Macroeconomics

National income accounting

The identity that output equals expenditure equals income, which constrains every macroeconomic claim before any theory is chosen.

Sum consumption, investment, government spending, and net exports on the expenditure side, and label the same total on the income side as wages, profits, rents, and taxes net of subsidies. The three ways of counting the same flow must agree.

Every dollar spent is a dollar received, and every dollar produced is a dollar sold or added to inventory. The identity is a matter of definition, so a theory cannot violate it and must instead say which term is doing the moving.

The big idea

Any closed system of accounts with a defined boundary. Fails as a behavioral claim, since the identity is silent on causation and can be satisfied under contradictory stories.

Before any school of thought gets to argue, there is a piece of bookkeeping every school must obey. Whatever a country produced in a year was bought by somebody, and whatever that somebody paid became another person’s income. Count the output, count the spending, count the income, and all three totals must agree, because they are one flow measured at three points. The statement explains nothing on its own. Its value is that it constrains every explanation, since any story about one term moving owes an account of which other term moved with it.

Expenditure identity as constraint

The identity is written

$$Y = C + I + G + (X - M)$$

with output on the left and its four expenditure destinations on the right. Nothing in the equation is estimated. It is enforced by the statistical agency, which reconciles the production, expenditure, and income accounts through a residual so the three totals coincide. The Bureau of Economic Analysis publishes that residual, the statistical discrepancy, in the National Income and Product Accounts release each quarter.

symbol	meaning
$Y$	gross domestic product for the period
$C$	household consumption
$I$	gross private domestic investment, including inventory change
$G$	government consumption and gross investment
$X - M$	exports minus imports

Nominal against real

Nominal GDP mixes quantities and prices; real GDP strips the price movement out and tracks quantities only. The textbook version holds a base-period price vector fixed, but the published American series has not worked that way since 1996: the BEA chains together Fisher indexes built from each adjacent pair of years, which is why chained-dollar components do not sum to the chained-dollar total. The GDP deflator is the ratio of the two, and it moves for reasons the CPI does not: the deflator reweights each quarter to what was actually produced, while the CPI reweights on a lagged basket of consumer purchases. When investment prices fall and consumer prices rise, the two indexes disagree in sign, and the choice of which one deflates a nominal contract is a substantive one.

What GDP omits

Household production is outside the boundary, so a meal cooked at home does not enter GDP and the same meal bought at a restaurant does. The convention understates output in economies with more home production and overstates growth in economies moving from home to market. Depletion of natural capital is not deducted, so drawing down a fishery or an aquifer raises measured output while destroying the productive base. Quality change is adjusted for by several methods of uneven reach, hedonic regression being the most discussed but matched-model comparison and direct cost-based adjustment doing most of the work; the Bureau of Labor Statistics applies hedonics to computers, semiconductors, and a short list of other categories, and quality adjustment of any kind reaches most services poorly.

Revisions and the first print

The advance estimate of quarterly GDP is a projection built from partial data, and it is revised twice within three months and again in the annual and comprehensive revisions. The Federal Reserve Bank of Philadelphia’s Real-Time Data Set shows the historical distribution of revisions from advance to latest: the mean absolute revision in the quarterly growth rate is on the order of one percentage point at an annual rate, and revisions occasionally change the sign. A recession call cannot rest on an advance number, and macroeconomic models estimated on final data are being fit to a series the policymaker never saw.

Money, inflation, and the quantity theory

The relation between the money stock and the price level, stated as an identity and then given behavioral content by assumptions about velocity.

Write MV = PY as an accounting statement, then close it with an assumption: constant V gives the classical quantity theory, constant P gives full accommodation, and a stable money-demand function in between gives the monetarist prediction that money growth in excess of real growth becomes inflation.

Every transaction settles in money, so the flow of money spending equals the flow of nominal output by construction. The theory content is the claim about which term absorbs a change in another.

The big idea

Any economy with a single unit of account. Loses traction when velocity is unstable, when the definition of money shifts, or when the transmission from base money to broad money is broken, all of which happened after 2008.

How much money exists, how often each unit changes hands, what things cost, and how much gets made: those four quantities are tied together by an equation that holds by definition, in the way that distance equals speed times time holds by definition. No claim about the world has been made yet. The claims arrive when somebody asserts that one of the four stays put while another moves. Disagreements over which one stays put separate entire schools of macroeconomics from one another.

Identity and its schools

$$MV = PY$$

The identity has four terms, and each school picks one to hold approximately fixed and lets the others move. Irving Fisher, in The Purchasing Power of Money (1911), treated $V$ as fixed by payment technology and $Y$ as fixed by real factors, so a change in $M$ moved $P$ one for one. Milton Friedman, in the restatement collected in Studies in the Quantity Theory of Money (1956), replaced Fisher’s transactional velocity with a demand for real balances that is a stable function of a small number of variables, and drew the same conclusion by a longer route: sustained excess money growth becomes inflation.

symbol	meaning
$M$	money stock, on whichever aggregate is chosen
$V$	velocity, the average number of times a unit circulates per period
$P$	price level
$Y$	real output

Aggregates and the definition of money

The Federal Reserve tracks M1, which is currency, demand deposits and other liquid deposits, and M2, which adds small time deposits and retail money market funds. Those definitions are themselves recent, and the revision makes the point better than any argument: until May 2020 savings deposits sat in M2 and outside M1, and when Regulation D stopped limiting withdrawals from them the Fed moved them into M1, which more than tripled the reported M1 overnight without a dollar being created. Both velocities fell steeply after 2008, as households and firms held more deposits once interest rates sat near zero, and the Fed had already discontinued the broader M3 aggregate on March 23, 2006. That any of this is a decision about the boundary of “money” is the substantive point. The quantity theory says nothing until the aggregate is chosen, and the choice determines whether the theory looks right or wrong in a given decade.

