New Keynesian economics is a school of macroeconomics that explains economic fluctuations through the interaction of forward-looking decisions, imperfect competition, and frictions that prevent prices or wages from adjusting immediately. Its central proposition is that changes in spending can affect production and employment because not all nominal prices adjust at once. Consequently, monetary policy can influence real economic activity in the short run, even when households and firms act purposefully and understand the policy regime.
The approach emerged during the late 1970s and 1980s as economists sought to reconstruct Keynesian economics using explicit models of individual behavior and expectations. It subsequently became a major foundation of the models used by central banks to analyze inflation, business fluctuations, and stabilization policy. Its scope extends from theories of price adjustment and wage determination to large dynamic stochastic general equilibrium models, including models with financial intermediaries and heterogeneous households.
New Keynesian economics is not a single model or a uniform set of policy prescriptions. Early contributions emphasized the microeconomic foundations of economic rigidities; later work organized those mechanisms into quantitative models. Its conclusions depend on which imperfections are included, how expectations are formed, and whether the economy is experiencing ordinary fluctuations, financial disruption, or constraints on interest-rate policy.
Intellectual origins and historical development
The intellectual starting point is the problem associated with John Maynard Keynes: why can an economy experience sustained shortfalls in spending and employment rather than rapidly returning to full utilization of resources? Postwar Keynesian analysis often represented this problem through aggregate relationships, including the IS–LM model and the Phillips curve. These frameworks were influential but did not always derive their behavioral equations from explicit intertemporal choices by households and firms.
During the 1960s and 1970s, Milton Friedman and Edmund Phelps challenged the idea of a permanently exploitable trade-off between inflation and unemployment. New classical macroeconomics subsequently placed expectations and market-clearing behavior at the center of macroeconomic analysis. Robert Lucas’s 1976 formulation of the Lucas critique argued that estimated relationships could change when policy rules changed, because people would alter their decisions. A policy evaluation based on historically stable correlations was therefore not necessarily reliable under a new regime.
New Keynesian economists largely accepted the importance of explicit behavioral foundations and rational expectations, but rejected the conclusion that continuously clearing markets provided an adequate account of short-run fluctuations. Stanley Fischer’s and John Taylor’s work on staggered nominal contracts showed how anticipated monetary disturbances could have real effects when contracts were set at different dates. Guillermo Calvo’s 1983 pricing model supplied a tractable way to represent staggered price adjustment. George Akerlof and Janet Yellen, N. Gregory Mankiw, Olivier Blanchard, Nobuhiro Kiyotaki, and others investigated how small adjustment costs, imperfect competition, and related imperfections could generate substantial aggregate effects.
The label initially covered a broad research program rather than a standardized quantitative framework. Mankiw and David Romer’s two-volume New Keynesian Economics, published in 1991, collected work on nominal rigidities, market structure, coordination failures, and labor markets. During the 1990s, researchers increasingly combined these mechanisms with the intertemporal optimization and stochastic modeling techniques developed in real business cycle theory.
Marvin Goodfriend and Robert King described this integration as the new neoclassical synthesis in 1997. Richard Clarida, Jordi Galí, and Mark Gertler’s 1999 treatment of monetary policy, and Michael Woodford’s 2003 book Interest and Prices, helped establish the modern framework. This synthesis retained an important distinction: real factors govern productive capacity, but nominal adjustment frictions allow demand and policy to influence how fully that capacity is used.
Price setting and the foundations of nominal rigidity
The defining friction in the standard New Keynesian model is nominal rigidity: some prices or wages cannot, or do not, adjust immediately when economic conditions change. This need not mean that firms ignore profit opportunities. Repricing can require managerial attention, customer communication, changes to contracts, or revisions to information systems. Firms may also hesitate to change prices before competitors do, especially when customers compare prices or regard frequent increases unfavorably.
