NBER Macroeconomics Annual 2014: Volume 29

By Michael Woodford

The twenty-ninth edition of the NBER Macroeconomics Annual continues its tradition of featuring theoretical and empirical research on central issues in contemporary macroeconomics. Two papers in this year’s issue deal with recent economic performance: one analyzes the evolution of aggregate productivity before, during, and after the Great Recession, and the other characterizes the factors that have contributed to slow economic growth following the Great Recession. Another pair of papers tackles the role of information in business cycles. Other contributions address how assumptions about sluggish nominal price adjustment affect the consequences of different monetary policy rules and the role of business cycles in the long-run decline in the share of employment in middle-wage jobs. The final chapter discusses the advantages and disadvantages of the elimination of physical currency.

• Language: English
• Publisher: University of Chicago Press Journals
• Release date: Jun 2, 2015
• ISBN: 9780226268873

Book preview

Contents

Editors’ Introduction

Jonathan A. Parker and Michael Woodford

Abstracts

Productivity and Potential Output before, during, and after the Great Recession

John G. Fernald

Comment

Samuel Kortum and Unni Pillai

Comment

John Haltiwanger

Discussion

Quantifying the Lasting Harm to the US Economy from the Financial Crisis

Robert E. Hall

Comment

Martin S. Eichenbaum

Comment

Narayana Kocherlakota

Discussion

Information Aggregation in a Dynamic Stochastic General Equilibrium Model

Tarek A. Hassan and Thomas M. Mertens

Comment

Guido Lorenzoni

Comment

George-Marios Angeletos

Discussion

Whither News Shocks?

Robert B. Barsky, Susanto Basu, and Keyoung Lee

Comment

Franck Portier

Comment

Lawrence J. Christiano

Discussion

Effective Monetary Policy Strategies in New Keynesian Models: A Reexamination

Hess Chung, Edward Herbst, and Michael T. Kiley

Comment

Lars E. O. Svensson

Comment

Mark Gertler

Discussion

Labor-Market Polarization over the Business Cycle

Christopher L. Foote and Richard W. Ryan

Comment

Richard Rogerson

Comment

Fatih Guvenen

Discussion

Costs and Benefits to Phasing out Paper Currency

Kenneth Rogoff

Editors’ Introduction

Jonathan A. Parker

MIT and NBER

Michael Woodford

Columbia and NBER

The twenty-ninth edition of the NBER Macroeconomics Annual continues with its tradition of featuring theoretical and empirical research on central issues in contemporary macroeconomics. This volume’s papers not only address recent central developments, but also take up important policy-relevant questions and open new debates that we expect to continue in the years to come. Accompanying each paper are two excellent discussions, each written by a leading scholar in the area. Finally, following our recent tradition, this volume also contains the text of an invited speech by an NBER researcher who has been influential beyond the world of research, through public service and/or contribution to public policy debates.

Two papers in this year’s issue deal with recent economic performance, with one paper analyzing the evolution of aggregate productivity before, during, and after the Great Recession, and another characterizing the poor economic performance following the Great Recession. Another pair of papers tackles the role of information in business cycles, with one paper studying the dynamic properties of model economies with dispersed information, and the other addressing the extent to which news about future productivity causes business cycles in the present. Our final two contributions address central issues in monetary policy and labor markets—the robustness of the implications of different monetary policy rules to alternative models of sluggish nominal price adjustment, and the role of business cycles in the long-run decline in the share of employment in middle-wage jobs.

Perhaps the central question following the Great Recession is why the growth of the US economy, and the growth of aggregate productivity in particular, is so low. Productivity and Potential Output Before, During, and After the Great Recession, by John G. Fernald is a careful study of the dynamics of aggregate productivity through this volatile period. The paper shows that productivity growth slows to a new lower trend in 2005, well before the start of the Great Recession and even before the beginnings of the financial crisis. This 2005 slowdown is evident in the official measure of total factor productivity—which declines during the Great Recession—as well as a utilization-adjusted measure that is a better measure of productivity of factors in use—which booms during the recession. Further, and contradicting much popular opinion, the paper documents that after adjusting for utilization and local trends, productivity growth following the recession is typical of other recessions. That is, while this recession may have been special in many ways, it has not been associated with an unusually slow recovery in the productivity of employed factors. The paper then tackles the deeper question: If the Great Recession or its fallout did not cause the slowdown in growth rate of productivity, what did? The paper marshals evidence from different industries and from different US states that address specific hypothetical causes, such as the boom and bust in house prices and the growth and collapse of finance. On balance, the evidence suggests that the slowdown in productivity growth is due to technology and the waning of the information technology boom that began in the mid-1990s. Finally, most speculatively and provocatively, the paper extrapolates from its analysis to the implications for the future rate of US economic growth over the next decade.