Seigniorage and the fiscal theory

A government issuing base money earns seigniorage equal to the growth of that base times the real balances the public holds. When the primary deficit exceeds what taxes and non-monetary borrowing can cover, the residual is monetized and the inflation rate rises to whatever level is needed to close the fiscal gap. Thomas Sargent and Neil Wallace, in Some Unpleasant Monetarist Arithmetic (1981), showed that a central bank with an inflation target can lose control of inflation to the fiscal authority even when it holds money growth constant, because the eventual monetization is anticipated. The fiscal theory of the price level, developed by Christopher Sims and by Michael Woodford in the 1990s, pushes this further and treats the price level as the variable that makes real government debt equal to the present value of primary surpluses.

Hyperinflation

Hyperinflation is the natural experiment that separates real from nominal. Phillip Cagan, in a chapter of Friedman’s Studies volume, fit a money demand function to seven post-1918 hyperinflations and showed that real balances contract as inflation rises, so velocity is a function of expected inflation rather than a constant. Sargent, in The Ends of Four Big Inflations (1982), documented that the German, Austrian, Hungarian, and Polish hyperinflations of the early 1920s ended in weeks once the fiscal regime changed, not in the years a gradualist model of price rigidity would predict. The rational expectations reading of that record is that agents responded to the credible regime change rather than to the current money growth rate.

Business cycles and their measurement

The recurring but irregular fluctuation of output around trend, and the filters used to separate the two.

Choose a decomposition of the level series into a trend and a cyclical residual, then work with the residual. The choice is a modeling assumption disguised as a preprocessing step.

Long-run growth and short-run fluctuations are driven by different forces, so mixing them together makes both harder to see. The cost is that the trend is unobserved and every filter imposes one.

The big idea

Any macroeconomic time series with both a low-frequency drift and a higher-frequency cycle. Fails when the trend itself has structural breaks, which the standard filters treat as cycle.

An economy does two things simultaneously. It grows across decades and it wobbles across quarters. Nearly every question about recessions concerns the wobble, so the growth has to come out first. The trouble is that nobody ever observes the growth path on its own. It has to be estimated, and whichever tool does the estimating decides what is left over to count as a wobble. A step that looks like routine data cleaning is therefore a modeling decision with consequences for every result downstream of it.

Trend against cycle

A linear time trend, a deterministic polynomial trend, the Hodrick-Prescott filter, the Baxter-King band-pass filter, and the Beveridge-Nelson decomposition give different cyclical components on the same real GDP series, and the disagreement is not measurement error. It reflects different assumptions about whether shocks are permanent or transitory, which is the substantive question that macroeconomics is trying to answer with the residual. Whether a shock is permanent is itself a testable property of the series, and unit roots and cointegration are how the test is run.

Hodrick-Prescott and its critics

Robert Hodrick and Edward Prescott circulated their filter as a working paper in 1980 and published it in Journal of Money, Credit and Banking in 1997. The filter picks a trend $\tau_t$ that minimizes a weighted sum of squared deviations from the series and squared second differences of the trend, and the weight $\lambda$ is set to 1600 for quarterly data by convention. James Hamilton, in Review of Economics and Statistics in 2018, showed that the filter manufactures spurious dynamic relations that are not in the underlying data, that it produces end-of-sample values very different from the middle, and that the standard $\lambda$ has no basis in the frequency content of the series. Hamilton’s replacement is a regression of the series on its own lags from two years earlier.

Okun’s law

Arthur Okun, in a 1962 note for the Cowles Foundation, fit a linear relation between the output gap and the unemployment gap and reported that a one-point rise in unemployment coincided with about a three-point fall in output relative to potential. The coefficient has drifted since. Postwar United States data from 1949 to 2019 give an Okun coefficient closer to two, and the relation is more stable in first differences than in levels. The value of the coefficient is that it lets a nowcast of GDP be built from labor-market data that arrives sooner and is revised less.

Dating and the NBER committee

Recessions in the United States are dated by the Business Cycle Dating Committee of the National Bureau of Economic Research, chaired since 2024 by Valerie Ramey of Stanford. Robert Hall was a member from 1978 to 2024. The committee’s rule is not two consecutive quarters of negative real GDP growth. That popular formula would have missed the 2001 recession, in which quarterly real GDP declined in the first and third quarters but not consecutively. The committee weighs a menu including nonfarm payroll employment, personal income less transfers, industrial production, and real personal consumption expenditures, and announces dates months after the fact.

Persistence and asymmetry

Expansions in the postwar United States average about five years and contractions about ten months, so the cycle is asymmetric in duration. It is also asymmetric in speed: unemployment falls slowly during a recovery and rises quickly during a recession, a regularity that Salih Neftci tested formally in the Journal of Political Economy in 1984, and that James Hamilton’s two-state Markov switching model of output growth in Econometrica in 1989 built into the cycle itself. Any linear filter on the level series discards this asymmetry, which is one reason the postwar record looks smoother in filtered data than it feels in payroll data.

Keynesian economics

The claim that output is demand-determined in the short run because prices and wages do not clear markets fast enough, which makes aggregate demand a policy target.

Treat prices and money wages as sticky over the horizon of a cycle, so a fall in demand shows up as lower output and employment rather than lower prices. Fiscal or monetary policy that raises demand raises output until the resource constraint binds.

Coordinating a wage cut across every employer and worker is a problem no market instrument solves quickly. Firms cut hours before they cut nominal wages, and the resulting income shortfall in one sector becomes a demand shortfall in another.

The big idea

Economies with money contracts, discrete price adjustment, and a monetary or fiscal instrument that can shift nominal demand. Loses traction under hyperinflation, where price and wage rigidity break down and the classical dichotomy reasserts.

If demand for everything falls, the textbook response is that prices fall alongside it and the same quantity still sells. Keynes observed that this is not what occurs, because prices and wages above all are written into contracts and revised slowly. A firm facing weaker demand cuts output and staff long before it cuts anyone’s nominal pay. One firm’s wage bill is another firm’s customer base, so the shortfall travels. Under that account, total spending becomes a quantity a government can move and has reason to.