Monopolistic competition supplies a common market structure. Each firm produces a differentiated good and faces a downward-sloping demand curve. Unlike a perfectly competitive firm, it chooses its own price, ordinarily applying a markup over marginal cost. If that price is temporarily fixed, an increase in demand can make it profitable to increase production rather than ration customers. This provides a direct connection between spending and output that is absent from a frictionless market-clearing model.
Two important pricing mechanisms differ in the timing of adjustment. Under Calvo pricing, each firm has a constant probability of being permitted to reset its price in each period. If the probability of not adjusting is , the expected duration of a price spell is periods. A firm that resets today considers not only current costs but also expected conditions over the time its price may remain unchanged. This forward-looking decision produces a tractable aggregate inflation equation. The random adjustment opportunity is a modeling device, not a literal description of every firm’s pricing procedure.
In menu-cost models, a firm pays a cost when changing its price and therefore chooses when adjustment is worthwhile. This makes adjustment state-dependent: large shocks can prompt many firms to reprice simultaneously. Julio Rotemberg’s 1982 formulation offers another approach, assigning a smooth cost to changing prices. Calvo and Rotemberg models can deliver closely related linearized aggregate equations near a stable, zero-inflation equilibrium, but their nonlinear implications and interpretations need not coincide.
A small private cost of repricing can sometimes have a large aggregate effect because a firm does not internalize the effect of its price decision on economy-wide demand. This is a form of externality, although the precise amplification depends on market structure and other assumptions. Real rigidities—features that make desired relative prices or wages respond weakly to activity—can reinforce nominal stickiness. Examples include variable desired markups, firm-specific inputs, and wage-setting arrangements. Nominal rigidity concerns adjustment in monetary units; real rigidity concerns adjustment in relative prices or real incentives. The two are analytically distinct.
The canonical model and its equilibrium equations
The simplest modern New Keynesian model contains optimizing households, differentiated-goods producers, nominal price rigidity, and a monetary authority. Households choose consumption, saving, and labor supply over time. Firms hire inputs and set prices subject to adjustment frictions. These choices must be mutually consistent in general equilibrium. The benchmark typically abstracts from capital accumulation, financial distress, household differences, and government spending to isolate the monetary transmission mechanism.
Around a steady state, household consumption choices and goods-market equilibrium yield a forward-looking demand equation:
Here is the output gap, measured relative to flexible-price, or natural, output; is the nominal short-term interest rate; is next period’s inflation; and denotes expectations conditional on current information. The parameter governs intertemporal substitution. The natural real interest rate, , is the real rate consistent with the flexible-price allocation. A real interest rate above that benchmark encourages households to defer expenditure and lowers current demand.
Optimal staggered price setting produces the New Keynesian Phillips curve:
The parameter is the household discount factor, captures the responsiveness of inflation to activity, and represents a cost-push disturbance. More fundamentally, inflation depends on expected inflation and real marginal cost. Expressing marginal cost as an output gap requires assumptions about technology, preferences, and labor-market structure. Under a common Calvo specification with linear production and flexible wages,
where is the inverse elasticity of labor supply. Other production and wage-setting arrangements alter this expression.
A policy rule closes the model, often through an interest-rate reaction function:
The constants and denote the steady-state real rate and inflation target. The coefficients describe policy responses, while is a monetary policy disturbance. Actual estimated models often include interest-rate smoothing and additional variables.
These equations are a local approximation, not universal behavioral laws. They show why expectations matter: present spending depends on expected future activity and real rates, while present inflation depends on expected future inflation. They also distinguish natural output from socially efficient output. If monopoly markups or other distortions remain under flexible prices, eliminating the output gap does not necessarily eliminate the welfare-relevant inefficiency.
Monetary stabilization, commitment, and welfare
New Keynesian analysis explains the monetary transmission mechanism through real interest rates and expectations. When a central bank raises its policy rate and prices adjust incompletely, the expected real interest rate can rise. Households then postpone some consumption, firms face weaker demand, and inflation pressure declines. The full response depends on the expected path of policy rather than only the current policy-rate change.
