Three decades, three waves, one consensus broken. This chapter follows what happened to Keynesian macroeconomics between 1960 and 1990, when a coordinated assault dismantled the postwar synthesis from inside the discipline. Friedman supplied the empirical counter-narrative: the permanent income hypothesis attacked the consumption function, the natural rate hypothesis predicted that the Phillips curve tradeoff would break down, and the Friedman-Schwartz Monetary History rewrote the Depression as a monetary-policy failure rather than a market failure. Lucas supplied the methodological bomb: under rational expectations, the econometric apparatus the Keynesian profession had built for policy evaluation could not be used for policy evaluation, because the parameters were not invariant to the policy. Kydland and Prescott supplied the constructive program: a real-business-cycle model in which technology shocks and optimizing agents reproduced the statistical features of US business cycles without invoking demand shocks at all. By 1990 the counter-revolution had won the methodological war and was already losing the substance war. The chapter walks the three waves, then takes a position on what was accomplished and what was sacrificed.
In December 1967, Milton Friedman stood in front of the American Economic Association as its incoming president and told the discipline that the Phillips curve tradeoff it had relied on for a decade would break down. Within five years, stagflation proved him right. The address, delivered at the AEA’s Washington meetings and published as “The Role of Monetary Policy” in the March 1968 American Economic Review, is the cleanest opening salvo for what became the counter-revolution. The prediction was specific. The confirmation was rapid. The intellectual force of having said in advance what was about to happen, in a discipline that had organized its policy advice around the proposition that this could not happen, shifted the burden of proof in a way that the next two decades of macroeconomic argument would have to acknowledge.
What Friedman said was straightforward. The Phillips curve, the empirically observed inverse relationship between inflation and unemployment that A. W. Phillips had documented for British data in 1958 and that Samuelson and Solow had translated into a policy menu for American policymakers in 1960, was not a stable structural relationship. It was a short-run artifact that depended on inflation expectations being adaptive: workers and firms forming expectations of next year’s inflation by extrapolating from this year’s inflation. Adaptive expectations are backward-looking; agents update their forecasts gradually as new data arrives. While inflation surprises remained one-sided (actual inflation kept exceeding expected inflation), unemployment could be held below the rate the economy would otherwise gravitate toward. But surprises cannot be sustained indefinitely. Once inflation expectations adjusted to the realized inflation rate, the tradeoff would vanish, and any attempt to hold unemployment below its natural level would produce accelerating inflation rather than a stable lower unemployment rate.
The natural rate of unemployment is the rate that would obtain in the absence of monetary surprises, determined by structural features of the labor market: search frictions, demographic composition, the institutional set-up of unions and minimum wages, the matching of skills to jobs. Friedman in 1968 and Edmund Phelps in his 1967 paper “Phillips Curves, Expectations of Inflation, and Optimal Unemployment over Time” arrived at the same hypothesis from different directions. Phelps was building a microfounded labor-market model with informational frictions; Friedman was reasoning from the quantity theory of money and the proposition that real variables in the long run are determined by real, not monetary, factors. The convergence is the kind that gives a hypothesis force: two different analytical paths landing at the same conclusion in the same year, before the empirical event that would test it. The expectations-augmented Phillips curve replaces the simple inflation-unemployment menu with a relationship between unemployment and the gap between actual and expected inflation; only the gap moves unemployment, and the gap cannot stay one-signed forever.
What followed was the unusual experience of watching a prediction confirmed in real time. Through 1968 and 1969, the late-1960s expansion produced rising inflation alongside unemployment that had been pushed below the rate the structural features of the US labor market would have generated on their own. Through the early 1970s, inflation kept climbing while unemployment failed to stay where the standard Phillips curve diagram said it should. By 1974, after the first oil shock, the United States had simultaneously the highest inflation rate of the postwar period and unemployment well above its natural level. Stagflation, the simultaneous occurrence of rising prices and rising unemployment, was the empirical phenomenon the standard Keynesian model said could not occur, and it was occurring. The economic event itself, the wage-price spirals and the Volcker disinflation that ended them, lives in economic history ch. 16; this chapter takes the event as the empirical anchor for a prediction that had been on the record before any of it happened.
What gave the prediction its force was not Friedman’s rhetorical confidence but the structural argument behind it. The Phillips curve, on the Friedman-Phelps reading, had never been the policy menu the postwar synthesis had treated it as. It had been a temporary statistical regularity that emerged from a particular combination of conditions: roughly stable inflation expectations, an institutional environment in which monetary policy did not behave in obviously inflationary ways, and a sample period (the 1950s and early 1960s) short enough that the long-run relationship had not had time to assert itself. As soon as policymakers tried to exploit the regularity, the conditions that produced it would change. Workers would notice that inflation kept exceeding their forecasts, would adjust their forecasts upward, and would demand wage gains that protected against the higher expected inflation; firms would price accordingly; the curve the policymaker was trying to ride would shift up. The framework was self-defeating by construction. Policymakers could not buy a permanent reduction in unemployment by accepting a permanent rise in inflation, because the trade was structurally not on offer. (The textbook version of the expectations-augmented Phillips curve, with the formal expectations-formation rule and the algebraic statement of the natural-rate hypothesis, sits in economics ch. 14; the modern descendant inside the New Keynesian framework lives in economics ch. 15.)
The natural-rate argument was the second leg of the monetarist case. The first leg had been built a decade earlier, in Friedman’s A Theory of the Consumption Function (1957). The Keynesian consumption function (the proposition that current consumption is a stable function of current disposable income, with a marginal propensity to consume between zero and one) was the empirical foundation of the multiplier and therefore of the activist fiscal-policy program that the postwar synthesis carried as its operational core. Friedman’s permanent income hypothesis argued that consumers do not respond to current income as such; they respond to their estimate of long-run income, smoothing consumption across transitory income fluctuations by saving when income is unexpectedly high and dis-saving when income is unexpectedly low. A tax cut that consumers perceive as transitory will not raise consumption proportionately; a tax cut perceived as permanent will. Either way, the multiplier on a Keynesian fiscal stimulus is smaller, often much smaller, than the textbook framework suggested. The empirical evidence Friedman marshalled (consumption-income data across long time spans, cross-section data on consumption variation across income classes) was consistent with the permanent-income reading and inconsistent with the simple Keynesian one. (The formal model and the modern empirical literature on consumption smoothing live in economics ch. 14; the connection to the modern fiscal-policy debate runs through Walkthrough 01 on government spending.)
