Scott Sumner  

What's wrong with macro?

Uber, taxis and the issue of s... Correction on Abortion...

I see at least three problems with macro, as practiced at the elite level:

1. Academics who don't know the stylized facts of macro over the past 100 years. (Fifty years is not enough, as most of the great natural policy experiments occurred between the wars.)

2. Academics who don't know the history of thought in macro, and hence who keep re-inventing the wheel, or invent false doctrines which were considered and discarded decades ago.

3. Academics who rely on abstract mathematical models that are built on assumptions wildly at variance with the real world.

Let's think about the third problem. I recently spoke with someone who had mentioned NGDP targeting at a discussion involving elite macroeconomists. Several acted as if they had never even heard of the idea. Why might that be?

One possibility is that macro models are generally structured in such as way as to generate useful policy advice, and economists often assume that inflation variability and output gaps are the relevant problems. So P and Y appear in the models, but not NGDP. It is certainly true that nominal instability is a problem, but why assume that nominal instability is better proxied by inflation than NGDP? In other words, very few elite macroeconomists seem to have read George Selgin.

I've recently been corresponding by email with a very bright UC student named Basil Halperin. One issue that keeps coming up is the fact that NGDP targeting might result in inflation instability, and/or changes in the trend rate of inflation. It's not easy to explain why this doesn't matter to someone that has been trained to assume that the costs of nominal instability are best proxied by inflation. If you ask a typical economist why, they might mumble something about relative price distortions and/or menu costs, as if these are major economic problems at low inflation rates. Consider the following:

1. Suppose oil prices suddenly double due to less supply. In that case, under inflation targeting you'd have to suddenly reduce all non-oil prices by a small amount. Imagine the menu costs if they did so, and imagine the relative price distortions if they changed at uneven rates! Even under NGDP targeting one might have to reduce non-oil prices, but not as many and not by as much. (BTW, this is one reason why targeting total nominal labor compensation is probably better than NGDP targeting.) You might object that this example assumes a supply shock. Yes, but in the face of demand shocks inflation and NGDP targeting are exactly the same.

2. It seems likely to me that the preceding costs of inflation are utterly trivial compared to some of the other costs of excess/volatile nominal growth, such as excessive taxes on nominal income, or unfair redistribution between lenders or borrowers, or excessive risk in credit markets, or financial turmoil/banking crises. And yet in all those cases nominal GDP growth is a much better proxy for the harmful nominal instability than is inflation. Even "shoe leather costs" are better proxied by NGDP.

Of course I might be wrong, but where is the empirical evidence either way?

The elite macroeconomists will try to tell you that only complex mathematical models count. That those of us trying to figure out relevant characteristics of the real world that actually need to be modeled are not doing "real science."

I'm not opposed to mathematical models, and of course some of the market monetarists are much more technical than I am. I want people to develop good technical models of NGDPLT. I took a stab at it over at TheMoneyIllusion. But there's also a need to check out the underlying assumptions of these highly technical models.

It must be a major embarrassment to the profession that us lowly MMs turned out to be more correct during the crisis than any other major group (New Keynesians, New Classical, RBC-types, etc.) and indeed more accurate than other groups on the fringes (old Keynesians, old monetarists, Austrians, MMTers, etc.):

1. It's now obvious that Fed, ECB, and BOJ policy was far too tight in late 2008 and early 2009, but MMs were just about the only people saying so at the time.

2. We correctly pointed out that fiscal austerity in 2013 would not slow growth in the US because of monetary offset, whereas in a poll of 50 elite economists by the University of Chicago, all but one gave answers implying it would slow growth.

3. We pointed out that massive QE would not lead to high inflation, while many other economists on the right said it would.

4. We correctly predicted that the BOJ and Swiss National Bank could depreciate their currency at the zero bound, while many on the left said monetary policy was pushing on a string at the zero bound.

5. We pointed out that the ECB's tightening of policy in 2011 was a huge mistake, which now almost everyone recognizes.

6. We advocated NGDPLT before it was cool (i.e. before Woodford.)

And yet because we don't always express our ideas in the language of math, we are not doing "serious science."

PS. I'd like to thank Basil for correcting an earlier version of this post, which mischaracterized the standard model. All remaining mistakes are my own.

Comments and Sharing


COMMENTS (16 to date)
cthorm writes:

I'm all for "scientific" mathematical models in economics, especially micro. But you CAN'T judge the effectiveness of these models without well honed intuition about the dynamic relationships between the variables and the quality of assumptions. If you put the math before the intuition you end up worshiping relationships like the Phillips Curve or IS/LM.

james in london writes:

The Swiss National Bank has just dropped the peg and floated. Pretty dramatic currency moves as a result. SNB President says it is due to what other currencies/countries are doing, or going to do. Does he fear a too timid ECB or a shock and awe ECB? If the latter, why does he fear it?

amelanchier writes:

"in the face of demand shocks inflation and NGDP targeting are exactly the same"

Are you assuming that demand shocks don't affect RGDP?

Scott Sumner writes:

Chorm, Good point.

James, Those facts MIGHT justify a change in the peg, say a 5% revaluation (although I doubt it.) But it surely doesn't justify this insane decision. One of the worst unforced errors I've ever seen from a central bank. When was the last time a developed country stock market fell 10% on a monetary policy decision by a central bank?

amelanchier, No, I meant that central banks try to reverse, or offset, demand shocks regardless of whether they are doing NGDP targeting or inflation targeting.

