Crises in Economic Thought, Secular Stagnation, and Future Economic Research

Abstract

Previous article FreeCrises in Economic Thought, Secular Stagnation, and Future Economic ResearchLawrence SummersLawrence SummersHarvard Kennedy School and NBER Search for more articles by this author Harvard Kennedy School and NBERPDFPDF PLUSFull Text Add to favoritesDownload CitationTrack CitationsPermissionsReprints Share onFacebookTwitterLinked InRedditEmailQR Code SectionsMoreI am very flattered by the invitation to be the dinner speaker at this conference—now celebrating its 30th anniversary. I was proud to coauthor the lead article in the inaugural volume of this series on hysteresis and European unemployment with Olivier Blanchard (Blanchard and Summers 1986). Less successful in its original incarnation was a paper I presented a couple of years later on “The Scientific Illusion in Empirical Macroeconomics” that did not get published in this forum, but was published a few years later as Summers (1991). In ways I certainly did not expect, both these papers contain ideas that I believe are relevant to current policy dilemmas.More generally as someone who has held policy-making positions, I can attest to the influence of academic research of the kind presented and debated at these conferences. Over the years, the effect of research on policy has greatly increased. How could it be otherwise when the vast majority of the world’s central bankers are now economics PhDs with experience carrying out macroeconomic research?I have always believed that academics are most useful when they reconsider existing paradigms and least useful when they make recommendations on current policy where they often lack specific context. So, I have decided that I will try to behave in character and offer a few provocations regarding current macroeconomic thinking that come out of my work on secular stagnation and hysteresis. My hope is to make a case that analytical approaches quite different from those employed over the years at the NBER Macro Annual Conferences and by the world’s central banks are necessary to make sense of current events. I will explain why I believe that current inflation-targeting regimes will not survive the next decade in most countries. Finally, I suggest a variety of issues that I believe should preoccupy macroeconomic researchers in the years ahead.I. Crises and Macro ThoughtA. Financial Crisis and Recovery in Historical ContextFigure 1 points out what I believe should be a (if not the) principal preoccupation of contemporary macroeconomics. We know that the Great Depression started in the fall of 1929 and is normally thought of as having persisted for approximately 12 years, until the onset of World War II. We also know that the financial crisis started in 2007, and can project 12 years forward with a mixture of eight years of actuals and four years of CBO forecasts. The comparison is very sobering. Taking a 12-year view, and comparing 1929 to 1941 and 2007 to 2019 periods, US GDP will do no better over the entirety of the financial crisis and its aftermath than it did during the Great Depression. To be sure, things went down much faster during the Depression and they came back much faster, but overall the financial crisis decade is comparable to the decade of the 1930s. The rest of the industrial world had a somewhat better Depression and has performed relatively worse in recent years. So, if these comparisons were made on a global basis, the current episode would appear even worse compared to the Great Depression.Fig. 1. US real GDP per 18-to-64-year-olds, Great Depression and Great RecessionSource: Bureau of Economic Analysis, NBER, CBO, and Census.View Large ImageDownload PowerPointB. The Depression and Economic Thought: Keynesian Fiscal Stabilization as Growth EnhancerIn the wake of the Depression, economists came to accept that business fluctuations were more than a cyclical phenomenon. While there was much debate about the causes of the Depression and debate about how policy shaped its course, there was consensus that it did not have to happen. So, with better policy, the output and employment gaps associated with the Depression could have been mitigated, leading to higher levels of cumulative output and employment. It was the aspiration of Keynesian economics in the 1950s, 1960s, and 1970s to make societies richer over time by preventing costly downturns.The Keynesian aspiration was not to merely reduce the amplitude of cyclical fluctuations, but also to increase overall growth. For instance, President Kennedy’s 1963 Economic Report of the President argued that fiscal stimulus would boost long-run potential output: “among the costs of prolonged slack is slow growth. An economy that fails to use its productive potential fully feels no need to increase it rapidly. The incentive to invest is bent beneath the weight of excess capacity. Lack of employment opportunities slows the growth of the labor force. Defensive