Take the 1957 attack on the consumption function and the 1968 attack on the Phillips curve together, and the monetarist program had landed two body blows on the postwar synthesis before the events of the 1970s arrived to confirm them. The third was the historical reframing. Friedman and Anna Schwartz published A Monetary History of the United States, 1867–1960 in 1963. The book is 860 pages and statistically dense; its central claim about the Great Depression is compact. The Federal Reserve allowed the US money supply to fall by roughly one-third between 1929 and 1933, through a sequence of policy choices that included passive responses to bank failures, refusal to provide liquidity to the system, and adherence to gold-standard discipline that the magnitude of the contraction did not require. The collapse in the money stock turned what would otherwise have been a recession into the Great Depression. The Fed’s contractionary stance was not the only cause of the depth and duration of the slump, but it was the cause that monetary policy, properly conducted, could have prevented.
The Keynesian reading of the Depression, dominant in the postwar synthesis, ran in the opposite direction. The Depression was a demand failure on a scale that monetary policy could not fix. Interest rates were already low; the liquidity trap meant additional money creation simply piled up as idle balances rather than translating into spending; the central bank was “pushing on a string.” Recovery required fiscal expansion on the scale that wartime mobilization eventually delivered. The framework explained, on its own terms, why the New Deal’s fiscal interventions had not been enough (they were too small) and why activist fiscal policy was the appropriate tool against demand-driven slumps. The empirical history of 1929–33 was, on the Keynesian reading, the canonical case for the framework.
Both readings were available at strongest form, and each had real explanatory work to do. The Keynesian reading captures something the monetarist reading does not: the demand-side collapse of investment and consumption was substantial; nominal rigidities meant that price adjustment could not clear markets at full employment; private agents’ willingness to spend was genuinely impaired. The monetarist reading captures something the Keynesian reading does not: the magnitude of the monetary contraction was large enough that even granting all of the demand-side problems, a Fed that had supplied liquidity instead of withdrawing it would have prevented the cascading bank failures and the deflationary spiral that turned 1930 into 1933. What Friedman-Schwartz shifted, in the profession’s working understanding, was the question of what central banks could destroy. The Keynesian framework had treated monetary policy as ineffective in a deep slump; Friedman-Schwartz treated monetary policy as the slump’s active cause. The two propositions are different. After 1963 it was no longer possible to discuss the Depression in serious macroeconomic terms without engaging the monetarist counter-narrative. (Where Friedman, the work, and the Depression sit relationally on the timeline: the friedman thinker node, the monetary_history work node, the monetarism school node, and the great_depression crisis node.)
What this counter-narrative bought was the legitimation of monetary policy as the primary stabilization instrument. If the worst macroeconomic event of the twentieth century had been caused by monetary failure, then the monetary authority’s competence and constraints were not a peripheral question. Friedman’s constructive program, the case for a steady money-growth rule rather than discretionary policy, follows directly from the diagnosis: a Fed that had been bound by a rule to keep the money supply growing at a constant rate would not have been free to make the contractionary mistakes Friedman-Schwartz documented. The rules-versus-discretion argument that Kydland and Prescott would later formalize had its empirical foundation in the Monetary History’s reading of 1929–33. (The modern central-banking debate over rules versus discretion, central-bank independence, and inflation targeting runs through Walkthrough 06 on central banks.)
Friedman’s case against fine-tuning rests on a counter-intuitive claim: a dumb fixed rule can beat a smart central banker. Watch why. Leave the policy on the fixed rule and raise the shock slider — the path tracks a steady band, because the rule does not chase the noise. Now switch to discretion: the policy reacts to each shock, but with a built-in lag (Friedman’s “long and variable” lag). The correction lands after the shock has already reversed, and the swing gets wider, not narrower.
Figure 11.3 (interactive). A fixed money-growth rule versus discretionary fine-tuning under common shocks. The rule path is steady; the discretion path reacts late and overshoots. Switch on the time-inconsistency framing to see the second argument for rules — the discretion path also sits at a higher average level, because agents anticipate the renege. Toggle the regime; drag the shock slider; flip the framing. Stylized illustration of the qualitative result, not a calibrated money-supply series.
A fixed rule can’t chase a shock late and make it worse, because it doesn’t chase it at all. The discretionary banker, reacting on a lag, keeps arriving with the wrong correction at the wrong time. That is the first argument for rules — reaction lags make active stabilization destabilizing. The time-inconsistency framing adds the second: a banker who can renege will be expected to renege, so the discretion path floats up to a higher average inflation with nothing to show for it.
Two acknowledgments need to be on the record before the chapter moves on. The first is methodological. The counter-revolution this chapter walks is the formalist anti-Keynesian stream — Friedman, Lucas, Kydland-Prescott working inside the mathematical apparatus of the discipline and pushing it against the Keynesian synthesis. The Austrian tradition, covered in ch. 6, is the parallel anti-formalist anti-Keynesian stream, with the same hostility to the postwar synthesis and the opposite methodology. Hayek and Friedman shared a great deal of policy ground; they disagreed about whether macroeconomics should be done as deductive optimization theory or as praxeological reasoning about plans and discovery. The two streams converged at the political level (both contributed to the neoliberal turn that ch. 9 traces) and diverged at the methodological level (Hayek thought the Phillips curve debate was fought on the wrong ground because the ground itself was a model the framework should not have been carrying). The chapter is the formalist half of the story.
The second acknowledgment is scope. James Buchanan and Gordon Tullock’s public choice program, dating from the 1962 publication of The Calculus of Consent, was the third anti-Keynesian movement of the period, an analytical project that applied rational-actor reasoning to political behavior and argued that government failure was symmetric to market failure. Public choice contributed to the intellectual climate in which the counter-revolution flourished, but its center of gravity is in political economy and constitutional theory rather than in macroeconomic methodology, and the chapter does not develop it further. With those acknowledgments in place: Friedman challenged the Keynesian consensus empirically. Lucas would challenge it methodologically.
You are reading where the monetarist money-primacy frame was built. This walkthrough holds it against the fiscal-theoretic alternative.
Friedman’s “inflation is always and everywhere a monetary phenomenon” dictum, and the rational-expectations scaffolding Lucas and Sargent-Wallace later hardened around it, is where the modern money-primacy frame comes from. This chapter is its intellectual origin; the walkthrough sets it against the fiscal theory of the price level, where money is not the only lever on inflation.
Friedman’s rules-over-discretion case — the one you just drove in the rule-vs-discretion plot — is the backstory to the modern central-banking debate.