Our second paper also seeks to clarify the reasons for the US economy’s poor performance in the aftermath of the recent crisis. In Quantifying the Lasting Harm to the US Economy from the Financial Crisis, Robert E. Hall offers a decomposition of the 13% shortfall of real gross domestic product (GDP) (as of 2013) from the economy’s apparent trend prior to the crisis into parts due to each of four sources—below-trend growth of capital, of the labor force, of hours worked per member of the labor force, or of total factor productivity—through a standard growth-accounting approach, examines the quantitative importance of each of these factors and discusses the most likely reasons for departures from the prior trend of each kind. While continuing slack utilization of the labor force has been part of the problem, and one emphasized in much public policy debate, Hall argues that this factor accounts for only a relatively small part (only about one-sixth) of the GDP shortfall remaining in 2013; each of the other three factors accounts for a larger share, with below-trend growth of the capital stock (accounting for 30% of the total shortfall) the most important.

The paper is particularly concerned with the implications of its analysis of the sources of the persistent shortfall for the likelihood that the lost ground will be regained eventually. Hall expresses optimism that the part of the shortfall due to slower growth of the capital shock should eventually be corrected—even if not too rapidly, even under the most optimistic scenario—as he finds little ground to attribute the recent low rate of investment to factors that should permanently change the long-run capital/output ratio. He also argues that the part due to higher unemployment should eventually be eliminated, though the rate at which this can occur should continue to be slowed by a reduction in the efficiency of job matching that he attributes to a crisis-induced change in the composition of the pool of unemployed workers. He argues that there is no reason to expect the shortfall in total factor productivity (TFP) growth to be later reversed, even if reduced innovation has been caused by the crisis—that much of the below-trend growth of the labor force may be permanent as well. Hence the paper is relatively pessimistic about the ease with which the continuing output shortfall can be reduced; much of it should be expected to change only slowly, and some of it should not be reversible at all, even with aggressive demand stimulus policies.

An important recent development in macroeconomic theory has been a revival of interest in the role of information heterogeneity and the resulting differences in beliefs across market participants in generating and propagating fluctuations in economic activity. Much of this work, however, has offered purely qualitative results in the context of fairly stylized theoretical models; the quantitative significance of the effects under realistic assumptions is seldom addressed, largely because models with heterogeneous information have seemed intractable except under highly special assumptions. In Information Aggregation in a DSGE Model, Tarek A. Hassan and Thomas M. Mertens make important progress in the development of a more quantitative literature, by first incorporating a noisy rational expectations model of partial information aggregation in financial markets into a quantitative dynamic stochastic general equilibrium (DSGE) model of fluctuations in aggregate economic activity, and second by showing how an approximation method can be used to solve for the dynamics implied by the model.

Hassan and Mertens consider a real business-cycle model that has been extended in directions that allow it to match certain basic asset-pricing facts, such as the average returns on both riskless short-term assets and on the aggregate stock market, in addition to the facts about cyclical variation in aggregate quantities that were the focus of the classic real business cycle literature. These include assuming an exogenous process for aggregate productivity that includes both long-run and short-run risk components, and Epstein-Zin preferences for the representative household, as in the finance literature that attributes the equity premium to long-run consumption risk. Moreover, each household is assumed to receive a noisy private signal about the coming month’s shock to the long-run component of productivity; these signals are the source of belief heterogeneity in the model. Finally, as in the noisy rational expectations literature in finance, noise traders are also introduced into the equity market in order to prevent the price of stocks from fully revealing the aggregate productivity shock; hence, beliefs about future productivity remain heterogeneous in their model.

The paper sets out a methodology for assessing the consequences of such heterogeneity for the business cycle. The paper finds that allowance for dispersed information affects the quantitative predictions of this model in a number of respects. The risk-free rate falls while the predicted equity premium rises, thus helping to account for two basic facts that have been problematic for many macrofinance models. Many business-cycle predictions are modified as well; notably, even a moderate degree of heterogeneity of expectations substantially increases the predicted correlation between fluctuations in aggregate consumption and aggregate investment spending, again better matching the quantitative properties of observed business cycles. The model fails, however, to simultaneously match all of the data moments that are considered; notably, it is difficult for the model to account for the degree of dispersion in individual forecasts of GDP growth indicated by the Survey of Professional Forecasters. This suggests that the particular way of modeling belief heterogeneity considered here is not sufficiently noisy, but movement of the literature to a discussion of quantitative realism seems to us an important step forward.

