside a rational-expectations and microfounded frame, augmented after 2008 with financial frictions. The popular story — Keynes overthrown, then resurrected — gets the drama right and the substance wrong. The theory of involuntary unemployment did not come back from the dead. It survived by absorbing the revolution that was supposed to kill it, and the model running every inflation-targeting central bank today carries the founding insight at its heart. The frontier above it is genuinely open at the numbers; the frame and the method beneath it are settled. That is the honest verdict the next time someone tells you Keynes won, or that Keynes lost.