Wednesday, December 31, 2014

Farewell Secular Stagnation

Marc Andreessen recently looked at the arguments for and against secular stagnation. He cited my Washington Post article when examining the case against secular stagnation. One of the points I make in it is that the proponents of secular stagnation incorrectly invoke the long decline of real interest rates as prima facie evidence for their view. Where they go wrong is that they look at real interest rates without accounting for the long decline in the risk premium. Once the risk premium is stripped out of their real interest measure there is no downward trend in real interest rates. Rather, you get a stationary risk-neutral real interest rate measure that averages close to 2%. This can be seen in the figures below, drawn from my follow-up article with Ramesh Ponnuru:

Interestingly, this risk-neutral measure closely tracks the business cycle and suggests it was the severity of the Great Recession and not secular stagnation that explains the low interest rates over the past six years:

Given this business cycle-driven relationship, the recent spate of good economic news points to rising interest rates in 2015. In fact, the daily measure of the risk-neutral nominal 1-year and 10-year interest rates from Adrian, Crump, and Moench (2013) show just that. See the figure below and note that 2014 has seen a trend change in the path of neutral interest rates. If this continues--and the improved economic news suggests it will--then the Fed will have to raising its interest rate in 2015.

I bring all of this up as a way to motivate a prediction for 2015: secular stagnation will fade from the national conversation for the U.S. economy. Instead, the conversation will focus even more on how to handle the advent of an increasingly digitized, automated economy where productivity growth is rapid and neutral interest rates are rising. Secular stagnation, in other words, is about to experience the same fate it had when it was first pushed in the 1930s. Here is what Ramesh Ponnuru and I wrote about that experience:
"The business cycle was par excellence the problem of the nineteenth century. But the main problem of our times, and particularly in the United States, is the problem of full employment. . . . Not until the problem of the full employment of our productive resources from the long-run, secular standpoint was upon us, were we compelled to give serious consideration to those factors and forces in our economy which tend to make business recoveries weak and anaemic and which tend to prolong and deepen the course of depressions. This is the essence of secular stagnation— sick recoveries which die in their infancy and depressions which feed on themselves and leave a hard and seemingly immovable core of unemployment.” 
Thus wrote Alvin Hansen, a professor of economics at Harvard, in 1938. Slow population growth and the deceleration of technological progress, he argued, was leading to slow capital formation and weak economic growth. A program of public expenditures, though it had its dangers, was probably required to avoid this fate. 
Hansen’s article was of course spectacularly wrong as a guide to the next few decades. Instead of suffering through stagnation we entered an extended, broad-based, and massive economic boom. In hindsight we can see that his analysis, while thoughtful and intelligent, was unduly influenced by the depression he was living through, and can see as well that the depression was the result of specific policy mistakes rather than inexorable trends. Recent research by Alexander J. Field shows that the 1930s were actually a time of exceptionally high productivity growth. 
Hansen’s worry, some of his specific arguments, and his phrase “secular stagnation” are all making a comeback in our own day. Lawrence Summers, like Hansen an economics professor at Harvard, has sounded an alarm about the ability of industrial countries to achieve adequate economic growth. A new e-book, Secular Stagnation, includes chapters by Summers and other leading economists discussing the question. 
The fact that Hansen was wrong does not prove that contemporary stagnationists are. In this case, though, history is repeating itself rather exactly. We do not pretend to know what the future path of economic growth in the United States will be. But the case for stagnation is weak—and, as in the 1930s, it is getting undue credence because of a long slump caused by policy mistakes.
Farewell secular stagnation. Hello the second machine age.


  1. "I bring all of this up as a way to motivate a prediction for 2015: secular stagnation will fade from the national conversation for the U.S. economy. Instead, the conversation will focus even more on how to handle the advent of an increasingly digitized, automated economy where productivity growth is rapid and neutral interest rates are rising. Secular stagnation, in other words, is about to experience the same fate it had when it was first pushed in the 1930s....."

  2. I don't think risk-adjusted rates are the right measure. Or rather, what you have provided is not an argument against secular stagnation but an alternative explanation for it. To my mind, the problem of secular stagnation is essentially the problem of equilibrium (non-risk-adjusted) nominal interest rates being too low, so that (since they vacillate over time with the business cycle, etc.) they often go below zero. One solution to the problem (no matter what the cause) would be to make the inflation rate sufficiently unpredictable that people wouldn't trust fixed income assets and would demand a higher risk premium on those assets -- which is to say, a lower relative risk premium on real capital assets. So perhaps we've had the underlying conditions for secular stagnation all along, but it has been masked by the fixed income risk premium (my proposed "solution" having been unwittingly applied by careless central bankers). That doesn't make it any less of a problem now.

  3. Andy, what is central to this debate is the natural interest rate, the rate determined by expected productivity growth, expected labor force growth, and household discount rates. These determinants are what is being debated: has trend productivity changed, has the labor force growth slowed down, etc. Larry Summers and Robert Gordon say yes, I see evidence to the contrary. You can't answer this question without looking at the natural rate which is the same thing as risk-adjusted real rate. Not accounting for trend changes in the risk premium obfuscates the picture.

    Note that everyone in this debate points to the long decline in real interest rates as evidence for secular stagnation. I am just building upon what they have done by showing that a proper reading of the fisher equation (unlike the simplified textbook one) also has a term for the risk premium. So it's not like my point is exactly novel or that controversial.

