Goldman Sachs: "Welcome To The Great Stagnation"


A little under a year ago, with much pomp and circumstance, Goldman’s economic team proclaimed its multi-year long bearish outlook on the economy over, and on December 1 of 2010, issued a report in which it noted that it was turning a new leaf in its “bleak” perception of the economy, based in big part on its expectation that the combination of an ebullient monetary environment (QE2 has just been launched) and a cornucopic (sic) fiscal stimulus (the first, of many, payroll tax cuts had just been passed) would lead the economy to a sustained growth of not less than 4% (and led Zero Hedge to officially proclaim Goldman as having jumped the shark in conjunction with our prediction that a year hence the US economy would be contracting yet again). Zero Hedge was right, and Goldman was 100% wrong. Today, we however witness something different: in what seem to be a major paradigm shift within the firm’s strategic research team (not Jan Hatzius’ group but that of Dominic Wilson), the firm appears to officially give up on “recovery” as a modal outcome for both the US and the developed world, and instead aims a little lower. Far lower. The report is titled “From the 'Great Recession' to the 'Great Stagnation'” and in an extended analysis looking at 150 years of historical data, it concludes that “those historical lessons suggest that the probability of stagnation in the current environment is much higher than usual, at about 40% for developed markets. Trends in Europe and the US are so far still following growth paths that would be typical of stagnations.” Looks like Goldman just proved us at least half right when we said in January that the keyword of 2011 would be “stagflation.” Luckily, the Fed and the world’s central banks still have 3 months in which to prove the second half of our prediction correct as well.

Sure enough, such an outright capitulation would be quite uncharacteristic for Goldman, which does provide a last ditch deus ex machina: “whether these countries manage to avoid a ‘Great Stagnation’ by a pick-up in the recovery is likely to depend on policy being able to restore confidence and putting in place reforms that can decisively jolt growth.” We would say it otherwise, all hope abandon ye who put your hope in politicians. Goldman just proclaimed, in not so many words, that the great western world growth fairy tale is now over.

So without further ado, here is…

From the ‘Great Recession’ to the ‘Great Stagnation’, published September 29, 2011

Economies can either grow, contract, or stagnate. Since 2009, most concerns have centered on the second of those outcomes, and the prospect of a fresh recession. But with two years of positive but sub-par growth and cyclical sectors already depressed, it is now becoming more common to hear worries that we are heading not for a sharp renewed contraction but instead for a prolonged period of sluggish or stagnant growth. The poster child for this kind of outcome is Japan’s experience in the 1990s. And comparisons of the current experience with Japan have been rife for some time from policy makers, academics and market analysts alike.

How likely is a prolonged stagnation? The problem with answering this question through a focus on Japan is that it is just one episode, and quite a distinctive one. Relying on that to make an assessment of the current risks is an unnecessarily narrow prism through which to look at the current environment. We have already shown that the history of housing busts in the OECD since 1970 more generally is associated with long periods of sub-par growth. But the range of experiences and outcomes here too is quite wide.

Here we ask the question the other way round, focusing more directly on the risks of a prolonged (decade-long) stagnation. To do that, we crunch the numbers through 150 years of macroeconomic history. Taking such a broad sweep limits the kinds of analysis that can be done given the constraints of data. But it allows a much wider look at the conditions and features of periods of stagnant growth, with the following results:

  • Although the Japanese experience of the late 1990s has received a lot of attention, we find about 20 more episodes of that kind under a range of definitions, mostly in the post-war developed world. Japan may or may not be a paradigm but it is certainly not unique.
  • Per capita GDP growth during these episodes hovers steadily below 1% (at 0.5% on average). CPI inflation tends to be low and steady, signalling aggregate demand pullbacks. Unemployment is high and sticky. Housing prices tend to fall, especially in comparison to their pre-stagnation dynamism, while stock-market returns are far weaker than their historical averages.
  • These events tend to be correlated with, if not preceded by, financial crises, especially stock-market crashes. Put simply, the probability of stagnation is much higher after financial crises.

Those historical lessons also suggest that the probability of stagnation in the current environment is much higher than usual, at about 40% for developed markets. Trends in Europe and the US are so far still following growth paths that would be typical of stagnations. Given those risks, whether these countries manage to avoid a ‘Great Stagnation’ by a pick-up in the recovery is likely to depend on policy being able to restore confidence and putting in place reforms that can decisively jolt growth.

Stagnations Are More Common Than You Would Think….

Japan, you are not alone. From an analytical perspective, this is fortunate. Because episodes of stagnation are more common than is sometimes assumed, we can use information from a wider range of experiences to understand what the features of periods of stagnation normally are, and to assess the probability of stagnations occurring now.