Effective demand and the paradox of thrift

John Maynard Keynes, in The General Theory of Employment, Interest and Money (1936), rejected the classical proposition that saving and investment are equilibrated by the interest rate and argued that they are equilibrated by income. If every household tries to save more at once, aggregate income falls until desired saving matches investment at the new lower level, and the attempt to save more results in less actual saving. This is the paradox of thrift, and it is the mechanism through which a demand shortfall becomes a recession.

Consumption function and the multiplier

Keynes wrote consumption as a stable function of current income with a marginal propensity to consume between zero and one. If that propensity is $c$, an autonomous increase in spending of one dollar raises income by $1/(1-c)$ dollars through the multiplier, because the recipients spend a share $c$ of what they receive and the recipients of that spending do the same. Empirical multipliers estimated in the postwar United States are much smaller than the closed-economy textbook implies, on the order of 0.5 to 1.5 depending on the specification, because the propensity to consume is not stable, imports leak, monetary policy responds, and Ricardian households anticipate the tax bill.

Liquidity preference and the trap

Keynes derived money demand from three motives, of which the speculative motive is the analytically important one: when interest rates are expected to rise, holding money dominates holding bonds, and the demand for money becomes flat in the interest rate. The liquidity trap is the region where an increase in the money stock is absorbed as idle balances at unchanged interest rates, so monetary policy loses traction. The zero lower bound on nominal short rates, binding in Japan for most of the years from 1999 to March 2024, when the Bank of Japan ended its negative interest rate policy, and in the United States from December 2008 to December 2015, is the modern instance of the same condition, and it is what forced central banks into asset purchases as a substitute for further rate cuts.

IS-LM as a device

John Hicks, in Econometrica in 1937, formalized Keynes’s argument as the IS-LM diagram: an investment-saving curve on which the goods market clears and a liquidity-money curve on which the money market clears, with output and the interest rate determined by their intersection. Franco Modigliani, in Econometrica in 1944, sharpened the wage assumption and made explicit that the model behaves classically when wages are flexible and Keynesian when they are rigid. The device is a teaching tool that suppresses inflation expectations, has no explicit intertemporal structure, and treats the money stock as the instrument, none of which matches how policy is now conducted. The three-equation new Keynesian model has replaced it in graduate teaching for those reasons.

Original Phillips curve

A. W. Phillips, in Economica in 1958, fit an inverse relation between the unemployment rate and the rate of change of money wages in United Kingdom data from 1861 to 1957 and drew a scatter plot that appeared to show a stable tradeoff. Paul Samuelson and Robert Solow, in an American Economic Review Papers and Proceedings piece in 1960, translated Phillips’s fit into a tradeoff between unemployment and price inflation and treated the frontier as a menu for policy choice. The interpretation that policymakers could pick a point on the curve was empirically defensible in the 1960s data and empirically ruined by the 1970s data, which motivated the work the next section describes.

Monetarism

The position that inflation is a monetary phenomenon and that discretionary demand management does more harm than a rule, associated with Friedman and the Chicago department.

Restore the quantity theory as a long-run proposition, add a natural rate of unemployment to which the economy returns, and treat monetary policy's short-run traction as coming from unexpected money. Prefer a constant money-growth rule to discretion.

Policy lags are long and variable, so discretion aimed at smoothing output will as often amplify it. A rule removes the destabilizing move by removing the choice.

The big idea

Economies with a stable demand for money and a central bank able to control the aggregate. Both conditions weakened after 1980 and became severely strained after 2008, which is when the monetarist program lost its operational bite.

Grant the previous section its claim that demand management works, then ask a narrower question: does it work reliably enough to be worth attempting? Monetarism answers no. The effects of any measure arrive after a delay nobody can predict in advance, so an intervention aimed at cushioning a downturn often lands during the recovery and inflates the next boom instead. The recommendation that follows is to stop choosing at all. Expand the money supply at a fixed announced rate and eliminate the opportunity to mistime anything.

Friedman on consumption and the multiplier

Milton Friedman, in A Theory of the Consumption Function (1957), argued that households consume out of permanent income, meaning the annuity value of expected lifetime resources, rather than current income. The permanent income hypothesis attacked the Keynesian multiplier at its base: transitory tax cuts and transfers are mostly saved because they do not raise permanent income, so the multiplier from a temporary fiscal stimulus is small. The empirical debate is not settled. Modern estimates from Johnson, Parker, and Souleles on the 2001 and 2008 tax rebates find that liquidity-constrained households spend a large share of a temporary rebate, which reconciles a small aggregate multiplier with a large one for that subset.

Constant money growth rule

Friedman argued that the lags between a monetary action and its effect on output and prices are long and variable, on the order of six to twenty-four months and different from cycle to cycle. A central bank trying to lean against the current output gap will as often add to the next one, so a rule that grew a chosen monetary aggregate at a constant rate would beat discretion. The design recurs well outside central banking: Bitcoin fixes its issuance schedule at network launch and removes discretion entirely, a case cryptocurrency sets out. Paul Volcker’s Federal Reserve targeted nonborrowed reserves from October 6, 1979 to October 1982 in a partial trial of the idea, brought the federal funds rate above 19 percent in mid-1981, and drove the unemployment rate to 10.8 percent in November 1982. Inflation fell from 13.5 percent in 1980 to 3.2 percent in 1983. The disinflation succeeded, but the rule was abandoned because M1 velocity became unstable in the deregulated 1980s, which broke the link between the aggregate and nominal spending.

Natural rate hypothesis

Friedman, in his American Economic Association presidential address on December 29, 1967, published as The Role of Monetary Policy in American Economic Review in March 1968, argued that the Phillips tradeoff is a tradeoff between unemployment and unexpected inflation, and that unemployment returns to a natural rate determined by real factors once expectations catch up. Edmund Phelps, in Economica in 1967, reached the same conclusion from a different setup that made the tradeoff explicit as an intertemporal choice for the central bank. Both papers predicted, before the fact, that a policy exploiting the Phillips curve would end with the same unemployment rate and higher inflation.