restrictive practices—from featherbedding to market sharing—flourish when limited markets, jobs, and incentives shrink the scope for effort and ingenuity. But when the economy breaks out of the lethargy of the past five or six years, the end to economic slack will by itself mean faster growth” (Kennedy 1963). Even those like Milton Friedman, who opposed discretionary stabilization policy, argued that appropriate policy rules would prevent or at least limit downturns in ways that raised average levels of output.C. Stagflation and Economic Thought: Inflation Targeting as Output StabilizerThinking radically changed in the 1970s shortly after Milton Friedman (Friedman 1968) and Edmund Phelps (Phelps 1968) forcefully argued that there was no reason why a permanent increase in inflation should lead to a permanent reduction in unemployment. The prediction was dramatically confirmed. Some combination of reduced monetary discipline when the global monetary system became completely untethered to gold, adverse supply shocks, and policy misjudgments led to the emergence in the 1970s of high rates of inflation. Inflation coincided with what, at the time, seemed abnormally elevated unemployment and low rates of GDP growth.What followed was a new consensus that held that demand management policy could not affect the average level of output over long periods of time. It could only effect output for limited intervals, primarily due to expectational errors and temporary wage and price rigidities. Inflation targeting became the monetary policy approach of choice around the world since policy could not affect average unemployment, but could affect average inflation. Central bank economists relied almost without exception on models that saw policy as impacting the second moment (variability) but not the first moment (level) of output and employment. Fiscal policy was discarded as a stabilization policy tool. Little wonder that Robert Lucas (Lucas 2003) was famously able to dismiss the importance of macroeconomic fluctuations.D. Financial Crises and Economic Thought: Primacy of the First MomentAll of this appeared reasonable until 2008. Cyclical fluctuations were sufficiently mild and downturns sufficiently brief that one could reasonably believe that whatever was lost on the downswings was made up on the upswings. It was the era of the “Great Moderation.” Even after the sharp downturn of 2008–2009, by far the most serious of the post–World War II period, it was possible to suppose that the economy had been taken down by a kind of “financial interconnection failure.” This view stipulated that a breakdown in the financial intermediation system might be expected to cause a sharp, but transitory, fall in output, much as would be experienced during a telephone outage or pervasive power failure and normality would return when the connections were repaired.Unfortunately, as figure 1 makes clear, despite the fact that the credit system was stabilized by the end of 2009, levels of output remain very depressed more than seven years after the trough of the recession. Another way to make the point is to observe that if the economy had performed as it was expected to before the financial crisis, US real GDP would be $2 trillion higher today, translating to nearly $25,000.00 of personal income for the average family of four. As shown in Table 1, the CBO economic forecast completed immediately prior to the Great Recession projected that real GDP would grow 24% by 2015. It actually managed only around 10%. This disappointment is less than what has happened in Japan over the last generation or what has happened over the same interval in Europe.Table 1. Projected Versus Actual Growth, 2007–2015 Real GDP in Year 2007 (billions)14,74×CBO Expected Growth in Real GDP 2007 to 2015*24.2%=Expected 2015 Real GDP18,477–Actual 2015 Real GDP16,397=2015 GDP Shortfall2,080%US Population (Billions)0.321=Per Capita Real GDP Shortfall6,479×Personal Income Share of GDO86.2%=Shortfall Per Capita (2009 Dollars)5,587×2015 Value of a 20009 Dollar (PCE Deflator)$1.10=Per Capita Shortfall in Current Dollars$6,120*. CBO Budget and Economic Outlook: FIscal Years 2008 to 2018, released Jan 2008, Table E-1, line 3Sources: CBO, Bureau of Economic Analysis, CensusView Table ImageRecent events are as severe a challenge to current orthodoxy as the Depression was to the orthodoxy of John Maynard Keynes’ times or the inflation of the 1970s was to the orthodoxy of its time. I submit that our current experience cannot be seriously contemplated using orthodox models that assume that macroeconomic policy choices affect the second, but not the first, moment of output.If I am right, there are large implications for contemporary macroeconomics. There is the necessary reorientation from the study of the amplitude of fluctuations in output to the presence and persistence of output gaps between what economies produce and their potential. At the methodological level, for three decades graduate