To do that, we first need to define the kind of experience we want to capture from the historical record. We define episodes of stagnation as long-lasting periods of sub-par GDP per capita growth that are not interrupted by either sharp contractions or a return to mean growth. In practice, we devised an algorithm that filters growth through ‘stagnation bands’ (more details on these bands are provided in the box on the next page), accounting for differences in income levels and slow shifts in the patterns of growth across countries. We then define an episode of mild stagnation as one in which countries stay between an upper and lower bound for more than six years, and an episode of serious stagnation, our main focus here, as one that lasts more than 10 years. Sharp new recessions or significant recoveries are thus excluded from these definitions. In practice, any particular definition will have some strengths and weaknesses, but some plausible alternatives yield broadly similar conclusions.

We then apply this definition to a large data set, covering dozens of countries—both developed markets (DM) and emerging markets (EM)—since the late 1800s. Doing this, we find 93 episodes of stagnation, which Chart 1 splits into three classes. Of that universe, 24% have lasted more than 10 years, 60% have occurred in the post-WWII period and 67% have occurred in an OECD country. The major conclusions are that:

  • There are quite a large number of stagnation episodes in history.
  • The past 60 years have witnessed most of these episodes.
  • Surprisingly, the developed markets—especially Western Europe—have historically been more vulnerable to stagnation than their EM colleagues, even after accounting for differences in the volatility and levels of growth.

Japan’s identified experience (spanning 1992-2003) is representative but not unique. In comparison to these episodes, it is neither the longest-lasting nor the most severe. Table 1 shows the ranking of the top 10 stagnation experiences in these two categories. The longest-lasting experience was that of India, which ran for nearly two decades after the early 1930s. Japan’s experience takes ninth position, with a duration of 12 years. And the shortfall of growth from ‘normal’ does not place Japan in the 10 ‘most stagnant’ stagnations. This suggests it may be useful to broaden the focus to include this larger group of countries.

…More Sluggish than You Would Expect...

By construction, a stagnation period involves sluggish growth. But it is helpful to look at the features that characterise past stagnations in more detail. Chart 2 summarises that comparison for all stagnations: Japan’s experience, serious stagnations and post-WWII stagnations. Although data limits a full description of these episodes, it is straightforward to examine the behaviour of GDP growth, the CPI and stock-market returns over that period (both means and standard deviations). The key features of the ‘typical’ stagnation are as follows:

  • Lower and less volatile growth. Looking at real GDP per capita growth, serious stagnations were periods of growth averaging around 0.5% (Japan’s average of 0.6% was typical). These values are well below the historical averages for the full and post-WWII samples—at 2.0% and 2.8%, respectively. Growth was also substantially less volatile during stagnations (with a standard deviation of about 1%, well below the post-WWII average of 4.0%). These differences are costly. If a country grew at the mean stagnation rate over a period of 10 years, it would end up with a level of income more than 20% lower than it would have been had it grown at the post-WWII average.
  • Low inflation. CPI inflation tells a similar story. Japan’s case is extreme: it recorded an average 0.2% inflation rate during its stagnation period. The average for the other long-lasting stagnations is closer to 5% (median of 4.1%). Of course, these other episodes cover a wide range of time periods when inflation globally was much higher. But inflation clearly tends to be much lower and flatter during stagnations than during other comparable periods. Among other dynamics, this reflects demand pullbacks that may be difficult to correct if they persist for a sufficiently long period of time. It also underscores the importance of policy efforts focused on averting deflation.
  • Rising and sticky unemployment. Although data is more limited, for a subset of stagnation episodes, we found that unemployment rises during stagnations. Both for OECD and non-OECD countries, the unemployment rate generally rises by close to 3ppt on average compared with the three years before stagnation (from 4.6% to 7.3% for the OECD and from 8.1% to 10.9% for the Non-OECD sample).
  • Stagnant home prices. For a subset of stagnation episodes for which data is available, we found that housing price inflation tends to fall from an average of around 2% (1970-2010, across 20 OECD countries) to -0.1% during stagnation. Thus, stagnation periods are typically characterised by housing busts or, at least, by important housing price corrections.
  • Lower stock returns. Average returns in stagnation episodes are modestly positive (around 5% annual returns on average). But they are still below the relevant historical averages. And, without a proportionate fall in volatility, Sharpe ratios tend to be lower than average too. In terms of equity returns, the Japanese experience was significantly worse than the average (with a negative real total return of -3%). In that sense, the broader experience suggests that this may not be entirely representative.

