Emi Nakamura, Clark Medalist 2019

American Economic Association Honors and Awards Committee
April 2019

Emi Nakamura is an empirical macroeconomist who has greatly increased our understanding of price-setting by firms and the effects of monetary and fiscal policies. Nakamura’s distinctive approach is notable for its creativity in suggesting new sources of data to address long-standing questions in macroeconomics. The datasets she uses are more disaggregated, or higher-frequency, or extending over a longer historical period, than the postwar, quarterly, aggregate time series that have been the basis for most prior work on these topics in empirical macroeconomics. Her work has required painstaking analysis of data sources not previously exploited, and at the same time displays a sophisticated understanding of the alternative theoretical models that the data can be used to distinguish.

Nakamura is best known for her use of microeconomic data on individual product prices to draw conclusions about the empirical validity of models of price-setting used in the macroeconomic literature; this has been a critical issue for analysis of the short-run effects of monetary policy. Studies of the adjustment of individual prices—in particular, measures of the average time that prices are observed to remain unchanged—have long been a key source of evidence regarding the importance of price rigidity. However, until very recently most evidence of this kind came from studies of a very small number of markets, so that the question of how typical these specific prices were remained an important limitation. The availability of new data sets that allow changes in the prices of a very large number of goods to be tracked simultaneously has radically transformed this literature over the past fifteen years, and Nakamura, together with her frequent co-author Jón Steinsson, has played a leading role in this development.

In their most-cited paper, “Five Facts About Prices: A Re-Evaluation of Menu Cost Models” (QJE 2008), Nakamura and Steinsson study the BLS data on individual prices used to construct the published consumer and producer price indices for the U.S. economy, documenting a variety of facts about changes in individual prices that can then be compared to the implications of a popular theoretical model of price adjustment, the “menu cost” model. They give particular attention to the average frequency of price changes, an important issue in the numerical calibration of quantitative models of the effects of monetary policy. While past studies using other sources had concluded that the median time between price changes in the U.S. economy was nearly a year, the first work using the BLS microdata (by Mark Bils and Pete Klenow) had argued that the BLS data underlying the CPI showed that prices actually changed much more frequently (a median duration of prices only a little over 4 months). Bils and Klenow actually did not use the BLS micro dataset, but rather an extract from it for the period between 1995 and 1997 that reported average frequencies of price changes at a very disaggregated level. Nakamura and Steinsson instead obtained access to the actual micro data used by the BLS, which has all the price observations collected by the BLS and for the period from 1988 to 2005.

Revisiting Bils and Klenow’s conclusion using their superior dataset, Nakamura and Steinsson show that one’s conclusions about the frequency of price changes in the CPI data depend on the method used to distinguish sales from changes in “regular prices.” They also study the changes in individual wholesale prices and consumer prices. They find both that changes in regular prices occur much less often than price changes that include sales (they find a median duration of 8-11 months for regular prices, depending on the precise method used to classify price changes), and that producer prices (for which there is less of a need to filter out “sales”) also change quite infrequently. A first reason why this paper is so influential is that it gives very convincing microeconomic evidence for much more substantial “stickiness” of individual prices than the surprising results of Bils and Klenow had implied. The paper is also valuable for documenting features of the data on individual price changes that can be used to test the realism of specific models of price adjustment. Nakamura and Steinsson stress two features of the data that are contrary to the predictions of a popular class of models of price adjustment (“menu-cost” or “S-s” models): clear seasonality in the frequency of price adjustments, and the failure of the hazard function for price changes to increase in the time since the last change in price.

The ability of a “menu-cost” model to account for the quantitative characteristics of the micro data on price changes is considered further in “Monetary Non-Neutrality in a Multi-Sector Menu Cost Model” (QJE 2010, also with Steinsson). Prior numerical analyses of the implications of menu-cost models (such as the influential paper by Golosov and Lucas) had assumed that all goods in the economy were subject to menu costs of the same size (in addition to being produced with the same technology), with the parameters common to all goods being assigned numerical values to match statistics for the set of all price changes (such as the overall frequency of change in prices and the average absolute size of price changes). But one of the facts documented by Nakamura and Steinsson in “Five Facts” is that there is tremendous heterogeneity across sectors of the U.S. economy in the frequency of non-sale price changes.