Over the past decade, a number of papers have considered the possibility that some economic fluctuations may be driven not by innovations in policies or technologies today, but rather by changes in expectations today about policies or technologies in the future. In Whither News Shocks? Robert B. Barsky, Susanto Basu, and Keyoung Lee use identified vector autoregressions (VARs) to delineate the importance of these Pigouvian fluctuations for business cycles and then use a DSGE model to interpret their results. The first question that the paper addresses is whether, given minimal assumptions used to identify news shocks, these shocks account for a quantitatively important share of the fluctuations observed in the US economy. One might suspect that this answer is yes, given both the paucity of other measurable drivers of business cycles and the interpretation of recent fluctuations as expectations-driven, an interpretation in which expectations amplified growth before the Great Recession and, when determined to be overly optimistic, contributed to the severity of the downturn. But on the other hand, structural shocks identified by previous researchers have mostly been unable to account for much of the variance of output associated with the business cycle.

Barsky, Basu, and Lee show that there are quantitatively important movements in expected future total factor productivity that are unrelated to movements in current productivity, and moreover, that these innovations generate predictable movements in productivity that account for roughly half the variation in productivity at a horizon of five years. These identified news shocks track consumer confidence, are followed by lower inflation, and have prolonged real effects on productivity that differ from those commonly assumed in DSGE models.

The paper then uses structural macroeconomic models to help delineate and interpret the observed dynamics that follow a news shock. While consistent with previous research, there is evidence that news shocks increase consumption and reduce labor, as in the textbook neoclassical growth model and unlike in the typical business cycle. That said, this evidence is statistically weak and limited to the period immediately following a news shock. At the same time, the longer term dynamics exhibit comovement between consumption and labor, and the paper investigates the extent to which a New Keynesian model can account for these dynamics. In sum, this paper reinvigorates the literature on news shocks, and the discussions take up the debate.

Our fifth paper considers an issue that has been prominent in recent macroeconomic policy debates, both in the United States and a number of other countries: the way in which the effects of policy should be expected to be different when monetary policy is constrained by the zero lower bound (ZLB) on interest rates. A number of influential papers have argued, on the basis of analyses using New Keynesian DSGE models, that commitments to maintain future monetary policy looser in the future than would otherwise be justified by conditions at the time can be an effective source of aggregate demand stimulus, even when the current interest-rate target cannot be lowered. Analyses using similar models have also found that government purchases can be expected to have a larger stimulative effect on aggregate activity in the case of an economy at the ZLB, suggesting a more important role for counter-cyclical fiscal policy under such circumstances.

In Effective Monetary Policy Strategies in New Keynesian Models: A Reexamination, Hess Chung, Edward Herbst, and Michael T. Kiley consider the extent to which the conclusions obtained in these studies are robust to alternative specifications of the macroeconomic model used to predict the effects of alternative policies. They point out that all of the best-known analyses are based on fairly similar New Keynesian models, but argue that sounder conclusions can be reached by also considering the predictions of alternative models. The paper focuses in particular on the role of the New Keynesian Phillips curve specification, in which the short-run trade-off between inflation and the output gap depends on current expectations regarding future inflation, in generating the rather striking conclusions obtained in the ZLB literature. Chung and colleagues present evidence indicating that an alternative specification, a variant of the sticky information Phillips curve proposed by Greg Mankiw and Ricardo Reis, fits historical US data as well or better than a standard New Keynesian Phillips curve specification. The alternative Phillips curve makes relatively little difference for predictions about the effects of monetary policy under a policy rule of the kind estimated to describe US policy before reaching the ZLB; yet, they find that substitution of the sticky-information Phillips curve specification into an otherwise identical DSGE model makes a significant difference to quantitative predictions about the effects of policy at the ZLB.

In particular, they find that under the alternative Phillips curve specification the effects on current economic activity of an announced tightening or loosening of monetary policy several quarters in the future are only a fraction of the size of those predicted by a model using the more standard specification. This suggests that explicit statements about future policy may not be as powerful a policy tool as the New Keynesian analysis would imply. Nonetheless, their alternative model still implies that a change in anticipated future policy can influence economic activity at the ZLB, and it still implies that an optimal policy in response to a shock that causes the ZLB to bind would involve committing to keep interest rates lower for longer than would occur under a reversion to historical policy as soon as this was allowed by the ZLB, and would involve allowing inflation to temporarily overshoot its long-run target value as a result of the prolonged stimulus. A number of other conclusions about effective policy stressed in the New Keynesian literature are also found to be robust to use of the alternative model. In particular, they find that highly inertial interest rate rules that respond to deviations of the level of nominal GDP (or the price level) from a target path have desirable properties, both when the economy is hit by adverse supply shocks and when an unusually large demand shock causes the ZLB to bind. Thus, while the paper illustrates the value of checking the robustness of policy analyses to alternative model specifications, it suggests that the DSGE literature has yielded policy insights of some degree of robustness.