students have been taught that their goal as macroeconomists was to understand the time-series properties of key variables like output, inflation, and interest rates. The universal tool of choice has been some kind of DSGE (dynamic stochastic general equilibrium) model. If instead most of what is macroeconomically important involves occasional events that have profound effects over long periods of time, linear time-series modeling may be of limited relevance. The desire for microfoundations, linearity, and simplicity leads the vast majority of DSGEs to build in the assumption that monetary policies do not affect average output levels over time. More fundamentally, theories that posit that economies are naturally self-equilibrating without policy intervention likely assume away the most important issues. It may well be that in 1933 or 2008, the economy would not have found a bottom if policy had not reacted dramatically to events.In what follows I shall try to suggest some ideas that lie outside the prevailing orthodoxy and may be helpful in making sense of recent developments. My position is in some sense intermediate between that of Blanchard (2016), who critiques existing macroeconomic approaches centered on DSGEs but comes down in favor of eclectic modification rather than revolution, and advocates of a new heterodox macroeconomics, who often bring a negative attitude toward existing approaches without fleshing out any alternatives.II. Secular StagnationA. SymptomsEven at the depths of the financial crisis, no one remotely expected that in 2016 interest rates would be close to zero, expected inflation would be far below target, and GDP would be at current levels. As evidenced by the rapid improvement in the stock market, shrinking of credit spreads, and increases in the prices of real estate and housing wealth, financial conditions improved much more rapidly than market participants expected. Yet, Table 2 shows that over the next 10 years, the financial markets expect real interest rates in the industrial world to be negative and inflation rates to be substantially below central banks’ 2% targets.Table 2. Ten Year Interest Rates and Expected Inflation USA*JapanEuroNominal Swap Rate1.050.080.02Inflation Swaps1.390.211.06Real Swap Rate−0.34−0.13−1.04*. Adjusted for the 0.35 percentage point avg. difference between cpi and pce.Source: Bloomberg, Updated August 28, 2016View Table ImageFigure 2 shows that current real interest rates are the culmination of a trend that has been underway for quite a long time. Notice that the current rate is roughly what would be expected from fitting a linear trend to the precrisis (1999–2007) period. This invites the question: are real interest rates so low right now because of the crisis or because there were already a set of forces underway that were depressing rates substantially?Fig. 2. US TIPS 10-year real yieldSource: Bloomberg.View Large ImageDownload PowerPointMy judgement is that there is substantial merit to the latter view. Indeed, one simplified but largely valid interpretation of the US macroeconomic history of the last several decades follows. The “recovery” has been a period of very slow growth and abnormally low real interest rates. This period was preceded by a financial crisis. That was, in turn, preceded by a period of normal growth, in which the unexciting growth in aggregate demand was accomplished only through the erosion of credit standards and a mammoth housing bubble. That, in turn, was preceded by the 2001 to 2003 recession, which, in turn, was preceded by the Internet bubble. So the last time that the economy grew satisfactory with financially normal and sustainable conditions was probably the mid- or late 1990s.B. CausesThe essence of the secular stagnation hypothesis is that there has been a substantial backward shift in the IS curve, which explains much of what we have observed over the last two decades (Summers 2014, 2015, 2016). As is demonstrated by Rachel and Smith (2015), the shift to the IS curve can be convincingly explained by demographic changes, the impact of inequality on savings propensities, events in the developing world, changes to the relative price of capital goods, a sludging up of financial intermediation, and an increase in risk aversion and concomitant risk premiums.The market expects the shift in the IS curve to persist for a long time. This can be seen in figure 3, which depicts expectations for Federal Reserve policy as manifested in the OIS (overnight indexed swap) forward curve. The market is saying that the long-run normal rate will be something like 2% indefinitely. There are questions about how one should adjust this for a variety of technical factors such as convexity. But there is not a plausible set of adjustments that would get the number to the 350 basis points that would have been considered on the low side of normal nine years ago, or even to the 300 basis points that the Fed is currently projecting.Fig. 3. Overnight