The National Association for Business Economics recently published Larry Summers’ remarks on secular stagnation earlier this year. The conversation picked up after his speech at the IMF conference last year, though has been floating around for some time. Excluding gloomy prognostications from Alvin Hanson in the 1930s, I can trace back the idea that the US has fallen into a secular trap back to Paul Krugman at the Munk Debates in 2011.
Though Summers is seen today as the ringleader of the stagnationistas, Krugman’s formidable opponent was none other than Larry Summers. Summers has since undergone, as I’ve documented before, a remarkable intellectual evolution making very similar claims to Krugman in 2011 (though with more rigor and a stronger platform). While even the chance of a somewhat permanent slump is sobering, there are reasons to doubt both the theory and evidence behind this claim.
Let’s start with the labor force. Summers is understandably unhappy about the recent decline in participation rate. He controls for demographic factors by looking only at working age (25-54 yo.) men which should account for an older population and students staying in school longer:
But that is not the largest part of the story. Even if one looks at 25-to-54 year-old men, a group where there is perhaps the least ambiguity because there is the greatest societal expectation of work, Figure 3 shows that the employment/population ratio declined sharply during the downturn, and only a small portion of that decrease has been recovered since that time. The recovery has not represented a return to potential; and, according to the best estimates we have, the downturn has cast a substantial shadow on the economy’s future potential.
With the graph:
While there is certainly a sense in which this has not recovered to trend level, there is also an argument that the crash just accelerated an already fermenting labor force. While Summers argues this is due to features of the recovery, and not structural or technological issues, the data may suggest otherwise:
For example, the graph above demonstrates that the rough participation rate of those with a college degree or higher fluctuates within a very narrow band and has, if anything, continued on an increasing trend. This militates in favor of a skills-biased technological change hypothesis which is inconsistent with secular stagnation (of output at least).
The evidence is stronger still. Take these two charts from JP Morgan:
Discouraged workers are peripheral to the question of labor force participation. Nor is “working age men” the representative statistic it once was, as Summers analysis makes it out to be. If anything, females are obtaining college degrees at higher rates and are employable in a wide array of increasingly-important occupations, prominently healthcare and education. Traditionally male fields, on the other hand, are in largely tradable sectors and probably not coming back. This has worrisome implications for the distribution of income and the underlying social structure itself. And we should do something about that. But robots and cheap, Chinese goods are not “secular stagnation”.
This brings us to our next point:
Third, changes in the distribution of income, both between labor income and capital income and between those with more wealth and those with less, have operated to raise the propensity to save, as have increases in corporate-retained earnings. These phe- nomena are shown in Figures 9 and 10. An increase in inequality and the capital income share operate to increase the level of savings. Reduced investment demand and increased propensity to save operate in the direction of a lower equilibrium real interest rate.
This sounds good in theory, but in reality the very period over which income inequality has increased the most, total savings rates have fallen precipitously:
Gross private saving is the broadest measure of saving and accounts for increased savings in the form of corporate profits. Government saving, to nobody’s surprise, has also fallen over this period. While richer people most certainly do have a lower propensity to consume, the magnitude of this effect is clearly dwarfed by forces like the Asian saving glut and potentially overvalued dollar. Luxury markets are booming: and it may not be to a good liberal’s taste, but demand for chauffeurs and butlers creates employment just like that for apparel and electronics. Indeed, to the extent richer people consume at the margin on non-tradable services, the leakage is also lower.
But Summers’ grander point here remains: real interest rates have been on a secular decline. This strikes at what may be the heart of the problem: low investment demand. While low equilibrium interest rates have other problems, like increasing the risk of a liquidity trap, the structural impact isn’t necessarily felt in investment. The interest-rate elasticity of investment has always been tough to measure and Treasury yields may not be the best gauge: lets look at the corporate bond market once yields start to diverge. This isn’t too important (yet).
Rather, advanced economies currently function on what Paul Krugman has called a “bicycle model”. Firms invest on the expectation of higher returns in the future due to economic growth. The accelerator model of output, therefore, frames investment demand as a function of long-run potential growth. When this expectation falls, so does investment. In other words growth is necessary for growth. If long-run growth expectations fall substantially the US is by the principle of recursion screwed. But the bicycle doesn’t have to fall if we transition to a steadier, consumption driven model. And that’s just what we’ve been doing:
So long as we’re able to keep consumption growth to meet a lower steady-state investment rate, the economy won’t self-implode (though, definition, will grow slower). Given that the largest projected increases in future private and public spending come from education, healthcare, and a geriatric economy this transition will be rather easy for the United States to make. (Not to mention the ample space for public investment, something I know Summers is on board with).