Expectations-augmented Phillips curve

The augmented form is

$$\pi_t = \pi_t^e - \beta (u_t - u^*)$$

where $\beta > 0$, so inflation runs below expectations when unemployment sits above its natural rate and above them when it sits below. Inflation depends on expected inflation and on the unemployment gap, not on the level of unemployment. If expectations are formed rationally and $u_t = u^*$ on average, the long-run Phillips curve is vertical at the natural rate. The 1970s stagflation, in which United States unemployment rose from 3.5 percent in 1969 to 8.5 percent in 1975 while CPI inflation rose from 5.5 percent to over 11 percent at its 1974 peak, is the episode the Friedman-Phelps prediction turned on. The original Phillips relation could not accommodate the joint move, and the augmented version could.

Stagflation as the turning episode

The response inside the profession was fast. By the mid-1970s the expectations-augmented Phillips curve was in graduate textbooks, and by the early 1980s the vertical long-run curve was the mainstream view across the ideological range. The monetarist program was in that sense a decisive intellectual victory. Its operational program, targeting a monetary aggregate as the instrument of policy, was abandoned within a decade of Volcker because the aggregates lost their empirical connection to nominal spending, and the successor at every major central bank is an interest rate rule of the kind Taylor formalized in 1993.

Rational expectations and the Lucas critique

The argument that agents do not repeat forecast errors, so the estimated coefficients of a macroeconomic model shift when the policy rule that generated the data changes.

Impose that subjective expectations equal the mathematical expectation implied by the model itself, then check whether the estimated behavioral relations are invariant to the policy rule. Reduced-form Phillips and consumption relations fail this check.

Agents' decisions depend on the policy they expect. When the policy changes, the decision rules change, so a regression estimated under the old rule mispredicts under the new one. Deep parameters, such as tastes and technology, are what survive the change.

The big idea

Policy simulation and counterfactuals. Requires agents to have enough information and computation to solve the model, an assumption weakened by the behavioral and bounded-rationality literatures.

A model fitted to past data quietly assumes people will go on behaving as they did while that data was being collected. They will not, if the reason they behaved that way was the policy in force at the time. Announce a new policy, people adjust, the historical relationships shift, and the model that fit yesterday mispredicts tomorrow. The argument bites harder than a complaint about imprecision. It says the estimated coefficients of a macroeconomic model are not the stable objects a generation had treated them as.

Muth’s consistency condition

John Muth, in Econometrica in 1961, defined rational expectations as the requirement that agents’ subjective probability distribution over an outcome equal the objective distribution implied by the model. Rational expectations is a consistency condition on the model, not an optimality claim about agents, and it is compatible with agents making forecast errors as long as those errors are unbiased and uncorrelated with the information used to form the forecast. The condition disciplines the modeler more than the agent. The warning generalizes past macroeconomics to any learned policy deployed into an environment that reacts to it, which is where Goodhart’s law and the Lucas critique sit as the same objection stated twice.

Content of the critique

Robert Lucas, in Econometric policy evaluation: A critique (1976), argued that the parameters of the estimated Phillips curve, the estimated consumption function, and the estimated money demand function are convolutions of preferences and the policy rule. Change the rule and the parameters change. A simulation that fixes those parameters and varies the rule is fitting a stable relation to a series whose stability depends on the rule being fixed, and its projections are unreliable in exactly the case that motivated the simulation. The critique invalidated a generation of large-scale econometric models used for policy analysis at the Federal Reserve and elsewhere.

Anticipated money is neutral

Thomas Sargent and Neil Wallace, in Journal of Political Economy in 1975, showed that in a model with rational expectations and a Lucas-style aggregate supply curve, only unanticipated money moves output. Any systematic policy rule is embedded in expectations, so its predictable component has no real effect. The policy ineffectiveness proposition was a strong empirical claim that made the natural rate hypothesis into a testable prediction: the correlation between output and the anticipated component of money should be zero, and the correlation with the surprise component should be positive.

Time inconsistency and commitment

Finn Kydland and Edward Prescott, in Rules Rather than Discretion: The Inconsistency of Optimal Plans (1977), showed that a benevolent policymaker who cannot commit will produce systematically worse outcomes than one who can, because the best action at each date is not the action promised at the previous date. Once expectations of low inflation are set, the policymaker gains from surprise expansion, and rational agents anticipate the surprise, so equilibrium inflation is above the target and output is at the natural rate. The result gives the case for a central bank with an inflation target enforced by an accountable mandate, and it is the intellectual basis for the operational independence granted to the Bank of England in 1997 and written into the European Central Bank a year later. The Federal Reserve is the awkward case for that story: its independence was settled by the Treasury Accord of 1951, a quarter century before the paper, so there the argument rationalized an arrangement that already existed rather than producing it.

On June 29, 2026, the Supreme Court denied the government’s stay application in Trump v. Cook (No. 25A312) by a 5-4 vote, leaving Governor Lisa Cook in office pending litigation. On October 9, 2026, the White House named a three-member committee to investigate mortgage-fraud allegations against her and report on her removal, with a closed hearing set for November 5, 2026. On October 14, her lawyers told the White House that she would be at financial and monetary policy meetings in Europe on that date and could appear no earlier than after the Federal Reserve’s December 8–9 policy meeting. The Court said that accepting the government’s removal arguments would turn the Federal Reserve’s statutory for-cause protection into at-will employment.

Deep parameters

The critique implied a research program: rebuild macroeconomic models on deep parameters, meaning the preferences, technology, and information structure that a policy rule cannot shift. The dynamic stochastic general equilibrium models that followed, and that dominate central bank practice today, are the direct answer to Lucas’s challenge. Whether their behavioral relations are actually invariant to policy shifts is contested and depends on which parameters are called deep, but the direction of the answer is the one the critique demanded.

Real business cycle theory

The model in which cycles are the efficient response of optimizing agents to technology shocks, so fluctuations require no correction.

Take a stochastic version of the neoclassical growth model with intertemporal labor supply, drive it with shocks to total factor productivity, calibrate parameters to microeconomic evidence and long-run averages, and simulate. Compare the second moments of simulated series to their empirical counterparts.

If productivity varies and agents optimize, hours worked and investment respond in the direction that maximizes expected utility. What looks like a recession is the efficient response to a temporary fall in productivity.