indexed swap forward curveSource: Bloomberg (updated August 2016).View Large ImageDownload PowerPointFrom this perspective, much of what happened leading up to the financial crisis was not exogenous monetary and regulatory policy errors. It was authorities supporting a set of policies that were necessary to maintain something approaching full employment and reasonable growth given this large-scale leftward movement in the IS curve. In the face of reductions in demand, policymakers accepted the lower real rates necessary to assure full employment. These rates in turn had consequences for financial stability and set the stage for the crisis.C. Objections and ResponsesThe most common argument against the secular stagnation hypothesis is that the United States is at or near full employment, so cannot be afflicted with secular stagnation. Though the unemployment rate and most other labor market indicators do not suggest large amounts of slack, I nevertheless am more convinced of the validity of the secular stagnation thesis today than I was when I first put it forward in 2013. Growth has been much lower than was expected in 2013, despite interest rates being much more stimulative than market or central bank expectations. Central banks have been unable to raise even 10-year inflation expectations to target levels. Whether or not full employment is being supported by unsustainable financial conditions remains an open question. As I discuss below, the effective lower bound on interest rates will likely inhibit an appropriate monetary policy response when the next recession arrives. I do not see how anyone can prudently be confident that the demand side will not constrain economic growth in the years to come.Another common argument associated with Gordon (Gordon 2015) is that stagnation is a problem on the supply rather than demand side. I do not view this as a critique of, or mutually exclusive with, the idea of demand-side secular stagnation, and agree that the apparent collapse of productivity growth is a highly troubling phenomenon. Economists have a standard way of distinguishing changes in quantity that come from supply and demand shocks—looking at prices. Whereas the view that slow growth was coming from the supply side would suggest that prices should be higher than was anticipated in 2013, they are in fact substantially lower, as shown in Table 3.Table 3. Expected 2012–2018 Inflation* 2013NowCanada1.8%1.2%France1.6%0.9%Germany1.3%1.7%Italy1.4%0.9%Japan1.1%1.0%United Kingdom2.0%1.5%United States1.9%1.4%*. GDP deflator, annual rateSources: IMF WEO, Oct 2013 and April 2016View Table ImageOn the basis of the evidence laid out in Blanchard, Cerutti, and Summers (2015), I believe there is a substantial chance that productivity growth has been reduced by the shortfall in demand that we have already suffered. A corollary of this view is that if demand is increased, the result will be more rapid productivity growth. If weak demand inhibits the development of the economy’s supply potential, then the possibility of a low-growth trap caused by insufficient demand emerges.III. Over-Optimism and the Precrisis ParadigmThe failure of precrisis modes of thought, and concomitant necessity for a paradigmatic shift, is apparent from the economics profession’s repeated failure to predict what have turned out to be persistently disappointing outcomes. Experts, be they in the private sector, at central banks, or in the financial markets, have been consistently overoptimistic.A. ForecastsProfessional forecasters have been consistently wrong, as shown in figure 4, which depicts repeated downward revision to the IMF’s world growth forecasts. Nordhaus (1987) observed many years ago a tendency for revisions in economic forecasts to be serially correlated. That has been substantially borne out for both the world as a whole and for the United States. Table 4 shows the same pattern for the US economy, with continual negative revisions.Fig. 4. World GDP growth forecastsSource: IMF.View Large ImageDownload PowerPointTable 4. U.S. Year Ahead GDP Forecasts Prior JuneActualDiff20114.01.6−2.420123.52.2−1.320132.51.5−1.020143.32.4−0.920153.12.4−0.720162.6? Source: Federal Reserve Summary Economic ProjectionsView Table ImageThis over-optimism is quite broad. A similar pattern is present in figure 5, which shows analysts’ forecasts for companies, even though corporate profits on the whole have been a relatively bright spot; every two years analysts have been too optimistic about the next two years, and the moment when that has been truest is 2016.Fig. 5. Concensus bottom-up EPS estimatesSource: Financial Times, April 15, 2016; First Cell I/B/E/S, Goldman Sachs Global Investment ResearchView Large ImageDownload PowerPointB. Markets and Central BanksPerhaps what is most striking is how wrong the market has been about monetary policy. The market has consistently thought that normalization was coming. Each one of the dashed lines in figure 6 is the Fed Funds Futures, as of a particular moment in the past. And the market has very consistently been wrong. The only group that has been more wrong than the market has been the central banks. At every moment the Fed has seemed perplexed that the market did not believe its forecasts and did not recognize the degree and strength of their commitment to normalization. That continues to be the case today. Figure 7 depicts what Fed officials predict will happen and what the market expects. There are 19 dots, one for each Federal Reserve Governor and bank president. None of them are anywhere near what the market is saying.1Fig. 6. Forward OIS curveSource: Bloomberg.View Large ImageDownload PowerPointFig. 