VoxEU Column Financial Markets Global crisis Monetary Policy

Secular stagnation: A review of the issues

Pessimists have been predicting slowing rates of invention and innovation for centuries, and they have been consistently wrong. This chapter argues that if the US does experience secular stagnation over the next decade or two, it will be self-inflicted. The US must address its infrastructure, education, and training needs. Moreover, it must support aggregate demand to repair the damage caused by the Great Recession and bring the long-term unemployed back into the labour market.

When late last year, former US treasury secretary Lawrence Summers suggested that the advanced economies, and the US in particular, might be suffering from “secular stagnation,” his remarks struck a chord. The idea that the US and the other advanced economies might be suffering from more than the handover from a financial crisis resonated with many observers.

But while the term ‘secular stagnation’ was widely repeated, it was not widely understood. Secular stagnation, we have learned, is an economist’s Rorchach Test. It means different things to different people. In weighing the question of whether slow growth in the US and other advanced countries reflects some kind of ongoing stagnation problem, it’s important to be clear on the concept.

A first possible explanation for slow growth is that all the great inventions have been made. The respected Northwestern University macroeconomist Robert Gordon (2012) argues that electricity, the internal combustion engine and indoor plumbing were infinitely more important for boosting productivity and enhancing living standards than anything produced by the dot.com boom. Personal electronics may be great for playing games, but they are not so good for raising productivity. And there is no great invention equivalent to electricity or the internal combustion engine on the horizon.

For economic historians, this argument flies in the face of 200 years of experience. Pessimists have now been predicting slowing rates of invention and innovation for centuries, and they have been consistently wrong. Looking ahead, it seems clear that the productive potential of robotics and the human genome, for example, have only begun to be realised. Evidence that we are learning how to use intelligent machines to replace first unskilled and eventually skilled labour suggests that we have a distribution problem, not a growth problem.

Gordon bases his argument on the fact that productivity growth, in the US in particular, has been slower in the four decades since the early 1970s than in the century (or a bit less) that preceded them. Only the decade from 1995 to 2005, when, paraphrasing Robert Solow (1987), “you could see the computer age everywhere…. [including] in the productivity data,” did the rate of total factor productivity growth rival what had been achieved over the prior century.

This conclusion, in my view, ignores the fact that the preceding century, Gordon’s golden age of technical progress, also saw periods of slow productivity growth, notably when new network technologies were being rolled out but the economy had not yet adapted to their availability. (The period of electrification starting in the 1890s stands as a classic case in point.) This is an argument for not making too much of the slowdown prior to 1995, when adaptation to the availability of computers and the internet first got underway, or of slow productivity growth now, when we are potentially on the eve of a robotics and human genome revolution.

A second version of secular stagnation argument holds that we have a problem of stagnant aggregate demand – that households are not spending enough and firms are not investing enough even at near-zero interest rates. Those with very high incomes have a relatively low propensity to consume, and virtually all the income gains in the US have gone to those with very high incomes. (There’s that pesky problem of income distribution again.) The result is a glut of savings that firms are unable to invest at a positive interest rate. The advanced countries therefore find themselves with extraordinarily low interest rates as this glut of savings floods the market, and yet with not enough investment to absorb it or to sustain a respectable rate of growth.

This appears to be the version of the argument that Professor Summers prefers. I have my doubts. What matters for interest rates is not US saving but global saving, since funds in the 21st century can move across borders. And, in fact, global saving has basically held stable for the last decade and a half at 23 to 24% of global GDP. At most, global saving rates have risen only modestly. And looking ahead, with China rebalancing its economy toward consumption, there is every reason to think that the global saving rate will come down.

One can of course argue that the ratio of global savings to global GDP is determined in general equilibrium – that the numerator as well as the denominator is an endogenous variable. But that only reinforces the point. In a situation of near-zero interest rates and deficient demand (the current situation and the one that secular stagnationists foresee for the medium term), an increase in savings will have a negative impact on GDP. That effect will be unusually strong, because interest rates can’t fall further and because the global economy is closed to trade, increasing the size of the multiplier.

