Tuesday, January 10, 2017

The Trouble with Paul Romer

Please bear with me, as I'm out of practice. My last blog post was September 8 (seems like yesterday). As a warmup, we're going to get into Paul Romer's "The Trouble with Macroeconomics." This is somewhat old hat (in more ways than one), but this paper got a lot of attention in various media outlets, and still comes up occasionally. The titles of the articles are entertaining in themselves. For example:

"The Rebel Economist Who Blew Up Macroeconomics"
"It's Time to Junk the Flawed Economic Models That Make the World a Dangerous Place"

And who can forget:

"Famous Economist Paul Romer Says Macroeconomics is All Bullshit"

Paul has some rules for bloggers writing about his stuff. These are:
1. If you are interested in a paper, read it.
2. If you want to blog about what is in a paper, read it.
I've taken these to heart, and have indeed read the paper thoroughly. Comprehension, of course, is another matter altogether.

Paul's conclusions are pretty clear. From his abstract:
For more than three decades, macroeconomics has gone backwards...A parallel with string theory from physics hints at a general failure mode of science that is triggered when respect for highly regarded leaders evolves into a deference to authority that displaces objective fact from its position as the ultimate determinant of scientific truth.
So, those are strong charges:

1. The current stock of macroeconomists knows less than did the stock of macroeconomists practicing in 1986 or earlier.
2. A primary reason for this retrogression is that junior macroeconomists are unabashed butt-kissers. They can't bear to criticize their senior colleagues.

If you are a macroeconomist, one question that might occur to you at this point is how Romer views his role in all this macro bullshit. As we know, Romer's early work was highly influential. It's hard to say where the endogenous growth research program would be without him. This paper and this one have more than 22,000 Google scholar citations each. That's enormous. In order to collect that many citations, those papers had to be on many PhD macro reading lists, had to be read carefully, and had to form the basis for much subsequent published research. It's not for nothing that NYU, in the manner of various obituary writers, has a blurb ready to go should Paul ever collect a Nobel. Thus, it would appear that Paul singlehandedly shaped a large piece of modern macroeconomics as we know it, and should take credit for some of the bullshit he claims we are mired in. But, apparently he thinks it was someone else's fault.

A key part of Paul's paper has to do with what he views as flaws in the approach to identification in some modern macroeconomics, which he thinks yields bizarre results:
Macro models now use incredible identifying assumptions to reach bewildering conclusions.
He then sets up a straw man:
To appreciate how strange these conclusions can be, consider this observation, from a paper published in 2010, by a leading macroeconomist: "... although in the interest of disclosure, I must admit that I am myself less than totally convinced of the importance of money outside the case of large inflations."
In a paper that aims to debunk modern macro as pseudoscience, Romer here commits his first sin. Where's the citation, so we can check this quote? Of course, through the miracles of modern search engines, it's not hard to find the paper in question. It's this one, by Jesus Fernandez-Villaverde. Jesus is indeed a "leading macroeconomist," and I highly recommend his paper, if you want to learn something about the technical aspects of modern DSGE modeling.

It's useful to see the quote from Jesus's paper in context. Here's most of the paragraph in which it is contained:
At least since David Hume, economists have believed that they have identified a monetary transmission mechanism from increases in money to short-run fluctuations caused by some form or another of price stickiness. It takes much courage, and more aplomb, to dismiss two and a half centuries of a tradition linking Hume to Woodford and going through Marshall, Keynes, and Friedman. Even those with less of a Burkean mind than mine should feel reluctant to proceed in such a perilous manner. Moreover, after one finishes reading Friedman and Schwartz’'s (1971) A Monetary History of the U.S. or slogging through the mountain of Vector Autoregressions (VARs) estimated over 25 years, it must be admitted that those who see money as an important factor in business cycles fluctuations have an impressive empirical case to rely on. Here is not the place to evaluate all these claims (although in the interest of disclosure, I must admit that I am myself less than totally convinced of the importance of money outside the case of large inflations). Suffice it to say that the previous arguments of intellectual tradition and data were a motivation compelling enough for the large number of economists who jumped into the possibility of combining the beauty of DSGE models with the importance of money documented by empirical studies.
Paul wants us to think that the view expressed in the quote - that monetary factors are relatively unimportant for aggregate economic activity - is: (i) a mainstream view; (ii) a standard implication of "DSGE" models; (iii) total nonsense. First, the full paragraph makes clear that, as Jesus says, there is "an impressive empirical case" that supports the view that money is important. Jesus's parenthetical remark (the quote) states his skepticism, without getting into the reasons. Surely Paul should approve, as he thinks we're all too deferential to authority. Second, the New Keynesian research program is primarily engaged with the question of how and why monetary policy matters. We may quarrel about their mechanisms and the conclusions, but surely Mike Woodford can't be accused of arguing that monetary policy is irrelevant.