The rules-versus-discretion question and the time-inconsistency result behind central-bank independence both originate here — Friedman’s “long and variable lags” case in §11.1, Kydland-Prescott’s commitment argument in §11.3. The interactive shows why a fixed rule can beat a discretionary central banker; the walkthrough carries the live modern debate over how policy should actually be set.
The decade this chapter takes as its empirical anchor is the case this walkthrough works through.
The 1970s made the natural-rate and rational-expectations moves look inevitable: the stable trade-off the synthesis relied on broke exactly as Friedman had predicted in 1967, and the discipline rebuilt its apparatus around the lesson. This chapter is the intellectual revolution the decade forced; the walkthrough is the case itself, with the counter-revolution lineage as its verdict.
Lucas asked a question so simple it was devastating. What happens to your model when the people in it figure out what you’re doing?
The Keynesian econometric apparatus, in its 1960s form, treated the economy as a system of structural equations. A consumption function. An investment function. A money-demand function. A Phillips curve. Each equation had parameters that had been estimated from the historical data; together they constituted a model that could be simulated forward under counterfactual policy assumptions. The Federal Reserve Board’s MPS model, the Wharton model, and the smaller institutional models that proliferated through the decade all worked this way. To evaluate a proposed change in monetary or fiscal policy, the analyst plugged the new policy rule into the model, simulated the resulting paths of inflation, output, and unemployment, and reported what the policy would do. This was the operational core of Keynesian policy advice. The technical apparatus was sophisticated; the methodology was internally coherent; the framework had institutional backing in the central banks and treasuries that used it. Robert Lucas, in 1976, published a thirty-page paper in the Carnegie-Rochester Conference series that demonstrated the apparatus could not be used for the purpose for which it had been built.
The argument is best walked from where the Keynesian world ended. Take the Phillips curve as ch. 8 left it: a downward-sloping short-run relationship between inflation and unemployment, augmented (after Friedman) with an inflation-expectations term where expectations are formed adaptively. The policymaker observes that pushing inflation a bit higher pushes unemployment a bit lower. The relationship has held, on average, in the historical sample. The model used to evaluate the policy treats the parameters of the Phillips curve and the expectations-formation rule as fixed. The simulation reports that a permanent increase in the inflation rate from two percent to four percent would buy a permanent reduction in unemployment of, say, half a percentage point.
Lucas’s point is that the simulation is wrong about its own coherence. The parameters that were estimated from the historical sample were estimated under a particular monetary-policy regime, one in which inflation was, on average, low, and in which surprises were one-signed. Agents in the historical sample were forming inflation expectations under that regime. Their expectations-formation rule, the rule by which last year’s realized inflation predicted this year’s expected inflation, was an optimal forecasting rule given the regime they lived under. Switch the regime, and the rule changes. Workers in a four-percent-inflation regime do not extrapolate from last year’s inflation to forecast this year’s; they update on the regime itself, factoring the higher steady-state inflation into their wage demands directly. The Phillips curve’s slope, the parameter the policymaker thought was a structural feature of the economy, is in fact a reduced-form artifact of the regime that produced the data. Change the regime and the slope changes.
The diagram below is the chapter’s central comparison. The intellectual content of the Lucas critique is what the comparison shows.
Drive the Lucas critique instead of reading it. The plane below is inflation against unemployment, with the long-run curve vertical at the natural rate u*. Start at the adaptive end and push the inflation-surprise control: the marker slides down and to the left — the policymaker buys lower unemployment. Now drag the expectations regime dial toward rational. The short-run curve rotates until it lies flat on the vertical line, and the same surprise control now moves the marker straight up — inflation rises, unemployment does not budge. Nothing changed but the assumption about how people form expectations, and that single change destroyed the policy lever.
Figure 11.1b (interactive). The expectations-augmented Phillips plane. At the adaptive end the short-run curve slopes down and the surprise lever moves unemployment; as the regime dial turns to rational, the curve collapses onto the vertical long-run curve and the lever moves only inflation. The one-off-shock button shows that even under rational expectations an unanticipated surprise briefly knocks the marker off the line. Drag the two dials; fire the shock. Model-generated illustration, not a fitted series.
Turn the expectations dial. When people stop being fooled — when they figure out the policy and price it in before it arrives — the policymaker’s lever stops working. Under rational expectations the Phillips tradeoff is vertical for any systematic policy; only surprises have real effects, and surprises by definition cannot be the systematic part of the rule.
Rational expectations sets $E_t[\pi_{t+1}] = E[\pi_{t+1}\mid\Omega_t]$, where $\Omega_t$ is the full information set including the central bank’s policy rule.
As the information set absorbs the rule, the short-run Phillips slope $\to 0$ for anticipated policy: the expectations term moves one-for-one with announced inflation, leaving the unemployment gap unmoved. The heavy derivation is economics ch. 14’s; here it is the dial.
The Keynesian forecaster had a fitted Phillips curve estimated from the data — so why couldn’t they just read a policy off it? Start in Regime A: a cloud of inflation-unemployment observations around a downward-sloping fitted line, with the policymaker’s plan drawn as a dashed arrow (“move two points down the line”). Now flip to Regime B — the new, exploiting policy. The cloud re-forms around a different relationship, because agents re-optimized; the old fitted line now sits off the new cloud, and the plan delivers only inflation. Press Refit and watch the best-fit line land somewhere visibly different. The parameters were never structural; they were an artifact of the regime.
Figure 11.1c (interactive). The Lucas critique as an econometric failure mode. The line fitted under Regime A summarizes the old regime; switch the policy and the cloud re-forms, leaving the old line off the data. The refit draws a visibly different relationship through the new cloud. Flip the regime; refit; vary how strongly agents re-optimize. A relationship estimated under one policy is not invariant to the policy — so it cannot be used to evaluate a change in the policy.
Your fitted line was a snapshot of the old rules. Change the rules and the snapshot is wrong — not because you measured badly, but because the people in the data re-optimized the moment the policy changed. That is why Sargent and Wallace concluded systematic policy is self-defeating: the act of exploiting the relationship is the act that shifts it.
Panel A is the world the Friedman-Phelps natural-rate argument operates in. Inflation expectations adjust adaptively, with a lag. The short-run Phillips curve slopes down: at any given level of expected inflation, a higher actual inflation rate is associated with lower unemployment, because the unanticipated component of inflation reduces real wages and induces firms to hire. A policymaker who pushes inflation from two percent to four percent moves along the short-run curve, trading higher inflation for lower unemployment. The trade is real but temporary. As workers and firms observe that inflation has risen and adjust their expectations, the short-run curve shifts up; at the new, higher level of expected inflation, the same unemployment rate now requires the same unanticipated inflation, which means actual inflation must be higher still to maintain the gap. Repeated exploitation produces accelerating inflation, with unemployment returning to its natural rate. The long-run Phillips curve is vertical. The short-run curve is exploitable temporarily but only at the cost of repeated upward shifts that the policymaker eventually cannot stay ahead of. This is Friedman’s world; the policy menu still exists, but the menu prices the meals at increasing inflation rates.