It is also useful to map out the path and not just the averages of stagnation episodes. Chart 3 shows the evolution of a typical stagnation path, starting from 5 years prior to the beginning of an episode. The grey areas represent +/- 1 standard deviations from the average. We highlight three features:

  • The typical stagnation entails a noticeable flattening of the growth trajectory within relatively tight bands.
  • The Japanese experience follows the typical stagnation pattern closely but also shows that stagnations can be worse (in the sense of an even more flattened pattern).
  • The chart illustrates two contrasting examples that are at the core of why we care about these historical experiences: EMs and DMs, anchored at 2008, and projected to 2012 using our own forecasts. While the EM aggregate is on a growth path that could not be further from stagnation, DMs are well within its confidence bands. Both Europe and the US are remarkably close to the typical stagnation trajectory. Unless growth picks up substantially, in the next couple of years, we are likely to have mapped out a 5-year period of stagnant growth.

... And Now More Likely than You Would Like

This raises the question of how likely it is that we enter (or have entered) a period of prolonged stagnation now. The good news is that, in normal circumstances, the probability that a country will fall into stagnation in any given year is relatively low, at 2%. (The probability per time span—that is, any 6-, 8- or 10-year period chosen at random—is 16%, 10% and 6%, respectively.) But that is the unconditional probability. What we want to understand is the conditions that make the kinds of stagnations we have identified more likely. The bad news is that the major developed economies have not been experiencing normal circumstances in the past few years. Moreover, the ways in which the current episode is unusual make the probabilities of stagnation higher than normal too.

Defining the preconditions of stagnation across history is limited by the kinds of data that are available. But a lot can still be said. Our first approach is to match the relative frequency of stagnation episodes and periods of crisis. Our menu of crises depends on data availability, but it is sufficiently rich to give helpful answers. Chart 4 shows that stagnation periods tend to be either preceded or coincident with stock-market crashes more than 50% of the time. They are also related to currency crises and to a lesser extent to banking crises. The correlation with macroeconomic disasters (defined as 10% cumulative contractions in GDP per capita) is lower, and the connection to inflation crises or war is weak. The main signal we extract from this exercise is that market crashes, precisely of the type observed during 2008-2009, are highly informative about the prospects of economic stagnation.

To look at these factors together more formally, we have estimated a probability model that allows us also to generate forecasts (the details on the model are provided in the box on the bottom of this page). These models largely confirm the intuition from the simpler exercise. Stock-market crashes, currency crises, external debt crises and preceding periods of bad growth tend to raise the probability, and probabilities of stagnation are also higher in general for richer countries. That general story is consistent with—though broader than—the work we have done showing that growth recoveries are generally weaker than normal after major housing and banking busts.

We can use these results to see what the model says about the probabilities that a given country has entered or will enter a period of stagnation now. Chart 5 presents the results from our probability model applied to stagnation periods lasting for at least 8 years. While the differences between individual countries and the exact rankings should not be leaned on too heavily, overall the results are striking. In particular, the chances that EM economies will stagnate are much lower than for DMs. For example, all the BRICs are in the lower ranges of our probability estimates, with China, Russia and India at the bottom of the list. In contrast, developed economies populate the right-hand side of the scale, with probabilities in Europe and the US of around 40%. While this suggests that a long period of stagnation is far from a certainty, these probabilities are uncomfortably high relative to normal experience.

More Policy Action Likely to be Needed

History suggests that the prospect of a long period of stagnant growth is a plausible risk and a legitimate concern for the major developed economies. So what can be done to avoid it?

Given the difficulty of finding reliable data over long histories, it is harder to be clear about what was necessary to end stagnations in the past, although we plan to focus more on that question in future research.

And, of course, stagnation—as we have defined it—can be resolved either positively (through recovery) or negatively (through a fresh sharp contraction).

In the current environment, with financial conditions still generally tightening and banking stresses increasing again, our main focus is on policy actions that might stabilise these dynamics. But, as we have highlighted before, constraints on policy—both central banks and beyond—are tighter than they were in 2008. And with monetary policy up against the zero bound, truly unconventional ‘unconventional’ monetary policies may become more central to the debate. Given those constraints, and the starting points for inflation and the unemployment rate, the damage from a prolonged period of stagnant growth will probably be significant—in line with the historical experience we have detailed here. The good news is that policymakers are more aware—thanks to Japan’s experience—of at least a part of that historical experience, if not all that we present here. The bad news is that it is still far from clear whether enough has been done to jolt economic growth upwards and outside the zone where prolonged stagnation is a serious risk.