On the third point, that Jesus's skepticism is nonsense, Paul looks at some data from the Volcker disinflation. Volcker, as we all know, was Fed chair during the disinflationary period during the early 1980s. Here are Paul's conclusions:
The data displayed in Figure 2 suggest a simple causal explanation for the events that is consistent with what the Fed insiders predicted:
1. The Fed aimed for a nominal Fed Funds rate that was roughly 500 basis points higher than the prevailing inflation rate, departing from this goal only during the first recession.
2. High real interest rates decreased output and increased unemployment.
3. The rate of inflation fell, either because the combination of higher unemployment and a bigger output gap caused it to fall or because the Fed’s actions changed expectations.
So, that's a standard narrative that we hear about the Volcker era. What's wrong with it? First, it's incorrect to say that the Fed was "aiming" for a fed funds rate target at the time. It's quite interesting to read the FOMC transcripts from the Volcker era, in the context of modern monetary policy implementation. Volcker and his FOMC colleagues were operating on quantity theory principles. They aimed to reduce inflation by reducing the rate of money growth, without regard for the path of the fed funds rate, and you can see that in the data. Here's the fed funds rate during Volcker's time as Fed chair:
You can see that it's highly volatile leading up to the 1981-82 recession. And here's the growth rate in the money base, along with the unemployment rate:
The latter chart shows us that Volcker did what he said he was going to do - the money growth rate fell from about 10% in 1979 to about 3% in 1981. What about the real interest rate, which Paul focuses on?
I don't know about you, but the "simple causal explanation" in Paul's point #2 above doesn't jump out at me from the last chart. We could also look at the scatter plot of the same data:
Do high real interest rates make the unemployment rate go up? It's hard to conclude that from looking at this picture. But, the second chart (money growth rate and unemployment rate) does lend itself to a causal interpretation - money growth falls, then unemployment goes up, with a lag. That said: (i) The Volcker episode is basically one observation - we need more evidence than that; (ii) Raw correlations are suggestive only, as Paul evidently knows; (iii) If we had used the money growth/unemployment rate correlation observed in the second chart as a guide to monetary control in the post-Volcker era, we would have done really badly. As is well known, the relationship between monetary aggregates and other aggregate variables fell apart post-Volcker. I have never heard anyone in the Fed system mention monetary aggregates in a substantive policy discussion.

Paul's point #3 above reflects a Phillips curve view of the world - a higher "output gap" causes lower inflation, and lower expected inflation causes lower inflation. Though that story may appear to fit the Volcker experience, Phillips curves are notoriously unreliable. Just ask the people who keep predicting that low interest rates will make inflation take off. A more reliable guide is the Fisher effect, which fits this episode nicely. This is a scatter plot of the inflation rate vs. the fed funds rate, from the peak in the fed funds rate in June 1981 until the end of Volcker's term, with the observations connected in temporal sequence:
More often than not, the fed funds rate and the inflation rate are moving in the same direction. An interpretation is that Volcker reduced inflation by reducing the nominal interest rate - that's just Neo-Fisherism 101. Indeed, most of our models have neo-Fisherian properties, even when they have Phillips curve relations built in, as in New Keynesian models.

What's the conclusion? None of what Paul claims to be obvious concerning the effects of monetary policy during the Volcker era is actually obvious. Many hours have been spent, and many papers and books have been written by economists about the macroeconomic effects of monetary policy. But, given all that work, Jesus is not being unscientific in his skepticism about the importance of monetary factors for aggregate economic activity. It's widely accepted that central banks can and should control inflation, but there is wide dispersion - for good reasons - in views about the quantitative real effects of monetary policy.