Panel B is the world Lucas’s critique operates in. Inflation expectations are formed rationally, which is to say agents use all available information and the correct model of the economy — including the model of the central bank’s policy rule — to forecast inflation. The expectation of inflation matches its statistical conditional mean given the information set; on average, agents are not fooled. Rational expectations as a hypothesis is John Muth’s 1961 contribution, originally proposed in the context of agricultural-price forecasting and largely ignored by macroeconomists for a decade; Lucas’s 1972 paper “Expectations and the Neutrality of Money” introduced it to macroeconomic theory and the 1976 critique paper drew the methodological consequence. Under rational expectations, anticipated changes in monetary policy, the systematic component of the central bank’s rule, are anticipated. Workers do not need to wait until next year’s data arrives to figure out that a four-percent steady-state inflation rate implies four-percent expected inflation. They figure it out in advance. They demand wage gains that incorporate the four-percent expectation directly. Firms price accordingly. The short-run Phillips curve for anticipated policy collapses onto the long-run curve. The diagram’s vertical line is the entire Phillips curve for systematic policy; only unanticipated shocks — deviations from the systematic rule the central bank could not credibly commit to — can move agents off the line.
The comparison is the chapter’s point, and what changes between the two panels is not the diagram but the expectations-formation assumption. Adaptive expectations to rational expectations is a single substitution in the model’s informational structure, and the substitution destroys the policy instrument the previous diagram had carried. If agents expect the policy, they adjust their behavior to it before it arrives, and the policy’s predicted effect on real variables vanishes. The slope of the short-run curve, which had been the parameter the policymaker was riding, is now a function of how surprised agents are by the policy. Systematic policy cannot generate surprises; surprises by definition are deviations from the system. So systematic policy cannot move unemployment, because the systematic component is the component agents have already factored into their expectations. The mathematical formalization of this argument runs through Muth’s representation theorem and the rational-expectations equilibrium concept, and the formal apparatus lives in economics ch. 14. What the chapter is doing is walking the idea, not the derivation. The idea is what survives at thought-history register, and the idea is what made the critique professionally consequential.
The Lucas critique, in its 1976 form, states the methodological consequence as a general result. Econometric models whose parameters were estimated under one policy regime cannot be used to evaluate a different policy regime, because the parameters that look structural are in fact reduced-form combinations of agents’ preferences, technology, and the policy rule the agents were facing. Switch the policy and the parameters move. The MPS model and its successors had been built by treating their estimated coefficients as deep parameters of the economy. They were not. They were policy-conditional regularities, and changing the policy invalidated the simulation. Policy invariance, the property that a parameter remains constant across policy regimes, is what an econometric model needs in order to be used for counterfactual policy evaluation. The Keynesian models lacked it. The critique was not that the Keynesian models had bad fit (their fit was often good); it was that good fit on the historical data, achieved by reduced-form estimation, was no warrant for using the model to predict what would happen under a policy that had not been in the historical data. (Where the work sits relationally: the lucas thinker node and the lucas_critique work node.)
What this required, if it required anything constructive, was that macroeconomic models be built from microfoundations: explicit specifications of agents’ preferences, technology, and information structures, with the equations of the model derived from optimization given those primitives. Parameters that lived in preferences and technology would survive policy changes; parameters that lived in agents’ expectations would adjust correctly when the policy adjusted. A microfounded model was the only kind of model whose simulations under counterfactual policy would not be contaminated by the Lucas critique. The methodological program that followed, the requirement that all serious macroeconomic models be built up from rational, optimizing agents, was the constructive consequence of the destructive demonstration. The 1976 paper did not, by itself, produce the alternative; it described what an alternative would have to look like.
The policy conclusion arrived in 1975, before the critique paper itself. Thomas Sargent and Neil Wallace published “Rational Expectations, the Optimal Monetary Instrument, and the Optimal Money Supply Rule” in the Journal of Political Economy, deriving what came to be known as the policy-ineffectiveness proposition. Inside a model with rational expectations and flexible prices, only the unanticipated component of monetary policy affects real output; the systematic, predictable component is fully neutralized by agents’ correct anticipation. Since systematic policy is by definition predictable, systematic monetary policy cannot produce real effects. Only surprise policy can, and surprises by construction cannot be systematic. The corollary, stated bluntly, is that the entire countercyclical-monetary-policy program of the postwar synthesis was empty. The Fed’s systematic responses to inflation and unemployment, the textbook stabilization rule, did nothing to the real economy. Whatever real effects monetary policy had came from random deviations from the rule, which could not be exploited.
This was a strong claim, and the assumptions it rested on (flexible prices and complete information, fully rational expectations under the model the agents are in, no nominal rigidities) were contestable then and remain contestable now. The proposition is the rational-expectations argument pushed to its logical limit. Whether the limit holds in the actual economy is a different question from whether the argument is internally valid; New Keynesian economics would later answer no to the limit while accepting the methodological force of the argument. What Sargent-Wallace did in 1975 was demonstrate that, taking Lucas’s expectational machinery seriously, the standard Keynesian conclusion that systematic monetary stabilization works is not robust. Whether the conclusion fails in degree or in kind depends on what frictions one adds; that systematic policy could fail in kind was the new fact in the room.
Lucas’s 1972 paper supplied the constructive frame within which the rational-expectations program could think about monetary non-neutrality without giving up the apparatus. The islands model imagines an economy of physically separated markets (“islands”) in which producers cannot directly observe the aggregate price level; they observe only the price in their own island. When the price in their island rises, they cannot tell whether the rise reflects an aggregate monetary shock that is raising all prices proportionately or a relative-demand shock that has shifted demand toward their good. The optimal response differs: a relative-demand shock should raise their output (reallocate resources toward the now-more-valued good); a proportionate aggregate shock should not (reallocating resources is wasteful when relative prices have not actually changed). Producers form their best estimate of which kind of shock is happening, given their information, and respond accordingly. Aggregate monetary shocks produce real effects in the short run because they cannot be perfectly distinguished from relative-demand shocks in real time; they do not produce real effects in the long run because, once aggregate data arrives, the producers’ mistakes wash out. The model demonstrates that rational expectations are compatible with short-run monetary non-neutrality through an information friction. Information, not irrationality, produces the real effects. The paper got Lucas the 1995 Nobel Prize.