In the “Monetary Non-Neutrality” paper, they calibrate a multi-sector menu-cost model to also match the distribution across sectors of both the frequency of price changes and the average size of price changes. They find that the real effects of a monetary disturbance are three times as large in their multi-sector model as in a one-sector model (like that of Golosov and Lucas) calibrated to the mean frequency of price change of all firms. Indeed, whereas Golosov and Lucas argue that price rigidity is not an empirically plausible explanation for the observed effects of monetary disturbances, if one takes account of the micro evidence on the frequency of price adjustments, Nakamura and Steinsson show that their calibrated multi-sector model (with nominal shocks of the magnitude observed for the U.S. economy) predicts output fluctuations that would account for nearly a quarter of the U.S. business cycle. This would be roughly in line with the fraction of GDP variability that is attributed to monetary disturbances in atheoretical vector-autoregression studies. The paper’s emphasis on the importance of taking account of sectoral heterogeneity when parameterizing the degree of price stickiness has been highly influential.

More recently, Nakamura and Steinsson have devoted considerable effort to extending the BLS micro-level data set on consumer prices back to 1977. This labor-intensive, multiyear data-construction project is of interest because the extended database now includes the period in the late 1970s and early 1980s when inflation was much higher and more volatile than it has been since 1988. The first paper making use of the extended data set is with Patrick Sun and Daniel Villar, “The Elusive Costs of Inflation: Price Dispersion During the U.S. Great Inflation” (QJE, forthcoming). The paper considers how the firms’ adjustment of their prices to changing market conditions differs in a higher-inflation environment. The authors find that “regular” (i.e., non-sale) prices were adjusted more frequently in the earlier (higher-inflation) part of their data set, and by about the amount that would be predicted by a model of optimal price adjustment considering a fixed cost (a “menu cost”) of adjusting the firm’s price. They conclude from this that it is important, when assessing the welfare costs expected to follow from choosing a permanently higher rate of inflation, to take account of the increased frequency of price adjustments that should be expected to occur, keeping prices from being as far out of line with current conditions as would otherwise be expected in a more inflationary environment.

The paper also seeks to measure the degree to which there is greater dispersion in the prices of similar products in a higher-inflation environment. Some common models of price adjustment imply that there should be: given staggering of the times at which different firms’ prices happen to be reconsidered, the price that is optimally chosen would vary depending on the rate at which prices in general increase from week to week. If so, this should be an important source of increased distortion of the allocation of resources in a higher-inflation environment. Measurement of the degree of dispersion in the prices of genuinely identical goods is difficult, since different prices for different firms’ goods might reflect heterogeneity of the goods, so that they would have different prices even with fully flexible prices.

For this reason, the authors propose instead to look at the how the average size of price changes (when prices are adjusted) differs between high- and low-inflation periods; the idea is that if prices are adjusted to their currently optimal level whenever they are changed, the size of the price changes that are observed indicates how far prices have drifted from their optimal level just before they are adjusted. They find that the average size of price increases, when they occur, is about the same (a 7 percent increase on average) in their pre-1988 sample as in their post- 1988 sample. Again, they interpret this as evidence that the timing of price changes adjusts endogenously when the rate of inflation increases, in such a way as to reduce the distortions created by inflation relative to what one would expect if the timing of price adjustments were independent of the degree to which a given firm’s prices have gotten out of line with current market conditions. The authors conclude that the welfare costs of chronically higher inflation may not be as large as welfare calculations based on sticky-price models with an exogenous frequency of price adjustment would suggest. The paper is simultaneously an important contribution to policy debates about the costs of inflation; to our understanding of historical facts about price adjustment in the US; and to the empirical basis for assessing the realism of alternative theoretical models of price-setting.

Another area in which Emi Nakamura has had a significant impact is in the study of the effects of government spending shocks, a classic issue in macroeconomics that has been of renewed interest following the widespread use of fiscal stimulus measures by governments in response to the global financial crisis. Estimates of the size of the government spending multiplier have been quite dispersed and remain highly controversial. Nakamura’s work with Jón Steinsson in “Fiscal Stimulus in a Monetary Union: Evidence from U.S. Regions” (AER 2014) brought new data and a fresh identification approach to an important debate.