The final research paper in the volume, Labor Market Polarization over the Business Cycle by Christopher L. Foote and Richard W. Ryan, uses a wealth of labor market data to examine the role of macroeconomic fluctuations in the long-run decline of jobs in the middle of the wage distribution. The literature on the polarization of the labor market has shown that jobs in the middle of the wage distribution are disproportionately characterized by routine tasks like those performed on assembly lines or in traditional clerical work. As such it has been hypothesized that these jobs are more easily displaced by automation or international trade, and the share of such jobs, defined as those in occupations primarily requiring routine manual work, has declined from nearly half of all employment to less than a quarter from 1950 to 2010. These losses are concentrated at the start of recessions, suggesting that middle-skill workers reallocate themselves to other jobs when aggregate productivity is low.

But Foote and Ryan demonstrate that there is a simple alternative explanation with a set of interesting implications. First, the middle of the wage distribution consists disproportionately of jobs in disproportionately cyclical industries like manufacturing and construction. Manufacturing in particular has both a procyclical employment share and a declining employment share. The procyclicality of middle-skill employment is quite stable over time, even as labor market polarization has become more firmly established. Thus the observed relationship is due to cyclical demand in industries that have made heavy use of middle-skill workers and have declining employment shares, suggesting not reallocation but a trend decline buffered by the business cycle. Second, the paper shows that these polarization job losses are permanent: cyclical job losses due to polarization are associated with declines in labor-force participation among middle-skill workers. The decline in participation associated with polarization actually accounts for a large fraction of the total decline in labor-force participation since the late 1970s. In conclusion, the paper argues that these results imply a more nuanced view of both the cyclical employment fluctuations and labor market polarization than contained in most current models. In particular, the participation margin is important for quantitative models of polarization and the concentration of cyclicality implies an important role for industry dynamics and recall unemployment in quantitative accounts of labor market fluctuations.

The final chapter is a speech by Kenneth Rogoff, longtime NBER member, former Chief Economist and Director of Research at the International Monetary Fund, author of This Time Is Different: Eight Centuries of Financial Folly (with Carmen M. Reinhart), and a former editor of the NBER Macroeconomics Annual. In his talk, Ken Rogoff discussed the advantages and disadvantages of the elimination of physical currency. As a value-weighted share of total transactions, actual currency is now seldom used to make payment in transactions. Most purchases use checks, credit cards, debit cards, and so forth. The long-standing trend away from cash transactions has been highlighted by the advent of digital currencies. Rogoff considers the costs and benefits of the elimination of physical currency—notes and coins—entirely. While noting that this idea is provocative from the standpoint of symbolism and perhaps trust in a currency, Rogoff argues that the existence of physical currency facilitates crime and hinders negative nominal interest rates. On the other hand, there is demand for physical currency so that eliminating it would reduce revenues from seigniorage, and issuance of physical currency funds the central bank, facilitating its independence. The published text of the talk provides a quantitative consideration of the costs of eliminating currency and contrasts them with the likely benefits.

Finally, the authors and the editors would like to take this opportunity to thank Jim Poterba and the National Bureau of Economic Research for their continued support for the NBER Macroeconomics Annual and the associated conference. We would also like to thank the NBER conference staff, particularly Rob Shannon, for his continued excellent organization and support. Financial assistance from the National Science Foundation is gratefully acknowledged. Mariana Garcia Schmidt and Ben Hebert provided invaluable help in preparing the summaries of the discussions. And last but far from least, we are grateful to Helena Fitz-Patrick for her invaluable assistance in editing and producing the volume.

Endnote

For acknowledgments, sources of research support, and disclosure of the authors’ material financial relationships, if any, please see http://nber.org/chapters/c13406.ack.

© 2015 by the National Bureau of Economic Research. All rights reserved.

978-0-226-26873-6/2015/2014-0001$10.00

Abstracts

1. Productivity and Potential Output before, during, and after the Great Recession

John G. Fernald

US labor and total-factor productivity growth slowed prior to the Great Recession. The timing rules out explanations that focus on disruptions during or since the recession, and industry and state data rule out bubble economy stories related to housing or finance. The slowdown is located in industries that produce information technology (IT) or that use IT intensively, consistent with a return to normal productivity growth after nearly a decade of exceptional IT-fueled gains. A calibrated growth model suggests trend productivity growth has returned close to its 1973–1995 pace. Slower underlying productivity growth implies less economic slack than recently estimated by the Congressional Budget Office. As of 2013, about three-fourths of the shortfall of actual output from (overly optimistic) prerecession trends reflects a reduction in the level of potential.