The profession’s response was the question whether the critique’s reach equalled its grasp. As destructive demonstration, the 1976 paper was decisive. Within a decade, the large-scale Keynesian econometric models had largely lost their authority as policy-evaluation instruments inside the academic mainstream; central banks continued to use them, with substantial revisions, for forecasting purposes, but the claim that they could be used to evaluate counterfactual policy regimes was not defended in the way it had been before. The critique was accepted because the argument was correct: parameters that had been estimated under one regime were not invariant across regimes, and a generation of policy-evaluation work had been resting on an assumption it could not support. As constructive program, the question was harder. It was one thing to show that the existing models could not do what they claimed to do. It was another to specify what kind of model could. The microfoundations requirement was the methodological direction of travel; the destination was unclear. Was every macroeconomic question answerable inside a fully specified general-equilibrium model with rational expectations? Were the implied empirical content and policy guidance of such models adequate to what macroeconomic policy needs to do? The chapter returns to these questions in §11.4 and §11.5; in 1976 the questions were live and the answers were not in. Lucas destroyed the old Keynesian econometrics. Kydland and Prescott built the replacement.
You just drove the Phillips diagram this comparison treats as canonical — the tradeoff vanishing as expectations turn rational.
The natural-rate move (Friedman 1968, Phelps 1967) made the Phillips menu vertical in the long run; Lucas and Sargent then made it vertical even in the short run for anticipated policy. That is the lineage behind the Keynesian-versus-monetarist framework pair the comparison runs — and the expectations-augmented Phillips diagram inside this chapter is the visual it treats as canonical.
The case against systematic fiscal stimulus has an intellectual backstory — and it runs through the policy-ineffectiveness result you just watched break a fitted relationship.
The crowding-out and policy-ineffectiveness arguments — why the counter-revolution held systematic fiscal stimulus to be self-defeating — are this section’s backstory. Sargent-Wallace showed that if agents anticipate the systematic policy, only surprises move output; the fitted-relationship-break interactive shows the mechanism. The walkthrough carries the live modern fiscal-policy debate this lineage feeds.
Friedman attacked the Keynesian consensus. Lucas destroyed its methodology. Kydland and Prescott built the replacement — and in doing so, showed what the new approach could and couldn’t see.
The constructive program arrived in 1982, in a paper called “Time to Build and Aggregate Fluctuations.” Finn Kydland and Edward Prescott specified a dynamic general-equilibrium model of the US economy with a representative household, a representative firm, capital accumulation, time-to-build investment frictions, and a single source of stochastic disturbance: an exogenous technology shock following a persistent autoregressive process. Households chose consumption and labor supply to maximize the expected discounted utility of an intertemporal stream of consumption and leisure. Firms chose capital and labor inputs to maximize profits given technology. Markets cleared. The model had no money, no nominal frictions, no demand shocks, no government spending, no taxes that mattered. It had, in other words, none of the apparatus the Keynesian framework had identified as the source of business fluctuations. What it produced, when its parameters were chosen to match observed long-run features of the US economy, were artificial time series with the same statistical signature as the actual US business cycle.
The methodology was as consequential as the result. Kydland and Prescott did not estimate the model econometrically; they calibrated it. The household’s discount factor was set to match the average real interest rate. The capital share of output was set to the observed labor-share complement. The depreciation rate was set to the observed capital-output dynamics. The persistence and standard deviation of the technology shock were set to the residual of a Solow growth-accounting decomposition. With every parameter pinned down by an external moment of the data and none free, the model was simulated forward and the moments of the simulated time series compared against the moments of the observed time series. Calibration is not estimation. It does not deliver standard errors on the parameters or formal tests of the model’s fit. What it delivers is a quantitative assessment of how much of the data the model can reproduce when its parameters are forced to satisfy long-run constraints. Real business cycle theory, the program Kydland-Prescott launched, treated calibration as the appropriate response to a research question that asked not “is this model the true data-generating process?” but “how much of business-cycle fluctuation can be accounted for by an optimizing-agent model with technology shocks alone?”
Microfoundations, the requirement that macroeconomic models be derived from explicit optimization by agents with specified preferences and technology, was the methodological discipline the program imposed. The Lucas critique demanded it; the RBC program delivered it. Every parameter in the model lived in preferences, technology, or the exogenous shock process. Policy regime changes would alter the agents’ choices but not the parameters, because the parameters were structural features of the agents and the technology, not reduced-form artifacts of a particular policy environment. The program inherited the constraint Lucas had identified and used it to build something. The formal apparatus that descended from the 1982 paper, the dynamic stochastic general-equilibrium framework that became the workhorse of modern macroeconomics, lives in economics ch. 14 for the canonical RBC model and economics ch. 15 for the broader DSGE class; the chapter compresses the formal model to its load-bearing structural moves and lets the table do the moment-matching work. (Where Kydland, Prescott, and the school sit relationally: the kydland thinker node, the prescott thinker node, and the new_classical_rbc school node.)
The figure below shows the comparison the calibration exercise produces. The point of the table is not the third decimal place; the point is that a model with no demand shocks, no monetary policy, no nominal rigidities, no financial frictions, and no government produces simulated time series that, on a list of standard business-cycle moments, are within shouting distance of the observed US series.
| Statistic | US data | RBC model |
|---|---|---|
| Std. dev. of output (%) | ~1.7 | ~1.4 |
| Std. dev. of consumption / std. dev. of output | ~0.5 | ~0.4 |
| Std. dev. of investment / std. dev. of output | ~3.0 | ~2.8 |
| Correlation of hours and output | ~0.8 | ~0.9 |
| Autocorrelation of output | ~0.9 | ~0.9 |
What this bought the program, on its own terms, was a radical claim. If a model whose only stochastic disturbance is a technology shock can reproduce the major moments of the US business cycle, then the standard Keynesian apparatus (demand shocks, sticky prices, monetary non-neutrality, fiscal multipliers) is not necessary to explain business-cycle fluctuations. The fluctuations may be efficient responses of an optimizing economy to real disturbances rather than failures of demand to clear markets. The policy implications follow with force. If business cycles are efficient, stabilization policy is at best ineffective and at worst welfare-reducing. The recessions the postwar synthesis had treated as coordination failures requiring fiscal and monetary intervention were, on the RBC reading, the optimal response of households and firms to bad-technology realizations. Letting the cycle run was the welfare-maximizing policy. The model said so.