An important implication is the Taylor principle, associated with John Taylor’s 1993 policy-rule proposal. In the simplest setting, the nominal rate must respond more than one-for-one to a persistent increase in inflation so that the real rate also rises. A passive response can permit self-fulfilling movements in inflation and activity. Nevertheless, the precise conditions for a unique equilibrium depend on the complete policy rule, the fiscal regime, and the model’s structure. The coefficient restriction is not a universal theorem applicable independently of those features.
Policy objectives can be derived from household welfare rather than imposed as arbitrary preferences for stability. Near an appropriately defined steady state, the welfare loss in a basic model often takes the approximate form
Inflation is costly partly because staggered price changes create dispersion in relative prices. Otherwise similar firms charge different prices, causing inefficient allocations of production and labor. Output fluctuations matter insofar as they move activity away from its efficient level. The weight is related to structural parameters; with additional distortions, the relevant target and loss function become more complicated.
In a restrictive benchmark without cost-push disturbances and with the efficient and natural allocations aligned, stabilizing inflation also stabilizes the welfare-relevant output gap. Blanchard and Galí termed this coincidence the “divine coincidence.” It generally fails when wage markups vary, real wages are rigid, or other distortions create conflicting stabilization objectives. A supply disturbance may then force policymakers to accept either more inflation or a larger decline in activity.
The distinction between commitment and discretion is equally important. A credible promise concerning future policy can affect today’s expectations and improve current outcomes. A policymaker acting period by period may later prefer to abandon that promise. Commitment solutions therefore often display history dependence: appropriate policy depends partly on past disturbances and previous promises. These results provide analytical foundations for discussions of inflation targeting, policy credibility, and central bank independence, without implying that institutional arrangements alone guarantee an optimal outcome.
Fiscal policy and the effective lower bound
The effects of fiscal policy depend on how public spending and taxation interact with nominal rigidity, household behavior, and monetary policy. In a representative-household model, an increase in government purchases uses resources that might otherwise support private consumption or investment. Financing through present or future taxes can reduce household wealth and increase labor supply. At the same time, when demand is deficient and prices are sticky, additional public spending can increase production rather than merely displace private expenditure.
The resulting fiscal multiplier is not a fixed structural constant. It varies with the persistence and composition of spending, how taxes respond, the degree of economic openness, and the response of the central bank. If monetary policy raises interest rates aggressively after a fiscal expansion, private demand may be crowded out. If policy accommodates the expansion, the effect on output can be larger. Comparisons must also distinguish an impact multiplier from a cumulative multiplier measured over several years.
The effective lower bound on nominal interest rates changes this interaction. Currency and other institutional features limit how negative policy rates can become, although the practical bound need not equal zero. If the natural real rate becomes sufficiently negative, the central bank may be unable to reduce its nominal rate enough. Expected deflation then raises the real rate, reducing demand and potentially reinforcing the downturn. This mechanism formalizes an important version of the liquidity trap.
At a binding lower bound, government spending can raise output and inflation expectations. Higher expected inflation lowers the real interest rate when the nominal rate is unchanged, potentially supporting private expenditure as well. In some New Keynesian models this produces spending multipliers above one. That result is conditional: a short-lived constraint, changes in future taxes, expectations of earlier monetary tightening, or alternative household behavior can substantially change the magnitude. The timing of spending relative to the constrained period is particularly important.
Forward guidance operates through promises about future interest rates. Standard models can predict implausibly large effects from policy changes announced far in advance, a result known as the forward-guidance puzzle. Limited planning horizons, imperfect credibility, incomplete information, and household heterogeneity can moderate it. Quantitative easing requires additional mechanisms, such as imperfect asset substitutability or constrained intermediaries, to have effects beyond the signals it provides about future rates. In a frictionless benchmark, exchanging assets with equivalent risk-adjusted payoffs need not change private spending.
Labor markets, wages, and real rigidities
Labor-market analysis was central to the early New Keynesian program, although the simplest modern model represents employment through a competitive labor market and household labor supply. That simplification can account for changes in hours but does not by itself explain involuntary unemployment, job search, vacancies, or persistent differences between workers. More elaborate models therefore distinguish the wage-setting process from the process through which workers and jobs are matched.