2. Quantifying the Lasting Harm to the US Economy from the Financial Crisis

Robert E. Hall

The financial crisis and ensuing Great Recession left the US economy in an injured state. In 2013, output was 13% below its trend path from 1990 through 2007. Part of this shortfall—2.2 percentage points out of the 13—was the result of lingering slackness in the labor market in the form of abnormal unemployment and substandard weekly hours of work. The single biggest contributor was a shortfall in business capital, which accounted for 3.9 percentage points. The second largest was a shortfall of 3.5 percentage points in total factor productivity. The fourth was a shortfall of 2.4 percentage points in labor-force participation. I discuss these four sources of the injury in detail, focusing on identifying state variables that may or may not return to earlier growth paths. The conclusion is optimistic about the capital stock and slackness in the labor market and pessimistic about reversing the declines in total-factor productivity and the part of the participation shortfall not associated with the weak labor market.

3. Information Aggregation in a Dynamic Stochastic General Equilibrium Model

Tarek A. Hassan and Thomas M. Mertens

We introduce the information microstructure of a canonical noisy rational expectations model (Hellwig 1980) into the framework of a conventional real business-cycle model. Each household receives a private signal about future productivity. In equilibrium, the stock price serves to aggregate and transmit this information. We find that dispersed information about future productivity affects the quantitative properties of our real business-cycle model in three dimensions. First, households’ ability to learn about the future affects their consumption-savings decision. The equity premium falls and the risk-free interest rate rises when the stock price perfectly reveals innovations to future productivity. Second, when noise trader demand shocks limit the stock market’s capacity to aggregate information, households hold heterogeneous expectations in equilibrium. However, for a reasonable size of noise trader demand shocks the model cannot generate the kind of disagreement observed in the data. Third, even moderate heterogeneity in the equilibrium expectations held by households has a sizable effect on the level of all economic aggregates and on the correlations and standard deviations produced by the model. For example, the correlation between consumption and investment growth is 0.29 when households have no information about the future, but 0.41 when information is dispersed.

4. Whither News Shocks?

Robert B. Barsky, Susanto Basu, and Keyoung Lee

Does news about future productivity cause business-cycle fluctuations? What other effects might it have? We explore the answer to this question using semistructural vector autoregressions (VARs), where news is defined as the innovation in the expectation of total factor productivity (TFP) at a fixed horizon in the future. We find that systems incorporating a number of forward-looking variables, including stock prices, consumption, and consumer confidence and inflation robustly predict three outcomes. First, following a news shock, TFP rises for several years. Second, inflation falls immediately and substantially, and stays low, often for 10 quarters or more. Third, there is a sharp increase in a forward-looking measure of consumer confidence. Consumption typically rises following good news, but investment, consumer durables purchases, and hours worked typically fall on impact. All the quantity variables subsequently rise, as does TFP. Depending on the specification of the reduced form VAR, the activity variables may lead TFP to some extent—possibly lending some support to the hypothesis of news-driven business cycles—or they may move in lockstep with productivity. For the most part, the quantity and inflation responses are quite consistent with the predictions of a standard New Keynesian model augmented with real wage inertia.

5. Effective Monetary Policy Strategies in New Keynesian Models: A Reexamination

Hess Chung, Edward Herbst, and Michael T. Kiley

We explore the importance of the nature of nominal price and wage adjustment for the design of effective monetary policy strategies, especially at the zero lower bound. Our analysis suggests that sticky-price and sticky-information models fit standard macroeconomic time series comparably well. However, the model with information rigidity responds differently to anticipated shocks and persistent zero lower bound episodes—to a degree important for monetary policy and for understanding the effects of fundamental disturbances when monetary policy cannot adjust. These differences may be important for understanding other policy issues as well, such as fiscal multipliers. Despite these differences, many aspects of effective policy strategy are common across the two models: in particular, highly inertial interest rate rules that respond to nominal income or the price level perform well, even when hit by adverse supply shocks or large demand shocks that induce the zero-lower bound. Rules that respond to the level or change in the output gap can perform poorly under those conditions.