This is the place to register what the program was doing well and where it was overreaching. The methodological achievement was real. A complete microfounded general-equilibrium model that quantitatively reproduced business-cycle moments was something the discipline had not previously possessed; the RBC framework was the first attempt to take the Lucas-critique constraint seriously while delivering a working macroeconomic model. The constraint that the methodology imposes on the modeler is enormous. Every parameter must be tied to an observable feature of the economy. Every behavioral relationship must be derived from optimization. The calibration discipline forces the modeler to confront how much of the economy’s observed behavior can be accounted for by purely structural features rather than reduced-form fitting. As discipline, the program improved the standards of macroeconomic argument. As substantive claim, the program over-reached. The radical reading — that demand shocks are unnecessary, that monetary policy is irrelevant, that recessions are efficient — rested on the assumption that the technology shock the model needed to produce its moments was actually a technology shock, rather than a residual that captured everything the model had left out (financial frictions, nominal rigidities, unemployment fluctuations, anything that did not fit). That assumption became the program’s most contested commitment, and it is where §11.5 will pick up.
The same methodological moment seeded apparatus that outran the macro wars and became the load-bearing machinery of three other fields. Finance is the clearest case. Eugene Fama’s efficient-markets hypothesis (1970) carried the rational-expectations logic into asset pricing — if prices already incorporate all available information, the systematic component of any forecast is already in the price — and the Black-Scholes-Merton option-pricing model (1973) turned the same no-arbitrage discipline into a working pricing technology. The lineage is the rational-expectations program applied to financial markets; the formal apparatus is A’s to build.
Labor took the natural-rate idea and microfounded it. Phelps’s 1968 island parable, where workers search across employers under imperfect information, grew into the Diamond-Mortensen-Pissarides search-and-matching framework (the 2010 Nobel), which replaced the frictionless labor market with an explicit account of how vacancies and job-seekers meet. The natural rate stopped being a parameter and became the equilibrium of a matching process.
Trade got its modern foundation from the same cohort. Krugman’s new trade theory (1979–80) brought increasing returns and imperfect competition into trade models, explaining intra-industry trade between similar economies that the Ricardian comparative-advantage story could not. The optimizing-agent, explicit-structure discipline the counter-revolution had imposed on macro was, in these hands, the same discipline rebuilding the microeconomic fields it bordered.
Five years before the RBC paper, Kydland and Prescott had already published the result that gave the program its policy-design backbone. “Rules Rather than Discretion: The Inconsistency of Optimal Plans” (Journal of Political Economy, 1977) introduced the time inconsistency argument. Consider a central bank that has announced a low-inflation policy. Agents form their wage and price expectations in line with the announcement; with low expected inflation, the trade-off curve the central bank faces in the short run is favorable. Once the wage and price decisions have been made, the central bank’s incentive shifts. By engineering a small surprise inflation now, the bank can move output up along the now-favorable short-run curve. The optimal policy at the moment of action is no longer the optimal policy that had been announced. The announced low-inflation policy is time-inconsistent: optimal to announce, suboptimal to carry out.
The strategic implication is that agents anticipate the inconsistency. Knowing that the central bank’s ex-post incentive is to inflate, agents do not believe the low-inflation announcement; they form expectations consistent with the bank’s actual ex-post incentive, which is for higher inflation than was announced. The equilibrium is one in which the central bank produces the higher inflation rate, because once expectations are set at the higher level the central bank’s best response is in fact to validate them, and agents are not fooled. Inflation is higher than it would be under a credible commitment, output is no higher, and welfare is lower. The discretionary central bank is trapped by its own short-run incentives into delivering an outcome worse than the rule-bound bank could have achieved.
The same rule-versus-discretion plot returns, now read through Kydland and Prescott’s 1977 result. Leave the time-inconsistency framing on. Under discretion, the path does not merely overshoot — it floats up to a higher average level. The reason is the equilibrium just described: agents anticipate the banker’s incentive to renege, set their expectations at the higher inflation rate, and the banker validates them. The fixed rule, by removing the discretion, removes the temptation. Friedman’s lag argument and Kydland-Prescott’s commitment argument are two different roads to the same prescription.
Figure 11.4 (interactive). The rule-versus-discretion comparison reprised for the time-inconsistency argument. With the framing on, the discretion path sits at a higher average level even when its volatility is set aside — the inflation bias of a banker who cannot commit. Toggle the regime; drag the shock slider. Stylized illustration of the qualitative result.
A banker who can spring a surprise will be expected to spring it. So the public sets its expectations at the higher inflation rate up front, and the banker, facing those expectations, finds that delivering the higher inflation really is the best response. Everyone ends up at higher inflation with no extra output. A rule that takes the surprise off the table takes the bias off with it.
Friedman’s lag argument runs through the lag operator: a policy reacting to $\varepsilon_{t-k}$ adds variance when the reaction is mistimed relative to the shock’s own persistence.
Kydland-Prescott’s commitment gap: discretionary equilibrium inflation exceeds the rule-bound rate by the term that makes the bank’s ex-post incentive to inflate consistent with the public’s expectations — positive inflation bias, zero output gain. The formal treatment is economics ch. 15’s.
The argument provides the theoretical foundation for institutional reforms that constrain monetary discretion. Central-bank independence from political authority, inflation-targeting frameworks, statutory mandates that elevate price stability above other objectives, monetary-policy rules of the Taylor variety: each is a device for binding the central bank to a commitment that it could not credibly make on its own. The wave of central-bank-independence reforms that swept through the OECD economies and beyond from the late 1980s through the early 2000s (the Reserve Bank of New Zealand Act 1989, the Bundesbank model exported into the European Central Bank, the Bank of England’s 1997 operational independence) carries Kydland-Prescott as theoretical antecedent. The connection runs from the 1977 paper through to the modern central-banking debate that Walkthrough 06 walks at length. The constructive program was complete. The question is what it achieved.
The counter-revolution won the method war.
Below is the relational view of the arc the chapter has just walked. Three intellectual movements (Friedman’s monetarism, Lucas’s rational expectations, the Kydland-Prescott RBC program) unified by the expectations lineage that runs through all of them, with the pivotal counter-narrative work (the Monetary History) and the empirical event whose arrival the natural-rate hypothesis had predicted (stagflation).