So much for the case that modern macro leads to obviously false conclusions. What else does Paul have to say?

1. Paul doesn't like the notion of modeling business cycles as being driven by what he calls "imaginary shocks." Of course, economics is rife with such "imaginary shocks," otherwise known as stochastic disturbances. In macro, one approach - and an arguably very productive one - to constructing models that can be used to to understand the world and allow us to formulate policy, is to construct structural models (based on behavior, optimization, and market interaction) that are stochastically disturbed in ways that we can analyze and compute. Such models can produce aggregate fluctuations that look like what we actually observe. There are other approaches. For example, we can construct models with intrinsic rather than extrinsic aggregate uncertainty - models with multiple sunspot equilibria. Roger Farmer, for example, is very fond of models with self-fulfilling volatility. These models were popular for a time, particularly in the 1990s, but never caught on with the policy people. With respect to business cycle models with exogenous shocks, I can understand some of Paul's concerns. If I can't measure total factor productivity (TFP) directly, what good does it do to tell me that TFP is causing business cycles? If I have a model with 17 shocks and various ad-hoc bells and whistles - adjustment costs, habit persistence, etc. - is this any better than what Lawrence Klein was doing in the 1960s? But, and I hate to say this again, but economics is hard. In contrast to what Paul thinks, I think we have learned a lot in the last 30+ years. If he thinks these models are so bad, he should offer an alternative.

2. Paul thinks that identification - in part through the use of Bayesian econometrics - in modern macro models, is obfuscation. I might have been inclined to agree, but if you read Jesus's paper, he has some nice arguments in support of the Bayesian approach. Again, I recommend reading his paper.

3. The remaining sections of the paper, 6 through 10, are mostly free of substance. Paul asserts that macroeconomists are badly behaved in various ways. We're overly self-confident, monolithic, religious rather than scientific, we ignore parallel research programs and evidence that contradicts our theories. He seems to have it in for Lucas, Sargent, and Prescott, who apparently are engaged in some mutual fraudulent conspiracy. My favorite section is #9, "A Meta Model of Me." This paragraph sums that up:
When the person who says something that seems wrong is a revered leader of a group ... there is a price associated with open disagreement. This price is lower for me because I am no longer an academic. I am a practitioner, by which I mean that I want to put useful knowledge to work. I care little about whether I ever publish again in leading economics journals or receive any professional honor because neither will be of much help to me in achieving my goals. As a result, the standard threats ... do not apply.
This is supposed to convince us that Paul is being up front and honest - he's got nothing to lose, and what could he possibly gain from bad-mouthing the profession? Well, by the same logic, Brian Bazay had nothing to lose and nothing to gain by pushing me down in the school yard. But he did it anyway. In my experience, there is a positive payoff to demonstrating the weakness of another economist's arguments in public - and the bigger the economist, the bigger the payoff. If macroeconomists were actually as wimpy as Paul says we are, I wouldn't want to belong to this group. In general, economists are an argumentative lot - that's what we're known for. As Paul says, he's not an academic any more. Maybe he's been away so long he's forgotten how it works.

So, this is a pseudoscientific paper purporting to be about the pseudoscientific nature of modern macro. It makes bold assertions, and offers little or no evidence to back those assertions up. There's a name for that, but fear of shunning keeps me from going there. :)

12 comments:

  1. Dr. Williamson,

    The problem I see with your neo-Fisherian model is the same as that of many New Keynesian and old Keynesian models, and that is a focus on interest rates. Interest rates are a distraction. They are not the price of money. They are the merely the price of credit.

    The key to understanding the mechanisms of monetary policy, as practiced by the Fed for example, is to focus on the open market operations. The Fed funds rate target is just a target, not the transmission mechanism.

    If I were to accept your neo-Fisherian hypothesis, I'd have to believe that the Fed could buy up all the assets in the world, without significant increases in inflation. I'm pretty sure prices would start to rise well short of such an absurd ultimate goal.

    The problem many central banks around the world is having has to do with a lack of credibility regarding the willingness to facilitate sufficiently high inflation for long enough to actually get much traction.