6. Labor-Market Polarization over the Business Cycle

Christopher L. Foote and Richard W. Ryan

Job losses in the Great Recession were concentrated among middle-skill workers, the same group that has suffered the most over the long run from automation and international trade. How might long-run occupational polarization be related to cyclical changes in middle-skill employment? We find that middle-skill jobs have traditionally been more cyclical than other jobs, in part because of the volatile industries that tend to employ middle-skill workers. Also, unemployed middle-skill workers appear to have few attractive or feasible employment alternatives outside of their skill class, and the drop in male participation rates during the past several decades can be explained in part by a drying-up of middle-skill job opportunities. Taken together, these results imply that any model relating polarization to middle-skill employment fluctuations must go beyond pure search motives to include industry-level effects as well as a labor-force participation margin. The results thus provide encouragement for a growing literature that integrates macrolabor search models with macro-macro models featuring differential industry cyclicalities and convex preferences over consumption and leisure.

© 2015 by the National Bureau of Economic Research. All rights reserved.

978-0-226-26873-6/2015/2014-0002$10.00

Productivity and Potential Output before, during, and after the Great Recession

John G. Fernald

Federal Reserve Bank of San Francisco

I. Introduction

When we look back at the 1990s, from the perspective of say 2010, … [w]e may conceivably conclude … that, at the turn of the millennium, the American economy was experiencing a once-in-a-century acceleration of innovation.… Alternatively, that 2010 retrospective might well conclude that a good deal of what we are currently experiencing was just one of the many euphoric speculative bubbles that have dotted human history.

—Federal Reserve Chairman Alan Greenspan (2000)

Disappointing productivity growth … must be added to the list of reasons that economic growth has been slower than hoped.

—Federal Reserve Chairman Ben Bernanke (2014)

The past two decades have seen the rise and fall of exceptional US productivity growth. This paper argues that labor and total factor productivity (TFP) growth slowed prior to the Great Recession. It marked a retreat from the exceptional, but temporary, information technology-fueled pace from the mid-1990s to early in the twenty-first century. This retreat implies slower output growth going forward as well as a narrower output gap than recently estimated by the Congressional Budget Office (CBO 2014a).

Industry and state data show that the pre–Great Recession productivity slowdown was in sectors that produce information technology (IT) or that use IT intensively. Sectors that were obviously unusual or euphoric in the first decade of the twenty-first century—including housing and finance—were not the source.

Figure 1 illustrates that the mid-1990s surge in productivity growth ended prior to the Great Recession. The surge in labor productivity growth, shown by the height of the bars, came after several decades of slower growth.¹ But in the decade ending in 2013:Q4, growth has returned close to its 1973–1995 pace. The figure shows that the slower pace of growth in both labor productivity and TFP was similar in the four years prior to the onset of the Great Recession as in the six years since.

Fig. 1. Contributions to labor productivity growth, business sector

Source: Fernald (2014), BEA, BLS.

That the slowdown predated the Great Recession rules out causal stories from the recession itself. Theory and previous empirical literature (discussed in Section II.D) provides only limited support for the view that the Great Recession should have changed the underlying path of TFP. Figure 1 suggests no evidence that productivity was slower (or much faster) from 2007 to 2013 than in the several years before that. The evidence here complements Kahn and Rich’s (2013) finding in a regime-switching model that by early 2005—that is, well before the Great Recession—the probability reached nearly unity that the economy was in a low-growth regime.

A natural hypothesis is that the slowdown was the flip side of the mid-1990s speedup. Considerable evidence, discussed in Section III.A, links the TFP speedup to the exceptional contribution of IT—computers, communications equipment, software, and the Internet. Information technology has had a broad-based and pervasive effect through its role as a general purpose technology (GPT) that fosters complementary innovations, such as business reorganization (see Bresnahan and Trajtenberg (1995) and Helpman, ed., 1998).

Industry TFP data provide evidence in favor of the IT hypothesis versus alternatives. Notably, the euphoric bubble sectors of housing, finance, and natural resources do not explain the slowdown. Rather, the slowdown is in the remaining three-quarters of the economy, and is concentrated in industries that produce IT or that use IT intensively. Information technology users saw a sizable bulge in TFP growth early in the first decade of the twenty-first century, even as IT spending itself slowed. That pattern is consistent with the view that benefiting from IT takes substantial intangible organizational investments that, with a lag, raise measured productivity. By the middle of the first decade of the twenty-first century, the low-hanging fruit of IT had been plucked.

State data on gross domestic product (GDP) per worker rule out indirect channels through which the housing bubble and bust might have mattered. States differ in how much house prices ran up early in the twenty-first century and collapsed after 2006. Those differences could have influenced innovation through net-worth channels. There is little evidence that housing dynamics contributed much to the dynamics of the productivity slowdown. Rather, it is the common cross-state slowdown in IT-intensive industries that predominates.