By 1990, the methodological commitments the three waves had imposed were the operational standards of mainstream macroeconomics. Microfoundations, the requirement that macroeconomic models be built from explicit optimization by agents with specified preferences and technology, was no longer the new classical demand it had been in 1976. It was the entry condition for serious work. A graduate student writing a macroeconomic model that did not have microfoundations would not get the model into a top journal in 1990, regardless of which side of the macro wars the student was on. The Lucas critique was the binding constraint, accepted not because the new classical program had won the substantive argument but because the methodological argument was unanswerable on its own terms. Parameters that were not invariant to policy could not be used for policy evaluation; the only parameters that were plausibly invariant lived in preferences and technology; therefore the model had to be specified at the level where preferences and technology lived. Whether the model was new classical, new Keynesian, or neither, the requirement was the same.
Rational expectations as a hypothesis was similarly no longer contested. The original Lucas-Sargent-Wallace claim that expectations were correct on average given the agents’ information set was, by 1990, the modeling standard across the profession. The behavioral and bounded-rationality challenges were live, but they were challenges; the rational-expectations baseline was what the challenges were challenging. New Keynesian models, the synthesis that ch. 17 walks at length, used rational expectations as a matter of course. The microfounded household optimization. The microfounded firm optimization with sticky prices. The forward-looking Phillips curve in which expected inflation is the discounted present value of future expected marginal costs. Each of these is a New Keynesian construction, and each is built on the rational-expectations apparatus the counter-revolution forced into the profession.
What this looks like in graduate teaching is the cleanest evidence of the achievement’s permanence. The first-year macroeconomics sequence at every major US graduate program by the early 1990s opened with optimization by a representative agent over an infinite horizon, dynamic programming, the Euler equation, log-linearization around a steady state, and the rational-expectations equilibrium concept. The order in which these topics were taught was the order Kydland and Prescott had assembled them in. Whether the student went on to write New Keynesian dissertations with sticky prices, heterogeneous-agent dissertations with incomplete markets, or new-classical dissertations with full flexibility, the entry apparatus was the same. The Keynesian econometric tradition that had owned the first-year macro curriculum through the 1970s — structural-equation systems, identified through exclusion restrictions, estimated by two-stage least squares, simulated under counterfactual policies — had effectively disappeared from graduate training by 2000. The methodological replacement was complete.
What this means in operational terms is that the dynamic stochastic general-equilibrium framework, the heir of Kydland-Prescott’s 1982 model, became the workhorse of modern macroeconomic modeling. Central banks build DSGE models for forecasting and policy analysis. Academic journals publish papers that extend the DSGE framework with additional features (financial frictions, heterogeneous agents, fiscal-monetary interactions, open-economy considerations). The DSGE family tree, which economics ch. 15 traces in formal detail, has the RBC stem at its base; every modern macroeconomic model that takes optimization, rational expectations, and explicit shock structures seriously inherits, methodologically, from the program Kydland and Prescott launched. The New Keynesian models that absorbed sticky prices and demand shocks back into the framework did not abandon the framework; they extended it. The framework is what got inherited; what was contested was what to put inside it.
The rules-versus-discretion argument made by Kydland-Prescott’s 1977 time-inconsistency paper became the dominant framework for thinking about central-bank institutional design. Inflation targeting, central-bank independence, statutory price-stability mandates: each is in operation across the OECD and substantial portions of the rest of the world, and each carries the time-inconsistency argument as theoretical antecedent. The institutional landscape of monetary policy in 2010 looked nothing like that of 1970, and the difference is in significant part the working-out of an argument the 1977 paper had made. Friedman’s rules-based monetary policy did not carry through in the specific form he proposed (a constant money-growth rule, abandoned in practice through the early 1980s), but the broader Friedman commitment to rule-based rather than discretionary monetary policy, refined through Kydland-Prescott’s strategic argument and operationalized through inflation-targeting frameworks, is the dominant institutional inheritance. The forward connection to ch. 17 picks up at the New Keynesian synthesis — the repair project that absorbed the methodology while restoring substantive content the counter-revolution had excluded.
The anti-Keynesian critique that lost in real time in the 1930s is the one that came back, in formalist dress, in the chapter you are reading.
Hayek lost the 1930s argument to Keynes in real time, but the anti-Keynesian critique survived and returned — Friedman and Lucas formalized the assault the Austrians had made in praxeological terms (ch. 6 carries the anti-formalist stream). The counter-revolution this chapter walks is how the loser of the 1930s came back to win the method war by the 1980s.
The counter-revolution lost the substance war.
What was excluded from the benchmark new classical model is the catalogue of what the 2008 financial crisis was made of. Demand shocks were excluded; the RBC framework treated business fluctuations as supply-driven and treated demand-side disturbances as second-order phenomena that did not require their own modeling apparatus. Financial frictions were excluded; the canonical RBC and early DSGE models had no financial sector worth the name, no banks, no leverage, no margin constraints, no contagion mechanism through which the failure of one financial institution could threaten the solvency of others. Credit was abstracted away; the representative household borrowed and lent without regard to collateral, balance-sheet conditions, or counterparty risk. Systemic risk had no place in the framework, because the framework had no systems within which risks could become systemic. The benchmark model of 2007 had a representative household, a representative firm, a single technology shock, and rational expectations. It had nothing else.
The list of exclusions is not a list of oversights. It is a list of analytical choices, made deliberately, with reasons. The Lucas-critique constraint demanded microfoundations; microfoundations demanded that every relationship be derived from optimization; every additional friction added to the model was a parameter to be calibrated, an optimizing problem to be solved, and a layer of complexity that made the model harder to write down and harder to interpret. The new classical research strategy was to build the simplest microfounded model that produced the targeted statistical regularities, then add features incrementally as their absence became binding. Through the 1980s and 1990s, the additions came: nominal rigidities (the New Keynesian extension), heterogeneous agents (incomplete-markets models), financial intermediation (in a small subliterature). But the additions were marginal; the benchmark remained the representative-agent rational-expectations model, and the financial sector remained on the to-do list. By 2007, the canonical New Keynesian DSGE model used by central banks for policy evaluation still treated the financial sector as a residual. The Ricardian-equivalence argument, descended directly from the rational-expectations program, was one piece of this exclusion: if forward-looking agents anticipate that today’s deficit-financed tax cut implies tomorrow’s tax increase, the consumption response to the cut is zero, and fiscal stimulus through bond-financed deficits has no aggregate-demand effect. The argument’s working-out as a counter-revolution product to the Keynesian fiscal-multiplier program is part of what Walkthrough 01 on government spending traces in its modern form.