The big idea

Cycles driven predominantly by supply shocks. Loses traction on cycles driven by demand or by financial disruption, since both are absent from the baseline model.

Take the preceding critique seriously enough and macroeconomics has to be rebuilt from individual decisions upward, with people forecasting correctly and markets clearing. Build it that way and recessions stop resembling failures. If the economy has genuinely turned less productive for a spell, then working less and investing less is exactly what a well-informed person would choose to do. The conclusion is deliberately provocative, and its main effect was to force everyone who disagreed to name the assumption they were prepared to abandon.

Technology shocks and the Solow residual

Robert Solow, in Review of Economics and Statistics in 1957, decomposed output growth into contributions of capital, labor, and a residual he attributed to technical change. The Solow residual, computed as $\log Y - \alpha \log K - (1-\alpha) \log L$, is used in real business cycle work as the empirical proxy for the technology shock that drives the model. The residual is procyclical, has autocorrelation near 0.95 in quarterly United States data, and by construction absorbs any variation not accounted for by the measured inputs.

Intertemporal labor substitution

The baseline real business cycle model needs a large elasticity of labor supply with respect to the wage to match the observed volatility of hours worked. Finn Kydland and Edward Prescott, in Time to Build and Aggregate Fluctuations in Econometrica in 1982, built the canonical model with time-to-build in investment and a representative household choosing hours over time in response to expected wage movements. The elasticity the model needs, on the order of two to four, is above the elasticity that microeconometric studies on prime-age men have found, which is close to zero, and this wage-hours puzzle is the empirical strain that the intertemporal substitution mechanism has to bear.

Calibration against estimation

Kydland and Prescott chose parameters to match a small set of long-run moments in the United States data rather than estimating them from the second moments they were trying to explain. Calibration was defended as making the exercise a stress test rather than a fit: how much of observed fluctuation could a model with parameters chosen from independent evidence generate? Lawrence Christiano and Martin Eichenbaum, in a 1988 NBER working paper published in American Economic Review in 1992, argued that the models failed the test of the near-zero correlation between hours and wages, and by the 2000s the estimation of medium-scale extensions using Bayesian methods, as in Smets and Wouters in 2007, had displaced pure calibration in the frontier work.

Welfare and stabilization policy

The baseline real business cycle model is a first-best economy, so a stabilization policy that damped the cyclical response would move households onto a worse consumption path. Prescott, in Theory Ahead of Business Cycle Measurement (1986), stated this welfare implication squarely: the cost of business cycles is small and the cost of intervening in them is not. Robert Lucas, in Models of Business Cycles (1987), estimated the welfare cost of postwar United States consumption fluctuations at a small fraction of one percent of consumption per year, which sharpened the conclusion.

What the approach cannot explain

Money plays no role in the baseline model and is neutral by construction, so the model cannot generate an inflation-output correlation from monetary policy shocks. Unemployment is not a state; the labor market clears, and any spell of nonemployment is chosen leisure. And the model cannot generate the Great Depression, whose measured technology regress would have to be implausibly large. Prescott and Timothy Kehoe, in later Minneapolis work, extended the approach to the 1930s by adding policy distortions, but the exercise required layered additions that were not part of the original theory.

New Keynesian synthesis

The current workhorse, which keeps rational expectations and optimizing agents while restoring nominal rigidity, so demand matters and policy has traction.

Combine an intertemporal Euler equation, a forward-looking Phillips curve derived from staggered price setting, and a Taylor rule for the nominal interest rate. Estimate the parameters by Bayesian methods on postwar data and use the fitted model for forecasting and policy analysis.

Monopolistic competition gives firms a price to set, staggered contracts give a fraction of them a period in which they cannot reset, and the resulting sticky-price aggregate supply relation has the same form as the Phillips curve while inheriting rational expectations.

The big idea

Advanced economies with an interest-rate instrument and moderate inflation. Struggled at the zero lower bound after 2008, missed the housing boom, and required financial-friction extensions that the baseline model did not have.

The two preceding traditions each held something the other needed. One brought discipline about expectations and individual decisions. The other brought an account of why downturns hurt and why policy can relieve them. The synthesis keeps the first and readmits the second by giving firms a price of their own to set, then preventing most of them from resetting it in any given period. That one friction suffices to make demand matter again inside an otherwise fully optimizing model, and the result is the machinery central banks now run.

Calvo pricing

Guillermo Calvo, in Journal of Monetary Economics in 1983, wrote a model in which a fraction $1 - \theta$ of firms can reset their price in each period and the remaining $\theta$ hold the previous price. The staggered structure has no time index on which firm resets, which makes the aggregation tractable but implies an implausible age distribution of prices. Nakamura and Steinsson, in the Quarterly Journal of Economics in 2008, showed that United States consumer prices have a median duration of about eight to eleven months excluding sales, so a quarterly $\theta$ of 0.75 in the standard calibration, which implies a mean duration of four quarters, sits just above that range.

Monopolistic competition

Price setting requires that firms have a price to set, which requires that they face a downward-sloping demand curve, which requires monopolistic competition. Avinash Dixit and Joseph Stiglitz, in American Economic Review in 1977, provided the constant-elasticity aggregator that the new Keynesian literature adopted wholesale. The aggregate price level is a symmetric average of firm prices, and the fraction of firms that cannot adjust in a given period is the wedge through which nominal shocks have real effects. Without market power there is no price to set and the model collapses to a real business cycle model with money neutral.

Three-equation model and the Taylor rule

Richard Clarida, Jordi Gali, and Mark Gertler, in The Science of Monetary Policy: A New Keynesian Perspective (1999), wrote the compact three-equation form: a forward-looking IS curve, a new Keynesian Phillips curve, and an interest-rate rule. John Taylor, in Discretion versus policy rules in practice (1993), had already shown that the Federal Reserve’s federal funds rate from 1987 to 1992 tracked the rule

$$i_t = r^* + \pi_t + 0.5 (\pi_t - \pi^*) + 0.5 (y_t - y^*)$$

where $i_t$ is the policy rate, $r^*$ the equilibrium real rate, $\pi_t$ inflation and $\pi^*$ the target, and $y_t - y^*$ the output gap. The Taylor principle, that the coefficient on inflation exceeds one so real rates rise when inflation rises, is the stability condition for the model.