    The Fed is a good case in point. In September of 2008, the FOMC kept the Fed Funds target at 2% two days after Lehman failed. Especially if you think demand for money increased after this event, then monetary policy effectively tightened. In retrospect, given that NGDP was allowed to fall 6.3% below trend during the Great Recession, the Fed Funds rate should have been set to -4%.

    But again, -4% is merely a target. Had markets been convinced the Fed would do whatever it took to make sure new money ceeation would meet demand, less actual Fed action than this may have been required.

    Every central bank currently experiencing ZLB problems lost credibility at some point along the way, and has failed to demonstrate the understanding and resolve to correct the problem.

    Right off the top of my head, I can recall that the ECB did what neo-Fisherians would recommend in 2011, and raised rates. Immediate market reactions and the double-dip recession that followed don't speak well for that remedy.

    ReplyDelete
    Replies
    1. There is no "neo-Fisherian model." I work with all kinds of models, including ones that lay out the details of central bank balances sheets, open market operations, and multiple assets, some playing roles as transactions media. Essentially all of these models have neo-Fisherian properties.

      Empirically, the evidence is on the neo-Fisherian side. Go through my archive and read my posts. The ECB, which you mention, is a good case in point. Sustained low nominal interest rates inevitably lead to sustained low inflation.

      Delete
  2. Is there a published account of staff discussions during the Volker days? My recollection is pretty hazy. We were almost all thinking of money growth entering the Volker disinflation and we almost all were talking about setting interest rates after. Also, I recall looking more at GDP growth than unemployment. This was all well announced and well understood and pretty hard for us to reconcile with its having to be a surprise. (I also have no idea about neo-Fisherian models. My guess is the comment has to do with inflation adjustments to nominal rates.)

    ReplyDelete
    Replies
    1. You can find the FOMC transcripts here:

      https://www.federalreserve.gov/monetarypolicy/fomc_historical.htm

      I've only dipped into this stuff, but what I read was interesting. I think you're correct that, however this policy worked, that it was not a surprise.

      Delete
  3. Dr. Williamson,

    I have read many of your intro to neo-Fisherism you link to above, and am pretty familiar with neo-Fisherian views in general. While I agree with you that there is fairly often correlation between rate increases and increases in inflation, I think it's safe to say you have not demonstrated the direction of causation is as you claim. In my view, all the logic seems to suggest causation goes the other way.

    I'm curious as to how you respond to the reductio ad absurdum I mentioned. What of the implication I see in the neo-Fisherian view that, if correct, the Fed could just buy all the assets it wanted, while generating little or no inflation? Do you disagree that this is an implication of your perspective?

    Also, how do you defend the focus on interest rates? The price of money is simply the goods, services, and investment opportunities purchasable with that money. Interest rates are the price of credit, which is a very different thing.

    As I see it, when there is a negative nominal shock, it means the growth in the money supply versus money demand is not expected to be sufficient to support the previously expected rate of economic growth. If it is a shock of say, 2%, then interest rates can be expected to fall 2% to keep real rates constant(the value of each dollar is expected to rise 2%), assuming the monetary authority allows this. Otherwise, real interest rates rise, as they did for a time during the financial crisis. Also, consumption and investment fall by 2%. This is ultimately due to sticky wages.

    Do you agree with that outline? If so, how do we get from interest rate changes being a symptom of changes in expected money supply versus demand to interest rates causing changes in NGDP?

    ReplyDelete
  4. Nice post!
    Glad you are posting again.

    ReplyDelete
    Replies
    1. "I think it's safe to say you have not demonstrated the direction of causation is as you claim."

      I'm not sure what it would take for you to think that causation was "demonstrated." What's true is that: (i) Essentially all the models we work with have neo-Fisherian conclusions. (ii) Those conclusions are consistent with the data. All we need now is for a central bank to follow neo-Fisherian principles for 20 years or so, to see how it works.

      "...the Fed could just buy all the assets it wanted, while generating little or no inflation."

      Exactly. The Bank of Japan has been buying a lot of assets, while generating little or no inflation.

      Delete
  5. Dr. Williamson,

    On the reductio ad absurdum, if a central bank were legally able and willing to buy any and all assets, whether financial or otherwise, meaning literally every item in existence in the world, could it buy every item in the world without raising prices?