An important problem in estimating the fiscal multiplier is the difficulty of finding truly exogenous changes in government spending. Researchers have for a long time argued that changes in military purchases are a plausible candidate for exogenous variations in government spending. However, there have not been large variations in aggregate military spending since the Korean War, so that aggregate military spending is of limited use for identifying the government spending multiplier for the U.S. economy of the past 50 years. An important insight of Nakamura and Steinsson’s paper is that while in the aggregate there may not have been large variation in U.S. military spending, there has been sizeable variation in regional military spending, and those regional variations can thus be used to estimate a government spending multiplier. Another important problem with previous studies is that the output effects of government spending should very likely (according to standard theory) depend on the nature of the monetary policy reaction. Some have argued that typical studies under-estimate the multiplier by failing to take account of the extent to which output effects are reduced by the typical monetary response to output booms resulting from government purchases outside of deep recessions, even though the likely response during a deep recession would (arguably) be quite different. Nakamura and Steinsson’s strategy sidesteps this problem, since the monetary policy reaction is common to all states, and so should not be a factor in explaining the differential effects on output across states.

A further complication in estimating government spending multipliers is that their size depends on how changes in government spending are financed. Previous studies have struggled with how to take into account financing considerations. An advantage of Nakamura and Steinsson’s empirical strategy is that regional military spending is financed by federal taxation and thus regions that receive a large chunk of military spending will not have tax structures that are different from regions that do not receive military spending. Thus, considering variations in regional military spending and relating it to regional output variations should provide a much more reliable estimate of the government spending multiplier than previous studies.

The paper offers much more than a clever instrument for measuring the multiplier effect of government purchases. The authors point out that the multiplier estimated for the effect of relatively higher purchases in one state on relative economic activity in that state need not be the same as the multiplier for the effect on national GDP of a nation-wide increase in government purchases (the central issue for debates about the effectiveness of “fiscal stimulus” as a response to recession), because of spillovers between states of the effects of increased purchases in any given state. These spillovers occur not only because increased income in one state leads to increased purchases from out-of-state suppliers, while the national economy is less open, but also because increased relative government spending in one state is not financed by increased relative taxation of that state’s residents, while increased national spending will require increased revenue to be raised from US taxpayers in aggregate. Steinsson and Nakamura address the likely magnitude of the difference between the two multipliers by developing and analyzing a quantitative multi-region New Keynesian general-equilibrium model and asking what the national multiplier would be in the case of a model parameterization that can account for their estimated relative state-level effects. The paper provides an excellent example of work that combines non-structural empirical work with careful model-based analysis of what can be learned from the estimates and makes a substantial contribution to an applied literature of considerable importance for macroeconomic policy.

Nakamura has also made important contributions to empirical analysis of the effects of monetary policy. “High Frequency Identification of Monetary Non-Neutrality: The Information Effect’’ (QJE, 2018, also with Steinsson) studies interest-rate changes in a thirty-minute window around 106 scheduled Federal Reserve announcements between January 2000 and March 2014. As is standard in related literature, financial-market changes observed during this thirty-minute window are attributed to information released in the Federal Reserve announcement. However, unlike some earlier proponents of such “high-frequency identification” of shocks to monetary policy, Nakamura and Steinsson recognize that the news revealed need not only represent a change in expected monetary policy for given economic fundamentals; it could also contain news about the state of the economy that the Fed is aware of but the markets might not have been aware of yet, or news about how the Fed interprets the current state of the economy differently than markets had believed prior to the announcement. The paper's contribution is to draw inferences about monetary non-neutrality while allowing for the possible presence of such information effects, and to build and estimate a theoretical model that can explain the observed effects of Fed announcements.

This problem motivates the development of a model in which Fed announcements can have both an information effect and a pure monetary policy shock, allowing estimation of how big each component in the observed Fed announcements is. The results of this estimation suggest that the proposed model can explain well the observed effects of Fed announcement shocks; that about two-thirds of the announcement shock represents news about future economic fundamentals, and hence that only one-third represents a pure monetary policy shock; and that, despite the great importance of the information effect, the observed responses to Fed announcements are consistent with a high degree of monetary non-neutrality in the U.S. economy. These are important results about fundamental questions in monetary economics, and the paper represents a significant improvement upon prior methodology.

While Nakamura’s most characteristic contributions have been to empirical research, her work is always guided by a sophisticated understanding of the structure of theoretical models, and some of her contributions are primarily theoretical. An important example is her paper “The Power of Forward Guidance Revisited” (AER 2016, with Steinsson and Alisdair McKay). This paper addresses a question about monetary policy that has been a focus of considerable interest in light of central-bank responses to the recent financial crisis both in the US and elsewhere, namely, the extent to which central-bank commitments about future policy (possibly indicating that interest rates should remain at their current level for years into the future) can be an effective way of influencing financial conditions and stimulating aggregate demand, even in the absence of any change in the current level of short-term interest rates.

