New Keynesian economics

• 1. Introduction

  ›

  New Keynesian economics began as an answer to a challenge. In the late 1970s and 1980s, the new classical school, associated with Robert Lucas, argued that Keynesian ideas fell apart once people were assumed to hold rational expectations. The Lucas critique sharpened that attack. The New Keynesians replied by rebuilding the Keynesian case from the ground up, giving it explicit microeconomic foundations. Their wager was that everyday frictions, such as prices and wages that do not move instantly, could explain why economies sometimes leave willing workers idle. The wager paid off. Today the school ranks among the dominant paradigms in macroeconomic theory. Why would a firm leave a price alone when conditions change? How can a market keep willing workers jobless? And what should a central bank do about it?

  268053

• 2. Staggered Contracts

  ›

  In 1977, Stanley Fischer published an article titled Long-Term Contracts, Rational Expectations, and the Optimal Money Supply Rule. It contained the first model of sticky information. Fischer imagined two unions that take turns setting wages. Whenever a union's turn arrives, it locks in pay for the next two periods.

  John B. Taylor offered a variation in two articles, one from 1979 called Staggered Wage Setting in a Macro Model and one from 1980 called Aggregate Dynamics and Staggered Contracts. In his version, the nominal wage stays constant for the whole life of a contract, which lasts two periods.

  Both designs carry the same quirk. Only the union choosing in the current period works from the latest information. Wages across the other half of the economy still reflect what was known earlier.

  Fixed nominal wages gave these early theorists a lever. If a wage is set in money terms, a central bank can change the money supply and so move the real wage, which is pay adjusted for inflation. Moving the real wage, in turn, affects the employment rate.

  Guillermo Calvo added a different mechanism in 1983 with a paper called Staggered Prices in a Utility-Maximizing Framework. He wrote it in continuous time, though economists now mostly use a discrete-time version. In it, each firm faces a fixed chance in any period, called the hazard rate, of being allowed to reset its price. The chance that the price stays put is the survival rate. Economists sometimes call the hazard rate the Calvo probability. The decisive difference from Taylor's design is that a Calvo price-setter cannot know how long a price will last, whereas Taylor contracts have a known length. Calvo's setup has become the most common way to model nominal rigidity in New Keynesian work. The Calvo probability would later sit at the center of the equation linking inflation to output in most modern models.

  678249

• 3. Menu Costs

  ›

  A lump-sum cost for changing a price, now called a menu cost, first appeared in a 1977 paper by Sheshinski and Weiss. They were studying how inflation affects the frequency of price changes. In 1985-86, several economists simultaneously proposed using the idea as a general theory of why nominal prices resist change.

  George Akerlof and Janet Yellen tied it to bounded rationality. Their argument was that a firm will not bother changing its price unless the gain exceeds some small amount. That inertia in nominal prices and wages lets output swing even while those prices and wages stay fixed.

  Gregory Mankiw used the menu-cost idea to examine the welfare effects when output changes because prices are sticky. Michael Parkin also put forward the idea. Olivier Blanchard and Nobuhiro Kiyotaki then stretched the approach to cover wages as well as prices, in their influential article Monopolistic Competition and the Effects of Aggregate Demand. Huw Dixon and Claus Hansen showed that menu costs confined to a small sector would still ripple outward. Prices in the rest of the economy would become less responsive to changes in demand.

  Skeptics argued that menu costs are too small to matter for a whole economy. In 1990, Laurence M. Ball and David Romer showed that real rigidities could interact with nominal ones to produce significant disequilibrium. A real rigidity exists whenever a firm is slow to adjust its real prices as conditions change. A firm with market power faces one, as does a firm whose input and wage costs are locked in by contract. Ball and Romer argued that rigidities in the labor market keep a firm's costs high, so firms hesitate to cut prices and lose revenue. Combined with the menu cost, that makes price cuts down to the market-clearing level less likely.

  Dixon and Mankiw also worked independently on a puzzle about government spending. Their simple general equilibrium models showed that the fiscal multiplier can rise with the degree of imperfect competition in the output market, even when prices are perfectly flexible. The mechanism runs through wages. Imperfect competition tends to lower the real wage, so households shift away from consumption toward leisure. When government spending rises, the matching lump-sum tax cuts both leisure and consumption. With a lower real wage, more of the cut falls on leisure, meaning households work more, and less falls on consumption. The multiplier stays below one, though it grows with imperfect competition. Government spending would return to the story in the 2010s, when economists asked how households of different wealth respond to it.

  1112131416

• 4. Market Failures

  ›

  A factory goes idle although willing workers stand ready and would gladly buy its output if they had wages. New Keynesians called this a coordination failure. The invisible hand, they argued, sometimes fails to arrange the usual flow of production and consumption.