DSGE and the 2008 forecast failure

Frank Smets and Rafael Wouters, in American Economic Review in 2007, estimated a medium-scale dynamic stochastic general equilibrium model on United States postwar data using Bayesian methods, and versions of this model were in use at the European Central Bank, the Board of Governors, and central banks worldwide by 2008. None of them forecast the crisis, and their post-2008 forecasts of inflation and output overpredicted both. Volker Wieland and Maik Wolters, in Economic Theory in 2011, compared real-time forecasts from six models of the United States economy with Greenbook and Survey of Professional Forecasters forecasts across the 2007 to 2009 recession and the four before it, and found the mean model forecast about as accurate as the professional ones.

Financial frictions after the crisis

Ben Bernanke, Mark Gertler, and Simon Gilchrist, in The Financial Accelerator in a Quantitative Business Cycle Framework (NBER working paper 6455, 1998), had already built external-finance premia into a new Keynesian model, but their extension was not in the workhorse models the central banks used in 2007. After 2008 the frontier work added banking sectors and endogenous credit spreads (Gertler and Kiyotaki, in a chapter of the Handbook of Monetary Economics in 2010; Christiano, Motto, and Rostagno, in American Economic Review in 2014), and central bank models were re-estimated on data that included the crisis window. Whether the new versions would have flagged 2008 in real time is a counterfactual the discipline has not answered.

Austrian economics

A tradition rejecting aggregation and equilibrium modeling, treating prices as the mechanism transmitting dispersed knowledge no planner holds.

Reason from methodological individualism about how agents act on local knowledge, argue that the price system aggregates that knowledge in a way no central authority can replicate, and diagnose macroeconomic disturbances as consequences of interference with the interest rate.

Prices carry information about relative scarcity that is not available in any centralized form. A distortion of the interest rate by monetary policy misallocates investment across periods, so the correction is a recession.

The big idea

Explanations of the informational role of markets and the microeconomic incentives around monetary distortion. Loses traction as a program of empirical macroeconomic work because it rejects the aggregation and formalization that empirical work requires.

Every school above works with aggregates: total output, the price level, the money stock. The Austrian tradition objects to the aggregates themselves. Its central claim is that a price is a signal carrying knowledge existing only in scattered fragments across millions of people, not one of whom holds the whole picture. Averaging prices into an index discards exactly the information that makes the system function, and no planner can reconstruct it. That was the substance of a long argument over whether a planned economy could calculate at all.

Mises and socialist calculation

Ludwig von Mises, in Economic Calculation in the Socialist Commonwealth (1920), argued that a socialist economy without private ownership of the means of production has no market prices for capital goods, and therefore no way to compare the productivity of alternative uses. Oskar Lange, in Review of Economic Studies in 1936 and 1937, replied that a planner could iterate prices in a tatonnement algorithm and reach the same allocation, and the debate ran into the 1940s. The 1989 to 1991 collapse of central planning in Eastern Europe is read by Austrian writers as the historical vindication of Mises’s argument, though the interpretation is contested.

Hayek on knowledge

Friedrich Hayek, in The Use of Knowledge in Society in American Economic Review in 1945, sharpened the argument by locating the informational problem in dispersed, tacit knowledge rather than in the number of variables. A price system, Hayek argued, is a communication device that lets each agent act on what they know locally while adjusting to changes far away that they will never learn about explicitly. The essay is the most cited Austrian contribution and is influential outside the school; its argument about the informational content of prices is standard in mechanism design and in the literature on the theory of the firm.

Austrian business cycle theory

Hayek, in Prices and Production (1931), and Mises, in The Theory of Money and Credit (1912), argued that a central bank suppressing the interest rate below the natural rate induces firms to undertake longer-horizon projects than voluntary saving supports, so the boom builds a capital structure that the eventual return to a market-determined rate cannot sustain, and the recession is the unavoidable correction. The account was defeated in a famous 1930s exchange between Hayek at the London School of Economics and Keynes at Cambridge over how to interpret the Great Depression, and the profession went with Keynes. The theory has had a Ron-Paul-era political revival without a corresponding empirical one.

Methodological individualism

The Austrian program rests on methodological individualism, the position that macroeconomic aggregates are analytical fictions and that explanation must reduce to statements about individual choice. Mises, in Human Action (1949), rejected econometric testing on the ground that human action is not subject to constant parameters, so a coefficient estimated in one period carries no information about the next. The stance disallows the empirical machinery that the rest of the profession uses to adjudicate between theories, which is why Austrian claims cannot be refereed against the same criteria that separate a new Keynesian model from a real business cycle model.

Why it stayed outside the mainstream

Two forces kept the school outside the frontier. First, the refusal to formalize meant the Austrian arguments could not be stated as models that generate testable moments, so they could not enter the comparative exercises that decide which framework central banks adopt. Second, the same arguments about dispersed knowledge that give Hayek his enduring reputation were absorbed into mainstream micro theory without adopting the rest of the program, so a modern economist can accept the informational role of prices without accepting Austrian business cycle theory. The school retains an active research literature at George Mason and at the Mercatus Center, and a policy-outreach presence through the Cato Institute and the Mises Institute.

Post Keynesian economics

A heterodox tradition holding that fundamental uncertainty is not risk, so expectations cannot be rational in the technical sense and financial instability is endogenous.

Take money as endogenously created by bank lending, treat the future as uncertain in the Knight-Keynes sense rather than probabilistically risky, and let the financial structure of the economy evolve toward fragility during expansions.

Investment decisions are made under uncertainty that cannot be summarized as a probability distribution. Firms and households build up debt during good times because expectations of asset prices are self-confirming, and the resulting balance-sheet fragility is the endogenous source of the next crisis.