I then turn to two implications of the mid-2000s productivity slowdown. First, a multisector neoclassical growth model implies steadystate business-sector labor-productivity growth of about 1.9%, as shown at the far right of figure 1. Prior to the Great Recession, typical estimates were notably higher. Using demographic estimates from the CBO (2014a), my benchmark estimate implies longer-term growth in GDP of about 2.1% per year. As figure 1 shows, three out of the past four decades have shown this slower pace of productivity growth. That pace, rather than the exceptional 1995–2003 pace, appears normal.

Second, by 2013, the output gap, defined as the difference between actual and a production-function measure of potential output, is narrower than estimated by the CBO (2014a). I decompose the CBO’s gap into a utilization gap that reflects cyclical mismeasurement of TFP as well as an hours gap. The CBO estimates that the utilization gap in 2013 was as deep as any time in history other than 1982 and 2009, and was comparable to its level in 1975. In contrast, empirical estimates from Fernald (2014; following Basu et al. 2013) suggest a small utilization gap.

Figure 2 shows two alternatives to the CBO estimates of potential, with different estimates of the utilization gap. Both use the CBO labor gap to measure deviations of hours worked from steady state. One uses actual TFP, which imposes a utilization gap of zero. When utilization eventually returns to normal—as it plausibly did prior to 2013—this measure is appropriate. The second, labeled Fernald, uses my utilization estimate. By 2013, the alternatives imply that about three-quarters of the 2013 shortfall of actual output from the estimated precrisis trend reflects a decline in potential output. These estimates lie well below the CBO’s (2014a), which itself is well below its prerecession trend. The differences arise from the CBO’s assumed path for potential TFP. In contrast to the evidence in this paper, the CBO has no mid-1990s pickup in productivity and much less of a mid-2000s slowdown.

Fig. 2. Potential output and its pre-crisis trend

Notes: Annual data, 2009$. Figure compares actual real GDP to the CBO’s projections for potential prior to the Great Recession (the 2007 line) and the CBO’s (2014a) projection, as well as two alternative measures of potential that follow the CBO methodology but with different assumptions about utilization. Both alternatives use the CBO’s estimated hours gap (deviation of hours from steady state). Actual TFP assumes utilization is constant, so that actual TFP measures technology. Fernald uses estimated utilization and labor-quality gaps. The 2007 CBO estimates are from January 2008, which were based on data through 2007:Q3. Those estimates have been rescaled to 2009$ so that the 2007 value equals the level in CBO (2014a).

Source: CBO, BEA, and the author’s calculations.

An important caveat is that production-function measures of potential output are inherently cyclical because investment is cyclical. Slow aggregate-demand growth in the recovery has led to slow closing of the output gap. Cyclically weak investment, in turn, has contributed to slow potential growth; indeed, capital input grew at the slowest pace since World War II. Slow capital growth does not directly affect output gaps—in the CBO definition (as well as the usual dynamic stochastic general equilibrium [DSGE] definition), it affects both actual and potential output. In standard models, capacity should rebound (raising potential growth above its steady-state rate) as the economy returns toward its steady-state path.²

Section II discusses facts about the slowdown in measured labor and total-factor productivity, and compares the experience during and since the Great Recession to previous recessions and recoveries, finding that productivity experience was comparable. Section III assesses explanations for the productivity slowdown, using industry data and (maybe) regional data. Section IV uses a multisector growth model to project medium- to long-run potential output growth. The section also discusses key uncertainties. Section V then draws on the preceding analysis to discuss current potential output and slack in the context of the general methodology followed by the Congressional Budget Office. Section VI concludes.

II. Productivity Growth before the Great Recession

Trend productivity growth slowed several years before the Great Recession.

A. The mid-2000s Slowdown in Labor-Productivity Growth

Figure 3 shows the log level of business-sector labor productivity, which rationalizes the subsamples shown in figure 1.³ The mid-1990s speedup in growth is clear. The literature discussed in Section III.A links that speedup to information technology (IT). The slowdown in the mid-2000s is also clear. The dates of the vertical bars are suggested by the Bai-Perron test for multiple structural change in mean growth rates for the period since 1973. I have shown the traditional new-economy 1995:Q4 start date along with a slowdown date of 2003:Q4. The breaks are statistically significant.⁴

Fig. 3. Business sector labor productivity since 1973

Note: Log level (times 100), measured as cumulative growth since 1973:Q2.

Source: BLS and Fernald (2014).

The Bai-Perron results complement the findings of Kahn and Rich (2007, 2013). They estimate a regime-switching model, using data on labor productivity, labor compensation, and consumption. They find that productivity switched from a high-growth to a low-growth regime around 2004. By early 2005, the probability that the economy was in a low-growth regime was close to unity.