7. FOMC versus market expectationsSource: Federal Reserve; Bloomberg (updated to August 28, 2016 data).View Large ImageDownload PowerPointIf the Fed was determined to keep interest rates low and announcing that interest rates were going to stay low, and the markets were convinced that the Fed was well behind the curve and would have to raise interest rates much more rapidly, there would be an outcry from the markets and the economics community. But the current situation is largely passing without comment, despite the magnitude of the divergence and its persistence.IV. Policy OptionsJudging by the history of cyclical fluctuations in the United States and abroad as reflected in Table 5, the odds that there will be a recession within the next three years are significantly better than 50%. Conventional monetary policy, which as discussed above has been dominant in the demand-stabilization role since the stagflation of the 1970s, is challenged by extremely low interest rates. Conversely, low rates make government debt less costly. For both these reasons—the limitations on monetary policy and the relaxation of the intertemporal budget constraint implicit in low rates—I expect fiscal policy to play an enhanced role in the future.Table 5. Three+ Year Old Expansions Percent of Time Recession Within 2 years3 Years5 YearsJapan30%40%54%Germany53%74%98%UK28%40%63%US43%63%88%*. 1970–presentSources: NBER, Economic Cycle Research InstituteView Table ImageA. Monetary Ingenuity and Its LimitsThere is a substantial aggregate demand problem that is manifest in below-target inflation everywhere, in substantial slack in Europe and Japan, and in the prospect that the United States will, at some point in the not-too-distant-future, have slack that is not readily solvable. To counteract postwar recessions, the Fed has generally cut interest rates by at least 400 basis points. The prospect that there will be 400 basis points worth of monetary fuel to provide to the economy has to look quite remote, even considering that rates can be taken somewhat negative and risk premiums reduced through quantitative easing.How might monetary policy respond to all of this? One view taken by some central bankers is that if equilibrium interest rates are low, then we have to find a way to get interest rates down to equilibrium levels, and, so, you need monetary ingenuity. I think that is a reasonable point of view. It is constrained by real questions as to how much incremental monetary expansion is possible through available tools. Whether there is any political openness to the institutional reforms that would make substantially negative interest rates possible is a question that is very much in doubt. Forward guidance is another possible tool. But how effective can forward guidance really be when the futures markets are already judging that rates are going to be very close to zero? And even if low enough interest rates could be achieved, other concerns remain. It seems to me highly plausible, on the basis of all sorts of behavior that we observe, that abnormally low interest rates and extensive liquidity promote leveraging, promote reaching for yield, make bubbles more likely, and therefore run substantial risks of subsequent financial instability. Even if that is not the case, the investments that were not worthwhile at a 0% 10-year real yield, that suddenly become worthwhile at a significantly negative yield, are likely to be of low quality.Another possibility is quantitative easing policy, or in Fed parlance, “large-scale asset purchases” (LSAP). Here, too, there are reasons to doubt efficacy. Whatever one thought about the impact of LSAP on interest rates a few years ago, the observation that long rates have fallen sharply and yield spreads between short and long rates have fallen even more rapidly since QE was suspended in 2013 has to give pause. There is also the question of whether policies that work through putting price pressure on long-term Treasury bonds impact the longer-term rates that are most important for investment decisions. Finally, advocates of LSAPs have to contend with the observation made by Greenwood et al. (2014) about the complexities posed by the distinct role of the Federal Reserve and Treasury in these markets. The period when QE was