The big idea

Financialized economies with substantial private credit expansion. Fits crises with a leverage and asset-price component, of which 2007 to 2009 is the clearest case, less well those driven by supply or by monetary tightening.

Most macroeconomics treats the future as a lottery whose odds are unknown but real. This tradition denies the odds are there at all. Nobody ever assigned a probability to an invention not yet made, and pretending otherwise writes a false precision into every model built on top of it. From that starting point the emphasis shifts toward finance. If investment rests on confidence rather than calculation, a long stretch of good times makes borrowers and lenders steadily bolder, and the fragility that accumulates is manufactured by the calm rather than interrupted by the crisis.

Fundamental uncertainty

Keynes, in a 1937 Quarterly Journal of Economics reply to his critics, distinguished between risk that can be measured by a probability distribution and uncertainty that cannot, and argued that the interest rate, investment, and asset prices depend on the latter. The distinction is now called Knightian uncertainty after Frank Knight’s 1921 Risk, Uncertainty and Profit, and it puts a wedge between the rational expectations program and the post Keynesian program: if the distribution to be expected does not exist, requiring subjective probabilities to equal it is not a coherent condition.

Endogenous money

Nicholas Kaldor, in Lloyds Bank Review in July 1970, argued that in a modern banking system the money stock is determined by demand for credit rather than supplied exogenously by the central bank, so the causal arrow runs from bank lending to deposits to reserves. Michael McLeay, Amar Radia, and Ryland Thomas, in the Bank of England’s Quarterly Bulletin Q1 2014, adopted this account explicitly, describing loans as creating deposits and the deposit multiplier as backwards. The post Keynesian program had held this position for four decades before the operational description at a major central bank matched it.

Minsky and financial fragility

Hyman Minsky, in The Financial Instability Hypothesis (Levy Economics Institute Working Paper 74, 1992), distinguished three financing postures: hedge, in which cash flows cover interest and principal on debt; speculative, in which cash flows cover interest but not principal, so debt must be rolled; and Ponzi, in which cash flows cover neither, so the borrower depends on rising asset prices. During an expansion, agents shift from hedge toward speculative and Ponzi finance because the debts of the previous period have been successfully serviced, and the accumulated fragility becomes the endogenous source of the next crisis when asset prices fail to rise. Nothing in the mainstream model of 2007 had this structure. The run mechanics that turn fragility into failure, and the line between illiquidity and insolvency, are set out under financial fragility.

Effective demand without a labor market

The post Keynesian tradition retains Keynes’s proposition that output is demand-determined without importing the neoclassical labor market that new Keynesian models keep. Firms hire the labor needed to produce the demand they face; the wage does not adjust to clear a schedule of desired hours. This is closer to how firms describe their own behavior in the Bank of England Agents’ summaries and in the Atlanta Fed’s Business Inflation Expectations panel, and it is the reason the post Keynesian Phillips curve is written in cost-push rather than in expectations-augmented natural-rate form.

Where the tradition called it

Steve Keen, at the University of Western Sydney, and Wynne Godley, at the Levy Institute, warned in print through the early 2000s that United States and United Kingdom private-sector debt-to-income ratios had reached levels that a Minsky-hypothesis reading treated as unsustainable, and both predicted a recession triggered by a debt reversal. Godley’s Seven Unsustainable Processes (Levy Institute, 1999) laid out the sectoral-balance argument in detail. The mainstream forecast of 2007 did not include the crisis; the post Keynesian sectoral-balance forecast did. Whether that gives the tradition a general edge or one hit is disputed and depends on the base rate one credits.

Freshwater against saltwater

The literal geographic split in United States macroeconomics, between Great Lakes departments favoring market clearing and coastal departments favoring rigidity.

Take the label at face value: the market-clearing camp taught mainly at Chicago, Minnesota, Rochester, and Carnegie Mellon, and the sticky-price camp at MIT, Harvard, Yale, Berkeley, Princeton, and Stanford. The methodological gap ran through both the treatment of expectations and the assumption on price adjustment.

A department's paradigm is set by its senior faculty, its dissertation committees, and its seminar culture, and these persist for decades. The geographic clustering is a consequence of a small number of hires that then set the direction of each program's students and their placement.

The big idea

United States macroeconomics from about 1975 through the 2000s. The label survives past 2008 as shorthand for the methodological disagreement even after the substance has largely converged.

The labels are geographic and they mean precisely what they say. Departments near the Great Lakes leaned toward markets that clear and cycles needing no correction. Departments on the two coasts leaned toward prices that stick and policy that helps. Ideas in economics cluster by institution because senior faculty set the direction, dissertation committees enforce it, and graduates carry it to their first appointments, so a handful of hires can fix a program’s character for several decades.

Which departments the labels named

The Great Lakes schools that gave the freshwater label its content had a specific roster. Robert Lucas joined Chicago in 1974 after a decade at Carnegie Mellon. Thomas Sargent was at Minnesota from 1971 to 1987 and Neil Wallace until 1994; Edward Prescott was at Carnegie Mellon from 1971 to 1980 and at Minnesota from 1980 to 2003. Robert Barro and Robert King were at Rochester. The saltwater departments had Robert Solow, Paul Samuelson, Franco Modigliani, Stanley Fischer, and later Olivier Blanchard at MIT; Martin Feldstein and N. Gregory Mankiw at Harvard; James Tobin at Yale; George Akerlof and later David Romer and Christina Romer at Berkeley; Alan Blinder at Princeton; and John Taylor at Stanford.

Methodology beneath the geography

The methodological split beneath the geography had two joined pieces. The freshwater program insisted on optimizing agents with rational expectations and treated any deviation as an ad hoc addition; the saltwater program was willing to keep nominal rigidity as an assumption and to write down aggregate models with an old-Keynesian structure. The Lucas critique and Kydland-Prescott’s calibration methodology were freshwater innovations, and Fischer’s overlapping wage contracts (1977) and Taylor’s staggered contracts (1980) were the saltwater response that put stickiness on optimizing microfoundations.