B. Growth-Accounting Identities

Growth accounting provides further perspective on the forces underpinning the slowdown. Suppose there is a constant returns aggregate production function for output, Y:

Variable A is technology; K and L are observed capital and labor. Variable W is the workweek of capital and E is effort—that is, unobserved variation in the utilization of capital and labor; Ki is input of a particular type of capital—computers, say, or office buildings. Similarly, Hi is hours of work by a particular type of worker, differentiated by education, age, and other characteristics. Time subscripts are omitted.

, j ∈ K, L, j ∈ K, L. Differentiating logarithmically (where hats are log changes) and imposing the first-order conditions yields:

Various input aggregates on the right-hand side are defined as:

Growth in capital services, K̂, is shareweighted growth in the different types of capital goods. Similarly, growth in labor services, L̂, is share-weighted growth in hours for different types of workers. Total hours, H ≡ H1 + Hcaptures variations in capital’s workweek and labor effort.

TFP growth, or the Solow residual, is output growth not explained by (observed) input growth:

The second line follows from equation (2). I will always take TFP growth to be this measured Solow residual, defined by the first line in equation (4), and refer to  as utilization-adjusted TFP.

A large literature discusses why measured TFP might not reflect technology over the business cycle.. Basu, Fernald, and Kimball (2006) and Basu et al. (2013) implement a theoretically based measure of utilization. Their method essentially involves rescaling variations in an observable intensity margin of (detrended) hours per worker. I return to this measure below.

From equations (2) and (4), labor productivity growth, defined as growth in output per hour, is then:

Loosely, labor productivity rises if workers have more capital or better skills (quality), or if innovation raises technology. In the short run, cyclical variations in utilization also matter.

C. Aggregate Data and Growth-Accounting Results

Both TFP and capital deepening contributed to the mid-2000s slowdown in labor-productivity growth. Specifically, figure 4 shows components of equation (5) using the quarterly growth-accounting data set described in the appendix. These data provide quarterly business-sector growth accounting variables through 2013. Variables shown are in log levels (i.e., cumulated log changes). The utilization measure applies annual estimates from Basu et al. (2013) to quarterly data. Utilization is based on variations in industry hours per worker. Using restrictions from theory, Basu and colleagues relate unobserved intensity margins of capital’s workweek and labor effort to this observed intensity margin.

Fig. 4. Evolution of key growth-accounting variables

Note: Log levels (times 100), measured as cumulative growth since 1973:Q2. Level of utilization is set to zero in 1987:Q4.

Source: Fernald (2014).

Panel A shows TFP and utilization-adjusted TFP.⁶ These series grew rapidly from the mid-1990s to the mid-2000s, then essentially hit a flat spot. Panel B shows capital-deepening, K/(H · LQ). In the early twenty-first century, capital-deepening growth slowed (consistent, perhaps, with the slowdown in technology growth). Panel C shows that labor quality accelerated in the Great Recession as low-skilled workers disproportionately lost jobs. Finally, panel D shows utilization itself. This series is clearly highly cyclical. By early 2011, this measure had recovered to a level close to its prerecession peaks. Indeed by the end of the sample, labor productivity (figure 3) or TFP (figure 4, panel A) appear to lie more or less on the slow-trend line from the mid-2000s.

D. Productivity Growth during the Great Recession

That the slowdown predated the Great Recession suggests it was not a result of the recession itself. Still, if productivity during the recession were unusual, that might suggest a role for the recession. For example, a few years of bad productivity luck before the recession could have been followed by the greater, and more persistent, bad luck of a severe recession. This section argues this was not the case. Rather, productivity behaved similarly to previous deep recessions: TFP and utilization fell very sharply, but recovered strongly once the recession ended.⁷

Figure 5 shows spider charts comparing the Great Recession to the nine previous recessions (1953–2001). In each panel, the horizontal axis shows the number of quarters from the peak. In the Great Recession, for example, quarter 0 corresponds to 2007:Q4. The vertical axis is the percent change since the peak. I remove local trends from all data.⁸

Fig. 5. Comparing recessions

Notes: For each plot, quarter 0 is the NBER business-cycle peak which, for the Great Recession, corresponds to 2007:Q4. The shaded regions show the range of previous recessions since 1953. Local means are removed from all growth rates prior to cumulating, using a bi-weight kernel with bandwidth of 48 quarters.

Source: Fernald (2014).

Panels A and B show how unusual output and hours were, with steep declines in both. For the first three quarters (through 2008:Q3), the declines in output