  4. Can you elaborate on this adjusted interest rate that you are calculating? You variously call it the "risk-free" and the "risk-neutral" rate, which are two different things. The risk-free rate usually means something like the Treasury rate, since the US government has negligible default risk in this context. There is the term premium, but of course that can be avoided by just using short-term rates, so that can't be what you mean by the risk premium. There are risk-neutral expected (i.e. forward) Treasury rates at different tenors implied from the yield curve, but that can't be what you mean either.

  5. Eric, what I am doing is taking the observed treasury yield and subtracting from it (1) expected inflation and (2) the risk premium. The risk premium is a catch-all category that here means the same thing as the term premium. For (1) I use the Phildelphia Fed's Survey of Forecasters and for (2) the risk premium measure from Adrian, Crump, and Moench (2013) referenced above. (You can download their data at teh link.) What this creates is an interest rate measure that lines up with what economist call the natural interest rate. Here is a post I did that uses this composition and spells it out more explicitly:

    I would note that even short-term treasuries are not truly risk free, even though they are considered the 'risk-free' benchmark asset in finance. There is a small default risk and term premium. The ACM data, for example, have risk premium estimates for 1-year treasuries. Even with 3-month treasury bill you still have these risks and that makes it difficult to use the observed treasury yield as a measure of the natural interest rate.

    Here is a post where I talk more about the natural interest rate:

    Hope that helps.

  6. Thanks for the extra info. It does make sense that what you're really after is some estimate of the natural interest rate, and actually I like the methodology you outline in the first link--basically taking NGDP stability over the Great Moderation to imply that the natural rate is identical to the fed target rate, so that you can regress on fundamental measures of productivity, etc. Still, it would be more convincing that this estimate is correct if there were some kind of non-trivial backtesting that would have the power to falsify your methodology.

    So I think I get (and am sympathetic to) what you're trying to do here, but I don't understand your comment that the risk premium is just the term premium. Quantifying the term premium is a simple matter of just extrapolating to the short end of a yield curve. Clearly the natural rate estimation methodology you discuss in the first link, and that I summarize above, is not just that.

    1. Eric, the risk premium measure I use comes from the ACM source cited above. For them, it is an attempt at measuring the term premium. See

    2. Yes it's clear from that link and your other link on ACM that they are attempting merely to estimate the term premium. But you are making an extra assertion: that the natural rate is equal to real interest rate minus a risk premium, which you define as just the term premium. What is the justification for this asserted relationship between natural and real rates?

      In your other link I discussed above you make what looks to me like a much clearer argument about how to estimate the natural rate, which seems totally different from what you're doing here based on just the real rate and the term premium.

    3. Eric, a couple of points. First, the authors also call the term premium the risk premium. If you look at the Kim-Wright estimates they do the same. So I don't think I am being innovative here.

      Second, regarding this "But you are making an extra assertion: that the natural rate is equal to real interest rate minus a risk premium, which you define as just the term premium." This is a standard understanding in macro. Let me try to explain using the expectation hypothesis framework.

      Start with the following:
      (1) long-term interest rate = average short-term interest rate expected over same horizon + term premium

      This can be further defined as follows:

      (2) long-term interest rate = (average expected real short-term interest rate + average expected inflation) + term premium

      So if I can account for expected inflation and the term premium (or risk premium) then all that is left is the average expected path of the risk-free real short-term interest rate. That is the natural interest rate, for it determined by the economic fundamentals of productivity growth, population growth, and household time preferences.

    4. I suspect you agree that the expected real short rate is not identical to the natural rate. Maybe it's an impressionistic estimate of the natural rate--a rate corresponding to some hypothetical outcome of an unfettered market process that would equilibrate the supply and demand of loanable funds. Clearly adjusting the nominal rate by reference to a crude, ultra-aggregated measure of consumer goods inflation does not magically tell us what this outcome would have been.

      But maybe I am being too demanding. Your original point was just to add a term premium correction to this pre-existing impressionistic estimate, and I can't complain about that. For the record, the Great Moderation-based regression methodology you discuss in that other link still sounds more promising to me as a way to estimate the natural rate.

    5. Eric, to be precise, the expected real, risk-free rate--the measure I use--is an estimate of the expected average short-term natural rate over the same horizon. Thus, since I am using the expected real, risk-free 10-year rate is an estimate of the short-term real natural interest average over the next ten years.

      Though it is only an estimate, it gives really reasonable results empirically. It averages just under 2%--which is close to average growth of the U.S. economy-- and the variation around that mean tracks the business cycles. These are all things we would expect from an estimate fo the natural interest rate. That is, in theory there is a long-run natural interest rate that tracks the trend growth of the economy and a short-run one that tracks the business cycle.

      Note what I am using is a measure constructed using a macro-finance model. The hard part is done by ACM. All I did is subtract an estimate of the expected inflation. There are other ways to estimate the natural interest rate. One popular approach is use theoretical or 'structural' models. Here is on by Barsky et all (2014)

      My approach using the Great-Moderation as a benchmark is another, simpler way to do it. Ultimately, the natural rate is unobservable and all we have are estimates with varying degrees of standard errors.

  7. This comment has been removed by the author.