Efficiency-wage theory explains why firms might voluntarily pay wages above the level that would clear the labor market. Higher pay can reduce turnover, improve recruitment, or discourage shirking when effort is difficult to observe. Carl Shapiro and Joseph Stiglitz’s 1984 model showed how unemployment can help sustain effort incentives: losing a job is costly when finding another is difficult. The mechanism rests on information asymmetry and monitoring limitations rather than on nominal wage contracts. It can therefore generate real wage rigidity even when monetary prices are flexible.
Nominal wage stickiness provides a different mechanism. Workers or unions may reset wages only intermittently, while firms adjust employment in response to demand. Christopher Erceg, Dale Henderson, and Andrew Levin’s 2000 framework combined staggered wage setting with staggered price setting. In such models, stabilizing goods-price inflation alone generally fails to stabilize wage dispersion and employment efficiently. Monetary policy faces several distortions, and the desired trade-off depends on how readily wages and prices adjust.
Search-and-matching models give explicit roles to unemployment and vacancies. Firms incur recruiting costs; unemployed workers search; successful matches generate a surplus that is divided through wage determination. Hiring depends on expected future profits, so a demand disturbance can reduce vacancies before it fully reduces employment. Incorporating these mechanisms into New Keynesian models connects monetary disturbances to job creation, job destruction, and measured labor-market tightness. Wage adjustment is crucial because it determines how much a shock changes the expected surplus from hiring.
Labor-market institutions can alter these responses. Collective bargaining, unemployment benefits, contract duration, and employment arrangements influence outside options, wage adjustment, and recruitment incentives. Their effects cannot be inferred from nominal rigidity alone. A mechanism that increases wage insurance may also change employment responses, while a wage that appears rigid in aggregate data may partly reflect changes in the composition of employed workers.
These distinctions matter for interpreting an output gap. A decline in production may reflect fewer hours per worker, fewer employed workers, lower participation, or reduced productivity. Models that merge these margins can generate similar aggregate output movements while implying different welfare costs and different responses to policy.
Financial intermediation and open-economy extensions
The basic New Keynesian model often represents saving and borrowing through a single risk-free bond. This abstraction leaves no substantive role for banks, borrowing spreads, defaults, or intermediary capital. Models with financial frictions instead distinguish the policy interest rate from the financing conditions faced by households and firms. An unchanged policy rate can then coexist with a severe contraction in credit and expenditure.
The financial accelerator developed by Ben Bernanke, Mark Gertler, and Simon Gilchrist links borrowing costs to borrowers’ financial position. When borrower net worth falls, conflicts between borrowers and lenders become more consequential, and the external finance premium rises. Weaker investment reduces activity and can further damage borrower balance sheets. Collateral constraints create a related amplification mechanism: a decline in asset values reduces borrowing capacity, which can force spending cuts or asset sales.
Other models place the constraint within financial intermediation itself. Losses reduce intermediary capital, restrict balance-sheet expansion, and increase lending spreads. Such mechanisms became particularly prominent after the global financial crisis of 2007–2009. They help explain why changes in credit conditions can have effects that a model with only a policy rate misses. They also create a framework for studying lender-of-last-resort operations, asset purchases, and macroprudential policy, although the effectiveness of each instrument depends on the specific friction modeled.
Open-economy New Keynesian models introduce foreign demand, imported goods, international asset markets, and the exchange rate. A currency depreciation can increase export competitiveness while making imported consumption goods and production inputs more expensive. The consequences depend on invoicing currency and the degree of exchange-rate pass-through. A firm that fixes its export price in the buyer’s currency produces different short-run expenditure responses from one that fixes its price in its own currency.