thought to have been effective, the stock of long-term public debt that markets had to absorb increased sharply because of government deficits and the Treasury’s shift toward issuing more long-term debt.I am not sure what the future will bring with respect to monetary policy. The current paradigm of independent central banks politically accountable for the attainment of inflation targets follows naturally from New Keynesian models in which monetary policy does not have long-run effects on output and in which the dynamic consistency problem leads to a bias toward excessive inflation. In today’s world, where markets are indicating that insufficient rather than excessive inflation is likely to be the problem for the next decade, and where chronic output shortfalls are a real risk, I see no compelling argument for the current paradigm. Ideas that focus on credibly signaling a commitment to enough inflation to allow monetary policy flexibility, that involve taking more responsibility for avoiding output shortfalls, and that are more sensitive to financial conditions should, I predict, supersede inflation targeting over the next decade.B. Fiscal Policy and Lower RatesI believe that doubts about the efficacy of very low interest rates in stimulating demand, along with concerns about the possible adverse side effects, will eventually lead to substantially increased reliance on fiscal policy as a stabilization tool. Fiscal policy operates more directly and responds to the current market signals that bonds are extraordinarily high-priced, which suggests that more of them should be sold. Put differently, the nature of the government budget constraint is fundamentally different in an environment of zero real interest rates that can be locked in for long periods of time. And there are strong microeconomic arguments for believing that infrastructure investments will have a high social rate of return, at least in the United States.Many who might otherwise be in favor of fiscal activism worry that given current high levels of government debt, it may be infeasible or imprudent to rely on fiscal expansion as a stabilization tool. This is a legitimate concern and those who think it can be avoided by relying on helicopter money (and money financed fiscal expansion) are confused. Base money is now paid interest, so is best thought of as a form of floating-rate government debt. Moreover, because there exists today no preset path for monetary policy going forward, there is no way of meaningfully committing that a given fiscal expansion will be indefinitely financed with money.However, there are three other compelling responses to concerns about excessive borrowing. First, much incremental borrowing can be for outlays that would otherwise take place with a delay. There is a phenomenon in many countries that I term “repressed deficits,” which is analogous to “repressed inflation.” Repressed inflation is when a country would have inflation, but has imposed a price control. Repressed deficits take place when a government would have a budget deficit but instead defers maintenance, underpays the civil service leading to higher turnover, or does not make adequate contributions to pension plans. These actions all have the effect of pushing a liability into the future that will likely compound at a rate substantially greater than the zero real interest rate available on government debt. Increasing deficits in order to pull forward necessary investments would likely improve the government’s long-run financial position.Second, as Delong and I (DeLong and Summers 2012) demonstrate, in the current low-interest rate environment, fiscal stimulus in the form of public investment does not have to be hugely productive to pay for itself. This finding has been corroborated by the IMF (2014). Imagine a project with only a 5% return taking account of both any stimulative aspects and the output benefits of increased public capital. For each $100 invested, GDP will rise by $5 leading to about $1.50 in increased revenue collections. Leaving aside any hysteresis considerations that would just strengthen the case, this $1.50 is more than enough to cover the debt service on long-term inflation-linked bonds.Third, it is possible for fiscal policy to be expansionary and raise demand and neutral interest rates without increasing the deficit at all. Consider the possibility that the government expands pay-as-you-go social security—that is, a system where each year the young generation pay for the old age benefits of their parents’ generation. Such a system by the definition of pay-as-you-go does not impact the federal deficit. Yet its effect is to raise demand by obviating the need for as much private retirement saving as would otherwise take place. A similar impact can be achieved by deciding at present to rely more on tax finance and less on benefit cuts in dealing with funding gaps in existing entitlement programs.V. Future ResearchA. Limiting the Incidence and Da