In this situation, an increase in global savings will have a sharp negative impact on global GDP. Imagine, for example, an aggregate-demand multiplier of two. When savings rise by, say, 1% of initial GDP, that GDP then declines by 2%, and the resulting saving rate, as measured, is more than 1% greater than before. But this means that the change in the ex ante savings rate (the shock ostensibly responsible for secular stagnation) is smaller than the change in the ex post savings rate that is actually observed. And the increase in the savings rate we observe in the data, as already noted, is small (globally, 1-2% of GDP).

Even taking general equilibrium effects into account, then, changes in global savings do not appear to be a major factor in the persistently low level of interest rates and slow rate of economic growth.

A third version of the argument suggests that output and total factor productivity growth are stagnant because of the failure of countries like the US to invest in infrastructure, education and training. I have considerable sympathy for this view, given how nondefence, non-entitlement federal government spending, which is devoted heavily to infrastructure, education and training, has been cut to the bone. The empirical literatures on infrastructure, education and economic growth are less than fully conclusive. Intuitively we know that there is something here; we just don’t know how much.

A fourth and final version of the secular stagnation hypothesis argues that the US economy’s supply-side potential has been permanently reduced by the Great Recession and the slow recovery that followed. The failure of output growth to recover to the pre-Great Recession trend, instead moving in parallel with this trend at persistently lower levels, is consistent with this view. The mechanism in question is then high unemployment, which has permanently impaired the productive potential of the labour force through forgone on-the-job training and the atrophy of skills.

High long-term unemployment and a large number of discouraged workers are prominent features of the current recovery. The question is whether the damage to human capital accumulation due to being out of work is permanent or temporary – whether the effects are reversed easily or only with difficulty. This debate is raging within the corridors of the Federal Reserve System as we speak. The question there is whether many of the long-term unemployed have become essentially unemployable, in which case their being out of work does little to moderate upward pressure on wages, making the headline unemployment rate the best measure of slack in the labour market.

Here, neither contemporary nor historical evidence is definitive. Nick Crafts’ studies of long-term unemployment in Britain in the 1930s (e.g. Crafts 1989) confirm that the long-term unemployed had little impact on the behaviour of wages, as if those out of work for extended periods became effectively detached from the labour market. On the other hand, studies of the US during World War II by none other than Robert Gordon show that a positive labour-demand shock, if sufficiently strong, can draw the long-term unemployed back into work and quickly reduce the natural rate of unemployment to earlier low levels (Gordon and Krenn 2010).

So is there a secular stagnation problem? Yes, there are reasons to worry that the US’s growth rate over the next 10 or 20 years will disappoint by the standards of the 20th century. But this is not inevitable. It will not be because all the great inventions have been made or because there is a dearth of attractive investment projects and an overabundance of savings.

If the US experiences secular stagnation, the condition will be self-inflicted. It will reflect the country’s failure to address its infrastructure, education and training needs. It will reflect its failure to take steps to repair the damage caused by the Great Recession and support aggregate demand in an effort to bring the long-term unemployed back into the labour market. These are concrete policy problems with concrete policy solutions. It is important not to accept secular stagnation, but instead to take steps to avoid it.


Crafts, N (1989), “Long Term Unemployment and the Wage Equation in Britain, 1925-1939”, Economica 56(222): 247-254.

Gordon, R and R Krenn (2010), “The End of the Great Depression 1939-31: Fiscal Multipliers, Capacity Constraints and Policy Contributions”, NBER Working Paper No. 16380.

Gordon, R (2012), “Is US Economic Growth Over? Faltering Innovation Confronts the Six Headwinds”, NBER Working Paper No. 18315.

Solow, R (1987), “We’d Better Watch Out”, The New York Times Book Review, 12 July.


4,934 Reads