International financial arrangements matter as well. Complete cross-country insurance, incomplete asset markets, foreign-currency debt, and borrowing constraints generate different relationships between exchange rates, consumption, and external balances. Guillermo Calvo’s and Maurice Obstfeld and Kenneth Rogoff’s contributions supplied important building blocks for nominally rigid open-economy analysis, while subsequent work developed policy models for small open economies and monetary unions.
Within a currency union, member economies share a nominal policy rate but can experience different natural rates and inflation rates. Their real interest rates can consequently diverge in a destabilizing direction. Country-specific fiscal conditions, wage adjustment, and cross-border risk sharing become particularly important. These extensions show why conclusions drawn from a closed economy with a single representative borrower need not carry over unchanged to an internationally integrated economy.
Empirical evidence and quantitative implementation
Empirical evaluation takes place at several levels: individual prices and wages, aggregate inflation relationships, responses to identified policy shocks, and the performance of complete structural models. These forms of evidence answer different questions. Observing infrequent price changes supports the existence of nominal rigidity, but does not by itself identify how much a monetary disturbance changes output.
Microeconomic price data show substantial variation across products and sectors. Mark Bils and Peter Klenow’s 2004 study of United States consumer prices found more frequent price adjustment than some earlier macroeconomic calibrations had assumed. Emi Nakamura and Jón Steinsson’s 2008 research demonstrated the importance of distinguishing regular-price changes from temporary sales and product substitutions. A temporary discount may have different implications for aggregate adjustment from a persistent revision in the regular price. Neither a single average price duration nor a single adjustment mechanism captures all sectors well.
Estimating the New Keynesian Phillips curve creates additional difficulties. Expected future inflation is not directly observed, while marginal cost and the natural output benchmark must be measured or inferred. Research has used surveys, forecasting models, labor-income shares, and statistical output gaps, each with limitations. Galí and Gertler’s 1999 study used a hybrid specification combining forward-looking price setting with a backward-looking component. Such models can fit inflation persistence, but the interpretation depends on whether persistence comes from indexation, expectations, shocks, or measurement error.
Medium-scale structural models add features needed to match aggregate dynamics. These commonly include habit persistence in consumption, investment adjustment costs, variable capital utilization, wage stickiness, and persistent disturbances. Lawrence Christiano, Martin Eichenbaum, and Charles Evans’s 2005 model examined the responses to a monetary policy shock. Frank Smets and Rafael Wouters’s 2003 euro-area model and 2007 United States model helped establish Bayesian estimation of New Keynesian systems for policy analysis.
Bayesian estimation combines the likelihood of observed data with prior distributions over parameters. It permits joint estimation of structural features and unobserved shocks, but a close fit does not guarantee that every mechanism is correctly identified. Different combinations of rigidity, shock persistence, and expectation formation can generate similar aggregate movements. Identification therefore also benefits from evidence outside the model’s principal estimation dataset.
Central-bank models are consequently used alongside forecasts, sectoral information, and alternative empirical methods. Structural models make assumptions explicit and support counterfactual exercises, while reduced-form evidence can test particular predictions without accepting the entire system. Evaluation also distinguishes conditional policy scenarios from unconditional forecasts: success in one task does not automatically establish success in the other.
Household heterogeneity and alternative expectations
The representative-household assumption suppresses differences in wealth, income risk, borrowing constraints, and spending behavior. Yet a household with substantial liquid assets can respond to an income loss differently from one that must immediately reduce consumption. These differences are central to heterogeneous-agent New Keynesian models, commonly abbreviated HANK models. The framework combines incomplete household insurance with the nominal rigidities of New Keynesian macroeconomics.
Greg Kaplan, Benjamin Moll, and Giovanni Violante’s 2018 analysis of monetary policy provided an influential quantitative formulation. Their model distinguishes liquid assets, which can readily finance spending, from illiquid assets, which are costly to adjust. A household may hold substantial housing or retirement wealth while possessing little readily spendable wealth. Consequently, measured net worth alone does not determine sensitivity to disposable income.