However, we should still be uncertain about the extent to which potential growth has fallen. As Summers notes:
Ponder that the leading technological companies of this age—I think, for example, of Apple and Google— find themselves swimming in cash and facing the challenge of what to do with a very large cash hoard. Ponder the fact that WhatsApp has a greater market value than Sony, with next to no capital investment required to achieve it. Ponder the fact that it used to require tens of millions of dollars to start a significant new venture, and significant new ventures today are seeded with hundreds of thousands of dollars. All of this means reduced demand for investment, with con- sequences for equilibrium levels of interest rates.
It is valuable here to think in terms of inputs and outputs. We need less “debt” to achieve a similar valuation that once required tons of borrowing. What Summers is describing is technological, if not economic, progress. Indeed the interest-rate elasticity of demand has probably fallen, but only in a small segment of the economy. I’ve written about this in some detail before:
While he’s right that there are secular trends toward cheaper capital, I’m not sure we can attribute much of that to the information technology era. Even granting that the magnitude of technology is substantial, something I’ll get back to in a minute, it’s important to explore the consequences. I would argue that the information era doesn’t reduce the demand for investment, as Summers proposes, as much as the interest rate elasticity of demand – that is to say firms are less sensitive to changes in the cost of capital. The reason isn’t obvious. Interest rates matter a lot for long-term, capacity building investments—like opening a huge new factory. But software is usually rented on a monthly basis from the firm’s own cash—so interest rates matter a lot less. (Reduced demand and reduced need mean the same thing in English, but not economese).
But that’s coming from the same cash pile Summers is talking about—and it’s still investment. In general it’s not that there’s a lot less investment as much as the factors that affect its demand are changing. This is where we get back to magnitudes. Take a look at software’s share of total private investment:
There are two stories here. One, in sheer magnitude, software (the blue line) is still only 15 percent of private investment and not significantly higher than points in the past two decades when interest rates were a lot higher. On the other hand, residential investment as a share of private investment, hasn’t changed much in structure since the mid-’60s and is still very sensitive to changes in the interest rate.
The point here is that while WhatsApp didn’t need any real investment, a lot of the economy does, and as sexy as Silicon Valley is, main street America is not irrelevant. There’s another point here, and I’d file it under, as Scott Sumner might say, “never reason from a price change.” The Silicon Valley story tells us that the returns on very small investments can be huge and, under competitive markets, this would imply that the marginal return on capital is falling rapidly after a certain level. However, that isn’t consistent with the broad increase in capital’s share of income we’ve noticed over the past decade or so. In a simple (but empirically powerful) Cobb-Douglas production function, the exponent on capital is its share of total income, and the higher the exponent, the larger the marginal product at any given point (other things equal).
Of course, other things are never equal, and that’s Sumner’s point.
Summers is right, that to the extent software is significant, there are important implications for the equilibrium interest rate. As the elasticity falls (that is, as the demand curve becomes more vertical) changes in the supply of loanable funds will be felt entirely through interest rate adjustment. So an increase in supply has a lot more potential to keep the economy under a low interest rate environment than before.
Ultimately, this is all pretty speculative. Calculating the importance of interest rates, as an empirical phenomenon, is notoriously difficult (studies disagree, for example, on whether higher interest rates even increase savings)—and observing shifts in the shape of the determining curves is harder still. While interest rates are low for a number of reasons, technology per se may not be one.
The practical point here is that since long-term interest rates are what drive the kind of investment Summers says is dwindling, it’s ever more incumbent that the Fed doesn’t let the term premium on long bonds rise too much. That is, keep quantitative easing hard.
Lower interest rates and interest-rate elasticity of demand pose a real threat to the current model of monetary policymaking. These are good arguments for a slightly higher inflation target.
But maybe the most important problem with secular stagnation is the cyclical recovery, which to some extent are mutually exclusive. While the economy is still really tough for those at the bottom, there are signs of life with increasing quits, inflation, and pressure in the market in general. There’s a lot more we can and should be doing, but these should perhaps be considered for redistributive and supply-side – rather than purely demand-side – reasons.
Of course, until we do get consistently sufficient inflation, demand is still lacking. The question is whether that seems impossible, and I think it would be difficult to answer in the affirmative. Unfortunately, my post has none of the optimism with which Larry Summers and Ian Bremmer challenged Paul Krugman in the Munk Debate. The Depression was the last time we were in such a bad position. As horrific as that was, it was no secular stagnation.