  Russell Cooper and Andrew John gave the idea a general form in their 1988 paper Coordinating Coordination Failures in Keynesian Models. Their models have multiple equilibria, and agents might coordinate to improve, or at least not harm, their situations. They built on earlier work, including Peter Diamond's 1982 coconut model, a case of coordination failure involving search and matching theory. In Diamond's world, producers are more likely to produce when they see others producing, because more producers means more possible trading partners. This is a thick-market externality: markets work better as more people and firms join. Each agent's welfare depends on what the others decide.

  Self-fulfilling prophecies offer another route. Suppose a firm expects demand to fall and cuts back its hiring. Fewer vacancies worry workers, who then spend less. Demand falls just as the firm predicted, yet the firm's own caution caused the fall.

  In a Walrasian market, jobless workers bid wages down until the demand for labor matches supply. Under that logic the unemployed would be only people between jobs and people who refuse wages they consider too low. New Keynesians built several theories to explain how willing workers could remain jobless. The most important was efficiency wage theory. It also explained why a short-term rise in unemployment can become permanent and lead to higher unemployment in the long run.

  In efficiency wage models, firms pay for productivity rather than clearing the market. In developing countries, a firm may pay above the going rate so workers can afford enough food to work well. Higher pay may also build loyalty and morale. It may discourage shirking too, and shirking models proved especially influential.

  Carl Shapiro and Joseph Stiglitz published Equilibrium Unemployment as a Worker Discipline Device in 1984. Their model assumes employees avoid work unless firms can monitor effort and threaten slackers with unemployment. At full employment, a fired shirker simply finds another job. So each firm pays a premium above the market rate, to make workers prefer keeping their job to shirking. Because every firm does this, wages sit above the market-clearing level and the labor market as a whole does not clear. The resulting pool of jobless people raises the price of being fired, since a worker risks not just lower pay but being stuck in that pool. Some willing workers are left without jobs, yet the workers who have jobs are more efficient. By the early 1990s, economists were ready to stitch these strands onto a rival program built around real business cycles.

  2529402631

• 5. The Synthesis

  ›

  In the early 1990s, economists began to merge New Keynesian ideas with Real Business Cycle theory. RBC models were dynamic but assumed perfect competition. New Keynesian models were mostly static but built on imperfect competition. The new neoclassical synthesis took the dynamics of the first and the imperfect competition and nominal rigidities of the second. Tack Yun was among the first to do this, using the Calvo pricing model.

  Goodfriend and King listed four elements at the heart of the synthesis: intertemporal optimization, rational expectations, imperfect competition, and costly price adjustment. They also drew policy lessons from the consensus models. Monetary policy can affect real output in the short run, but there is no long-run trade-off. Money is not neutral in the short run, yet it is in the long run. Inflation has negative welfare effects, and central banks should keep their credibility through rules-based policy such as inflation targeting.

  John B. Taylor formulated his rule in 1993. It is a compact approximation of how a central bank's nominal interest rate responds to inflation, output and other conditions. For each one-percent rise in inflation, the bank tends to lift its rate by more than one percentage point, a feature often called the Taylor principle. The original version compares actual inflation with a target, and real GDP with potential GDP. The rate in question might be the federal funds rate in the US or the Bank of England base rate in the UK. Central banks do not explicitly consult such rules when setting rates. The rule serves as a concise descriptive proxy.

  Roberts derived the New Keynesian Phillips curve in 1995, and most state-of-the-art New Keynesian DSGE models have used it since. DSGE stands for dynamic stochastic general equilibrium. The curve says that this period's inflation depends on current output and on expectations of next period's inflation. It comes from the dynamic Calvo model of pricing. The less rigid nominal prices are, the greater the effect of output on current inflation.

  Richard Clarida, Jordi Gali and Mark Gertler surveyed these ideas in the Journal of Economic Literature, culminating in a three-equation model. It joined the Phillips curve and the Taylor rule with a dynamic IS curve, which comes from how households optimally plan consumption over time. The model was simple enough to analyze policy issues in theory. But it was oversimplified, with no capital or investment, and it does not perform well empirically. Those shortcomings pushed the next decade's builders toward far larger models.

  4544474851

• 6. Bigger Models

  ›

  In 2000, Christopher Erceg, Dale Henderson and Andrew Levin took the Blanchard and Kiyotaki model of unionized labor markets and paired it with Calvo pricing. They placed the result inside a New Keynesian DSGE model, adding imperfectly competitive labor to the sticky prices of the 1990s.

  To fit real data and run policy simulations, researchers then built quite complicated models. Seminal papers came from Frank Smets and Rafael Wouters, and from Lawrence J. Christiano, Martin Eichenbaum and Charles Evans. Their models shared several features. Habit persistence meant that the value of extra consumption depends on past consumption. Calvo stickiness applied to both wages and prices, with indexation so that wages and prices not explicitly reset are updated for inflation. Capital carried adjustment costs and variable use. New shocks included demand shocks, which affect the marginal utility of consumption, and markup shocks, which move the desired markup of price over marginal cost. A Taylor rule represented monetary policy, and Bayesian estimation methods fitted the models to data.