Heterogeneity changes the transmission of policy. A rate reduction affects intertemporal substitution, but it also changes employment income, business profits, interest receipts, debt payments, and asset valuations. The indirect effect through labor income can be more important than the direct incentive to shift consumption across dates. The aggregate response depends on which households receive additional income and their marginal propensity to consume. Redistribution is therefore part of the transmission mechanism, not merely a separate consequence of stabilization.
Expectations are another active area of development. Rational expectations require consistency between beliefs and the model’s probability structure; they do not imply certainty about future shocks. Nevertheless, conventional implementations often assume considerable knowledge of the economy and common understanding of policy. Imperfect-information models instead allow households or firms to observe noisy signals, update infrequently, or disagree about the state of the economy.
N. Gregory Mankiw and Ricardo Reis’s 2002 sticky-information model emphasized infrequent updating of information rather than infrequent opportunities to change prices. Other approaches introduce learning, limited attention, finite planning horizons, or forms of bounded rationality. These modifications can change inflation persistence and weaken the effects of distant policy announcements. Their empirical evaluation draws on household and firm surveys, expectation revisions, and responses to information treatments.
Heterogeneity also complicates welfare analysis. A policy that improves an aggregate stabilization measure can affect borrowers, savers, workers, and asset holders differently. The evaluation then requires explicit assumptions about social welfare weights and insurance arrangements. HANK models retain the characteristic New Keynesian mechanism of nominal rigidity, but expand the set of distributional and balance-sheet channels through which aggregate policy operates.
Criticisms, boundaries, and unresolved questions
A major criticism concerns the empirical interpretation of nominal rigidity. Calvo pricing is analytically convenient, but it assumes adjustment opportunities that are independent of a firm’s economic circumstances. State-dependent pricing models allow large shocks to trigger rapid adjustment, potentially changing monetary non-neutrality. Strategic complementarities, production networks, and heterogeneity in adjustment costs can restore stronger aggregate persistence, but the resulting predictions must be evaluated against detailed price data rather than imposed solely to reproduce aggregate behavior.
A second issue is the treatment of disturbances. Estimated models often require several persistent shocks to preferences, markups, investment efficiency, technology, and policy. These can summarize omitted influences and improve fit, yet their structural interpretation is not always unique. A “markup shock,” for example, may absorb changes in competition, taxes, production costs, or deficiencies elsewhere in the model. A decomposition of fluctuations into named shocks is informative only to the extent that those shocks are identified by credible restrictions or independent evidence.
The global financial crisis exposed limitations in models with frictionless finance, but it did not constitute a single test of every New Keynesian mechanism. Models available before the crisis already incorporated credit frictions, while many operational forecasting models gave them a limited role. The subsequent challenge has been to integrate financial distress, default, occasionally binding constraints, and nonlinear dynamics without making the model unmanageable or weakly identified. Linear approximations around a stable steady state are particularly ill-suited to some crisis mechanisms.
Inflation following the COVID-19 disruption renewed attention to sectoral bottlenecks, fiscal transfers, changing demand composition, and the interaction of supply constraints with aggregate spending. A single aggregate Phillips curve cannot fully describe changes in relative prices across energy, goods, housing, and services. Multi-sector models study how input linkages and uneven capacity constraints transmit those disturbances. They also distinguish a one-time relative-price adjustment from an ongoing process of generalized inflation, while allowing the two to interact through wages and expectations.
Other disagreements concern the breadth of the framework. Post-Keynesian economists criticize its frequent reliance on equilibrium optimization and its treatment of money and fundamental uncertainty. New classical and real business cycle approaches place greater emphasis on real disturbances and question some assumed adjustment frictions. Within New Keynesian research, disagreement persists over expectation formation, the measurement of slack, trend inflation, and how monetary and fiscal regimes jointly determine the price level.
These disputes attach to particular mechanisms and empirical claims rather than to a single universally agreed model. Current research tests how price adjustment changes across inflation regimes, how household distributions alter transmission, and how financial and production networks propagate shocks. Each extension changes which observations can discriminate among competing explanations of fluctuations.