What each side conceded after 2008

The synthesis that emerged in the 1990s new Keynesian program had already conceded rational expectations and microfoundations to freshwater and returned nominal rigidity to the workhorse model. After 2008 the concession ran the other way. Prominent freshwater voices, including John Cochrane and Casey Mulligan, both at Chicago, wrote extensively against fiscal stimulus in the recovery period; Paul Krugman, in his New York Times Magazine piece of September 6, 2009, gave the saltwater diagnosis a public form. The concession from the freshwater side was that a large financial disruption at the zero lower bound was a state the workhorse market-clearing model was not designed to describe. The concession from the saltwater side was that discretionary policy would not have been as effective as it was without the credibility that the freshwater program had established for the inflation-targeting mandate.

Convergence in current models

The DSGE models used at the Federal Reserve Board (SIGMA and EDO) and at the European Central Bank (the NAWM) share a common structural core: optimizing households, monopolistically competitive firms with Calvo pricing, a Taylor-type rule, financial frictions of the Bernanke-Gertler-Gilchrist type, and a zero lower bound treatment. FRB/US sits alongside them rather than inside that description: it is a large-scale econometric model that gets its nominal rigidity from polynomial adjustment costs and can be run with expectations formed either from a small VAR or from the model itself. The disagreements have moved to the calibration of the frictions, the specification of the financial sector, and the identification of shocks, not to whether the underlying paradigm should have forward-looking expectations or sticky prices. In that operational sense the geographic label describes a debate that ended.

Why the label persists

The label persists because the intellectual descendants of each camp still cluster where their advisors were, and because the methodological priors have not converged as far as the models have. A student trained at Chicago or Minnesota is more likely to build an equilibrium model with fully optimizing agents and to prefer identified structural shocks; a student trained at MIT or Harvard is more likely to reach for a reduced-form empirical strategy or to adopt behavioral departures from rational expectations. Neither is any longer the position on stickiness the label used to name.

Market clearing and the classical dichotomy

The assumption that prices adjust until quantity supplied equals quantity demanded in every market, which decides whether unemployment is voluntary and whether policy can help.

Postulate an equilibrium price vector at which all markets clear simultaneously, and let real allocations depend only on preferences, endowments, and technology. Nominal quantities are then determined separately, so the money stock scales the price level but does not change real outcomes.

If prices can adjust and every trader is a price taker, the only equilibrium is the one that clears markets, since a non-clearing price leaves gains from trade on the table. The dichotomy follows: money is a unit of account and does not enter the real allocation.

The big idea

Long-run analysis and comparative statics across steady states. Fails as a description of the short run wherever prices, wages, or credit terms are set by contract or by administrative decision and do not clear immediately.

Two linked assumptions divide this field more sharply than anything else in it. The first is that prices move until nothing is left unsold and nobody willing to trade is turned away. The second follows from the first: if that adjustment does happen, printing money raises every price in proportion and alters nothing real. Accept both and unemployment is a choice somebody made, leaving policy no useful work. Nearly every disagreement on this page reduces to a disagreement about how quickly the first assumption becomes true.

Walrasian equilibrium and the auctioneer

Leon Walras, in Elements of Pure Economics (1874), constructed the first general equilibrium model as a system of excess-demand equations, one per good, and postulated a tatonnement process in which a fictional auctioneer announces prices, hears no trade, and adjusts prices in the direction of excess demand until every market clears. Kenneth Arrow and Gerard Debreu, in Econometrica in 1954, proved that under convexity, continuity, and completeness conditions such an equilibrium exists. The proof is an existence theorem and gives no algorithm; without an auctioneer, actual markets have no obvious mechanism to compute the equilibrium prices, and the theory is silent on what happens when trade occurs at non-clearing prices.

Say’s law

Adam Smith, in An Inquiry into the Nature and Causes of the Wealth of Nations (1776), gave the analytical foundation on which classical political economy rested: division of labor and voluntary exchange coordinate production, and the invisible hand aggregates individual gains into a social one. Jean-Baptiste Say, in A Treatise on Political Economy (1803), added the proposition that supply creates its own demand, since the act of producing generates the income with which the product is bought, so general overproduction is impossible. Keynes, in The General Theory, took Say’s law as the target: with a demand for money, sellers of goods do not have to demand goods, so a general shortfall of aggregate demand can persist.

Classical dichotomy stated

The classical dichotomy separates real from nominal variables: real quantities, relative prices, and interest rates are determined in the goods and labor markets; the money stock determines the price level and no real variable. The dichotomy is exact in the long-run neoclassical model and empirically defensible over decades; the filtered money-growth and inflation series in Robert Lucas’s 1980 study of the United States give a slope close to one, and McCandless and Weber’s 1995 comparison of 110 countries over thirty years finds the two highly correlated. Over quarters and years the dichotomy does not hold, which is why the short-run macroeconomic model is not the long-run one.

Under the dichotomy, a doubling of the money stock doubles all prices and wages and leaves employment, output, and relative prices unchanged. Long-run neutrality of money is the version most macroeconomists accept; short-run neutrality is contested and separates the schools. Robert Lucas, in Journal of Economic Theory in 1972, built the first equilibrium model in which anticipated money is neutral even in the short run and only unanticipated money moves output.

What fails to clear

The empirical case for market clearing runs into a short list of concrete failures. The labor market shows persistent unemployment far above search frictions; the credit market rations by covenants and credit limits at posted rates rather than by an interest rate that equilibrates supply and demand; the housing market has a large stock of unsold inventory at any time and shows large trades at very different prices in the same neighborhood at the same time. These are the markets in which the assumption is empirically wrong at the short frequencies that matter for stabilization policy, and they are the ones that new Keynesian models introduce frictions to describe.

Why the assumption classifies the model

The single fastest way to place an unfamiliar macroeconomic model on the map is to ask whether it assumes markets clear. If yes, it is a real business cycle model or one of its extensions and treats fluctuations as efficient responses to shocks. If no, it is a Keynesian model, whether old or new, and treats stabilization policy as valuable. Almost every substantive disagreement in postwar macroeconomics reduces to this assumption and how it is defended, which is why an outline of the field can be organized around the schools that answered the question differently.