The 2008 financial crisis was the empirical event that the framework had no apparatus to explain. The bank-run dynamics, the leverage cycle, the cross-balance-sheet contagion, the freezing of the asset-backed commercial paper market, the collapse of confidence in counterparty solvency that turned a housing-market correction into a global financial crisis: each of these was a phenomenon the benchmark model could not generate, because the benchmark model did not have the financial-system structure within which they could occur. When the crisis hit, the macroeconomic models that central banks were running did not simulate forward into the trajectory the actual economy followed. The models had to be modified ex post, with new financial-frictions blocks bolted on, to produce simulations that resembled what was happening. The exercise was retrofit, not prediction. 2008 as a paradigm event was an event the dominant paradigm could not see. The narrative of the crisis itself (the housing bubble, the subprime crisis, the Lehman failure, the policy response) lives in economic history ch. 18 and ch. 19; the empirical event as a paradigm-breaking phenomenon is the subject of Walkthrough 08 on recessions. What this chapter takes from 2008 is the claim about the framework. The counter-revolution’s benchmark macroeconomic model was empty where the 2008 crisis lived. Whatever the framework was good at, it was not good at the most consequential macroeconomic event since the Depression.
Two readings of this failure compete in the historiographical literature. The mainstream reading treats the gap as an additive problem that the framework can solve from inside. Add financial frictions to the DSGE apparatus; add Bernanke-Gertler-Gilchrist financial-accelerator mechanisms; add Geanakoplos leverage cycles; add Brunnermeier-Sannikov continuous-time financial intermediation; add the He-Krishnamurthy intermediary asset pricing apparatus; the framework was incomplete in 2007, but the incompleteness can be filled in. By 2020, DSGE models with substantial financial-sector content were standard at major central banks, and the additive program had delivered a reasonable share of what was needed. The heterodox reading treats the gap as a structural problem the framework cannot solve from inside. The representative-agent assumption is incompatible with the heterogeneity that financial-system dynamics depend on; the rational-expectations assumption is incompatible with the panic-and-confidence behavior that drives bank runs; the equilibrium discipline is incompatible with the disequilibrium dynamics through which crises actually unfold. Adding ad-hoc frictions inside the framework is the same kind of move the Keynesian-econometrics tradition made when it added expectations-augmenting terms to the Phillips curve; it patches the symptom without addressing the framework’s analytical mismatch with the phenomenon. The mainstream reading is correct that the additions delivered something. The heterodox reading is correct that what they delivered is reactive rather than predictive: the framework can fit the crisis ex post, with the right additions, but the framework did not see the crisis ex ante, and the question is whether a framework that requires the right additions to be installed after the event is doing the predictive work macroeconomic theory is supposed to do.
The chapter’s position on the two evaluative claims is as follows. The methodological achievement was real. Microfoundations is a permanent constraint that has improved the rigor of macroeconomic modeling. The Lucas critique is an unanswerable result. Rational expectations as a baseline against which departures are measured is a more disciplined approach than the unspecified or implausible expectations the pre-1976 Keynesian apparatus carried. The constructive program (DSGE methodology, calibration, the discipline of tying parameters to long-run features of the data) was a step forward in how macroeconomics is done. The substantive claim, that demand shocks are unnecessary and recessions are efficient, was wrong, and the wrongness has been demonstrated by an event the framework was unable to predict and was retrofitted to explain after the fact. The two halves are independently assessable. Won the method, lost the substance.
Here is the close. Friedman and Schwartz published the Monetary History in 1963 with a central claim about the Great Depression: it was caused by monetary policy failure, by the Fed’s contraction of the money supply through 1929–33. The depth and duration of the slump were not a market failure that monetary policy could not fix; they were a Fed failure that monetary policy could have prevented. The book’s 860 pages amount to one long argument that financial-system breakdown, not demand-side stagnation, was the Depression’s engine, and that monetary policy is the lever through which financial-system breakdowns either turn into depressions or fail to. The intellectual tradition that the Monetary History helped found, the new classical and RBC programs that drew on Friedman’s methodological discipline and went further than Friedman in some directions, then built macroeconomic models with no financial sector, no mechanism for credit crises, and no role for the kind of financial-system collapse that the Monetary History had identified as the Depression’s engine. By 2007, the benchmark model the tradition had built was empty in exactly the place where its founding text had been most insistent that the lesson lived. The 2008 crisis was a financial-system collapse that the models did not simulate, because the models did not have the apparatus the Monetary History had said was the apparatus that mattered. The tradition forgot the lesson its founding text taught.
The next chapter traces the repair project — the New Keynesian synthesis that absorbed the counter-revolution’s methodology while restoring everything it had excluded. Ch. 17 picks up at the synthesis and runs through 2008 and after.
The empirical event this section turns on — 2008 as the crisis the counter-revolution’s benchmark model had no apparatus to see — is the paradigm case the recessions walkthrough works through, alongside the demand, supply, and financial-instability accounts of what a recession is.
You just read the chapter’s verdict on what the pre-2008 consensus could not see. This walkthrough asks whether the gap was fatal.
The Friedman→Lucas→Kydland-Prescott lineage built the pre-2008 consensus, and §11.5’s verdict is what that consensus could not see: a benchmark model with no financial sector and no crisis mechanism. The walkthrough takes the next step — whether the framework can repair the gap from inside (the additive reading) or whether the gap is structural (the heterodox reading).
Friedman, A Theory of the Consumption Function (1957); Friedman, “The Role of Monetary Policy” (1968); Phelps, “Phillips Curves, Expectations of Inflation, and Optimal Unemployment over Time” (1967); Friedman and Schwartz, A Monetary History of the United States, 1867–1960 (1963); Muth, “Rational Expectations and the Theory of Price Movements” (1961); Lucas, “Expectations and the Neutrality of Money” (1972); Lucas, “Econometric Policy Evaluation: A Critique” (1976); Sargent and Wallace, “Rational Expectations, the Optimal Monetary Instrument, and the Optimal Money Supply Rule” (1975); Kydland and Prescott, “Rules Rather than Discretion” (1977); Kydland and Prescott, “Time to Build and Aggregate Fluctuations” (1982); King, Plosser and Rebelo, “Production, Growth and Business Cycles” (1988); Cooley and Prescott, in Cooley ed., Frontiers of Business Cycle Research (1995). Historiographical: Snowdon and Vane, Modern Macroeconomics (2005); De Vroey, A History of Macroeconomics from Keynes to Lucas and Beyond (2016); Hoover, The New Classical Macroeconomics (1988); Mankiw, “The Macroeconomist as Scientist and Engineer” (2006); Krugman, “How Did Economists Get It So Wrong?” (2009).