Simple New Keynesian DSGE models imply that advance commitments to maintain a highly accommodative policy in the future should have a substantial stimulative effect; in fact, in the case of a commitment to low interest rates extending several years into the future, the models predict an immediate effect on both economic activity and inflation that is so strong as to make it difficult to regard this as a realistic prediction—and one that is certainly not consistent with the more modest effects of actual experiments with forward guidance. This has been called “the forward guidance puzzle.” Nakamura and her co-authors argue that the unrealistic implication of the simple New Keynesian models results from the feature that each agent has a single intertemporal budget constraint, as a result of assuming complete financial markets and no borrowing constraints. They analyze the effects of a long-horizon commitment to a fixed nominal interest rate in a model that instead allows for the existence of uninsurable income risk and borrowing constraints and find that while the effects of expectations about monetary policy at shorter horizons are similar to those predicted by the simpler model, the predicted effects of a long-lasting commitment to a fixed nominal interest rate are much weaker. Essentially, they find that in the case of a household with a significant probability of having a point in time over the next several quarters at which its borrowing constraint binds, expectations about monetary policy farther in the future than the time at which the constraint binds do not affect its current ability to spend, and this substantially reduces the predicted effects on current aggregate demand of commitments about policy years in the future.

Their alternative model thus implies that forward guidance is a less powerful tool for getting out of a sharp contraction than simpler models would imply, though it hardly implies that it is irrelevant. The paper is both a contribution to an important policy debate and a useful methodological contribution to the literature on the application of New Keynesian models to assess alternative monetary policies. It has stimulated an active recent literature on “heterogeneous-agent New Keynesian models,” which explores the implications for other aspects of macroeconomic dynamics of introducing income heterogeneity and borrowing constraints.

Nakamura has recently published a JEP article on “Identification in Macroeconomics,” with Steinsson and a new working paper on the role of women’s labor force participation in the slow recovery from recessions observed over the last few decades.  The former is an interesting generalization of the approach discussed above in her fiscal policy paper and also in the price-setting papers: using cross-section variation to identify macroeconomic phenomena and disciplining the aggregate implications with careful structural modeling.  This approach is common to several of Nakamura’s most influential papers and is methodologically eclectic.  It takes advantage of advances in the availability of new and larger data sets to explore cross-section variation, while also recognizing that this alone does not deliver the macroeconomic implications that are of interest to her.  The macro implications require modeling of aggregation that takes into account the heterogeneity in the micro data, and equilibrium considerations.  Moreover, the macro models have implications for the cross section that are testable and provide additional discipline and ability to distinguish competing macro hypotheses. This approach has also been applied in several of her recent papers on the wealth effect from housing, delivering significantly different implications from work focusing only on the micro data.

The working paper “Women, Wealth Effects, and Slow Recoveries” (with Fukui and Steinsson) on slow recoveries from business cycle downturns documents that the slow recovery phenomenon coincides with the convergence of female’s labor force participation to that of males.  That is, as female labor force participation rose during the mid and late-20th century, employment recovered quickly from downturns as women entered the labor force in higher numbers during recoveries.  However, as female labor force participation has risen and converged towards men’s, that dynamic has faded.  The paper argues that this effect alone accounts for 70 percent of the slowing of economic recoveries.  This is an interesting “opposite number” of another labor market finding: that firms adjust faster during downturns, as they adjust to long-run trends more when they are firing.  This result suggests a similar finding during upturns, when there is capacity to draw new workers into the labor market. 

Prior recognition for Nakamura’s accomplishments includes a CAREER Award from the NSF (2011), a Sloan Research Fellowship (2014), the Elaine Bennett Research Prize from the AEA (2014), being named a member of “Generation Next: Top 25 Economists Under 45” by the IMF (2014), and being named one of the decade’s top eight young economists by the Economist (2018). She serves as a Co-editor of the AER, on the CBO’s Panel of Economic Advisers, the AEA Committee on National Statistics, and the BLS Technical Advisory Committee; these appointments testify to the role she has quickly gained in the profession as an expert on issues relating to data construction. Her contributions to the general methodology of empirical macroeconomics, and to the empirical basis for analyses of the effects of monetary and fiscal policies, make Emi Nakamura an outstanding candidate for this year’s John Bates Clark Medal.