  Gregory Mankiw and Ricardo Reis later developed Fischer's sticky information idea. In their version, each period a fixed share of workers or firms can replan their wages or prices. Using quarterly data, they assumed 25 percent, so each quarter a quarter of randomly chosen firms or unions plans current and future prices on the latest information. The rest rely on information from when they last replanned. Mankiw and Reis found that the model explained inflation persistence well.

  The sticky information model has no nominal rigidity at all. It is information that is sticky, not prices. A firm that gets to replan will set a different price for every period in its plan. That conflicts with the evidence. Studies of the United States, the Eurozone, the United Kingdom and other places show that some sectors change prices often while others leave them fixed for long stretches. The mismatch has inspired attempts at a dual stickiness model that combines sticky information with sticky prices.

  5253555659

• 7. Households Diverge

  ›

  The 2010s brought models in which households face different circumstances. They are called HANK models, short for Heterogeneous Agent New Keynesian. Besides sticky prices, a typical HANK model includes uninsurable labor income risk that varies by household. That risk produces a wide spread of wealth.

  Early models with both features came from Oh and Reis in 2012, McKay and Reis in 2016, and Guerrieri and Lorenzoni in 2017. Greg Kaplan, Benjamin Moll and Gianluca Violante coined the name HANK in a 2018 paper. Their model also had households accumulate two kinds of assets, one liquid and one illiquid. It reproduces the evidence that many households hold little liquid wealth, the so-called hand-to-mouth households. About two-thirds of them still hold non-trivial amounts of illiquid wealth. Kaplan and Violante introduced the term wealthy hand-to-mouth in a 2014 study of fiscal stimulus policies.

  These households change how monetary policy works. Their spending reacts strongly to changes in disposable income, not to the interest rate, which is the price of future consumption relative to current consumption. So monetary policy travels mostly through general equilibrium effects on household labor income. In representative agent models, called RANK, the main channel is instead intertemporal substitution.

  There are two main implications. First, monetary policy interacts strongly with fiscal policy, because Ricardian equivalence fails when hand-to-mouth households are present. A change in the interest rate shifts the government's budget constraint, and the fiscal response to that shift affects household disposable income. Second, aggregate monetary shocks are not distributionally neutral. They move the return on capital, which affects households with different levels of wealth and assets differently. How a central bank should respond is the question the school's policy advice was built to answer.

  6663646568

• 8. Policy and Rivals

  ›

  Suddenly expanding the money supply just to produce a temporary boom is something New Keynesian economists do not recommend. Such a move would raise inflationary expectations, and removing them would be impossible without causing a recession. They advocate monetary policy for stabilization instead. When an unexpected shock hits, such as a fall in consumer confidence that lowers both output and inflation, expanding the money supply helps. Lower interest rates raise output while steadying inflation and inflationary expectations.

  Studies of optimal monetary policy in New Keynesian DSGE models have focused on interest rate rules, especially Taylor rules. Optimal rules usually react to the output gap rather than to output itself. In some simple models, stabilizing inflation is enough, because perfectly stable inflation also stabilizes output and employment as far as is desirable. Blanchard and Galí called this property the divine coincidence.

  They also showed that the coincidence breaks when a model has more than one market imperfection. With frictions in adjusting employment as well as sticky prices, stabilizing inflation and stabilizing employment become a trade-off. Alves showed in 2014 that the coincidence fails even in the non-linear standard model. It holds only if the central bank targets inflation of exactly 0%. At any other target, an endogenous trade-off appears. Some macroeconomists believe these models are close to giving useful quarter-to-quarter policy advice, but disagreement exists.

  After World War II, Paul Samuelson used the term neoclassical synthesis for the merger of Keynesian and neoclassical economics. The hope was that government and central banks would hold rough full employment, and John Hicks' IS/LM model was central to it. Later work by James Tobin and Franco Modigliani, stressing the microfoundations of consumption and investment, was sometimes called neo-Keynesianism. Paul Davidson's post-Keynesianism contrasts with it by emphasizing fundamental uncertainty, especially around private fixed investment.

  The new classicals combined a single market-clearing equilibrium at full employment with rational expectations. New Keynesians used microfoundations to show that price stickiness stops markets from clearing, so a rational-expectations equilibrium need not be unique. They saw full employment as automatic only in the long run, which is why policy is needed: the long run may be very long. The two camps merged in the new neoclassical synthesis of the 1990s, which underpins mainstream economics today. The 2008 global financial and economic crisis then renewed the Keynesian emphasis on coordinated macroeconomic policy, on institutions such as the World Bank and the IMF, and on a controlled trading system. That emphasis shows up in the work of IMF economists and of Donald Markwell.

  7174757778