    If not, then how does that not kill the neo-Fisherian hypothesis? Then, the explanation for the huge increase in the Fed's balance sheet, for example, with relatively small effects, would involve the zero lower bound problem/positive interest on reserves, and/or a credibility problem.

    It seems pretty clear to me that the QE programs had some modest, but important positive effects on economic growth, as evidenced in both market reactions and later GDP figures. 2013 seems a particularly good example of this, in which there was significant fiscal austerity.

    ReplyDelete
    Replies
    1. "...if a central bank were legally able and willing to buy any and all assets, whether financial or otherwise, meaning literally every item in existence in the world, could it buy every item in the world without raising prices?"

      This would mean that my title would be changed to "loan officer," in which case you have a lot more to worry about than what the price level would be as a consequence.

      "...then how does that not kill the neo-Fisherian hypothesis?"

      no.

      "It seems pretty clear to me that the QE programs had some modest, but important positive effects on economic growth, as evidenced in both market reactions and later GDP figures."

      I'm not sure why that's clear to you. What's your theory of QE? Why should it have big effects?

      Delete
  6. Dr. Williamson,

    I don't think QE has had big effects, at least relative to the size of the asset purchases involved, but I think they can have large effects in principle. The key from my perspective is in the creation of new money in exchange for assets, with credibility for leaving that new money in the economy long enough to restore previous NGDP growth trend crucial. I view the Fed Funds rate, and other interest rates as merely targeting/signalling tools.

    To explain, I should start with a thought experiment. If a monetary authority executed a helicopter drop, say crediting all deposit accounts with the total of 10% of NGDP, ceteris paribus, I would expect a 10% rise in the price level. However, if the helicopter drop occurred, but then the monetary authority said they would not expand the money supply again for a decade, policy would have relatively tightened, and the price level would rise less than 10%, or could fall outright, depending on prior expectations.

    In a situation in which a central bank has 100% credibility with regard to creating enough new money to return NGDP to trend, the amount of new money that would need to be created would be equal to both the percent of shortfall in the money supply versus demand and the downward movement in the Fed Funds rate.

    In other words, in the case in which NGDP fell 2% due to a nominal shock, money supply growth would need to increase 2%, and the Fed Funds rate would need to come down 2%, if the Fed Funds rate is changed at all and is to be a target/signal. In the case of QE, the Fed Funds rate need not change.

    The critical point here is that it's the asset purchases that cause the Fed Funds rate to change, not the reverse. One should focus on the amount of new money created, and how permanent that increase in the supply of money is deemed to be, and how conditional.

    Is that specific enough?

    ReplyDelete
    Replies
    1. "If a monetary authority executed a helicopter drop, say crediting all deposit accounts with the total of 10% of NGDP, ceteris paribus, I would expect a 10% rise in the price level."

      Over a six-year period, the Fed increased the quantity of outside money by almost five-fold. Over that same period prices increased by a factor of about 1.08. It wouldn't have made any difference if the Fed had made the large scale asset purchases with currency - what you might call a helicopter drop.

      Delete
  7. Dr. Williamson,

    Yes, the Fed's QE programs' limited effectiveness in my view was due to the percreived lack of the FOMC's determination to continue stimulus even in the face of temporary above-average NGDP growth. They are too fearful of inflation.

    I point out again that liquid asset markets certainly react to QE announcements as if a small pickup in economic growth is expected. This is true of stock indexes, commodity price indexes, and credit spreads. This is certainly not consistent with the strictest neo-Fisherian interpretation.

    I don't think it is a coincidence, for example, that the bottom of the Great Recession occurred in March of 2009, when the FOMC began QE. It's also not a coincidence that the economy actually grew a bit faster after fiscal austerity began in 2013, through 2014, in conjunction with another QE program.

    I argue that with a new targeting regime, like NGDP level targeting, or perhaps targeting nominal total labor compensation, the Fed's stimulus efforts would be just as effective at the zero lower bound as when that bound is not an issue.

    Also, I say that I'm general skeptical of old ideas that were rejected in the distant path within a scientific field, like neo-Fisherianism. It has been rejected by most of the best minds in economics for many decades. That doesn't mean it's wrong, but it would mean I'd be extra careful about examining such an idea.

    ReplyDelete