ThomasH writes:

I don't see why one can't have a model that has price level and real output uncertainty as separate arguments in a utility function and still look at the effects of NGDPL v inflation (v inflation ceiling) targeting. I don't think THAT is the reason for the slow adoption of NGDPLT.

Maybe a non-oil example would be better. Aren't petroleum products normally excluded from the index used for IT? Or is this just an argument for using non-traded goods prices as the I target?

Since there was political resistance to the Fed keeping to its IT, how do we know that resistance would be any less to hitting an NGDPLT?

[I still agree with NGDPLT>IT. The questions are to better refine the arguments.]

Scott Sumner writes:

Thomas, It's true that they often focus on core inflation, but oil does affect core inflation to some extent. In addition, there are other price shocks (such as housing) that have a big effect on core inflation.

Yes, you could enter inflation and output separately, but if you do then you are likely to assume inflation instability is a problem.

The Fed has usually been within one percent of its IT target. If it was within 1% of the NGDPLT target that would be wonderful!!

maynardGkeynes writes:

Professor Sumner, I think most observers, myself included, would look at what happened today as pretty strong evidence that the "worst unforced error" was starting the peg in the first place, rather than the ending it. Interventions always look great when they start ("No prob, we can buy as many Euros as we want!") but surprise! The price of oil drops, rich Russians decide to buy francs at the 25% discount kindly offered by the SNB, and the brilliant plan is screwed. What central bankers are in dire need of is a crash course in the the doctrine of unforeseen consequences. Although I very much enjoy your excellent commentary, I regret to say that you would not be my absolute first choice to teach that course

Roger McKinney writes:

Macro has so many more problems than these. Math is not one of them. The problem is the assumptions underlying the math. Correct the assumptions and the math will produce excellent results.

The problems with macro are the assumption of equilibrium, the fetish with perfect competition, the complete ignorance of the role of capital and the abstracting from time as if everything happened as fast as a computer solves macro equations, the assumption that the micro economy doesn't exist and all that matters is the money supply.

As someone said, mainstream macro is based on 19th century physics. It's time macro joined the 20th century, at least.

ChrisA writes:

@Maynard - "The price of oil drops, rich Russians decide to buy francs at the 25% discount kindly offered by the SNB, and the brilliant plan is screwed". Why is the brilliant plan screwed by Russians exchanging their Euro's for costlessly produced Francs? Seems like a pretty good deal for the Swiss, since these Euro's then exchanged for free can actually be used to buy real goods and services in Euroland.

The best I have been able to come up to justify the SNB action is the concern that there were increasingly more francs in existence so eventually these could come back to Switzerland and increase inflation there, necessitating higher interest rates. This concern was made worse by the announcement that the ECB was planning to do QE, meaning that SNB would have to buy even more to maintain the peg. The issue with this theory is that there is no sign of any inflation in Swiss bonds, indeed nothing but deflation is expected.

The other idea I had was that this was a morality issue. In effect the SNB was reflating the Euro area by itself. The SNB obviously felt that this was not its job.

One thing that does weaken a little from this learning (possible) is the Chuck Norris theory that the monetary base did not need to expand with a credible commitment from a CB. The SNB were obviously having to create a lot of money to keep the peg, despite the commitment. On the other hand, it wasn't credible, since it was broken! But this seems a little circular reasoning to me.

Michael Byrnes writes:

What is the SNB's rationale?

Scott Sumner writes:

Maynard, We both agree that pegging the exchange rate is not an optimal policy. I've told you what i think the Swiss should do (NGDPLT). What is your plan? You can't beat something with nothing. You can't say they should not do X, without saying what the SNB should do. You have not told me which policy is better than keeping the peg.

Roger, You said:

"Macro has so many more problems than these. Math is not one of them. The problem is the assumptions underlying the math. Correct the assumptions and the math will produce excellent results."

Isn't that what I said?

ChrisA, You said:

"The SNB were obviously having to create a lot of money to keep the peg, despite the commitment."

Actually the opposite was true, they had to create more money before the commitment, and their purchases dropped off after the peg was instituted. Now they may have to start up "QE"

Michael, See my posts over at Moneyillusion. Their rationale makes no sense. It's not even logical. They say they successfully solved the problem of an overvalued exchange rate, so their solution is to go back to an overvalued exchange rate!

Charlie writes:


Can you put together a set of forecasts/predictions? A lot of us weren't following you in real time in 2008, when you started pulling your hair out over Fed policy.

Or to put it another way, any reason I shouldn't expect:

1. -.3% Swiss change in CPI to decrease (to what - 2%...?
2. Swiss output to fall
3. Swiss unemployment to rise

I mean, is that "a test of MM" as much as there can be one? Well, it probably fits into a lot of paradigms, but at least a test of an unexpected, large, contractionary central bank move for an economy with low NGDP growth.

Shouldn't I be able to reason from something here...?

ChrisA writes:

Scott - I must admit I am now even more confused, this post from the FT ( says that the SNB "had stretched itself to buy up to €174bn in Euro-denominated assets in the third quarter of 2014, up from €89bn in mid-2011. This is equivalent to around 40 per cent of its CHF525bn balance sheet." implying that they were having to create lots of francs to support the peg. However the chart they provide of balance sheet as percent of GDP over time shows the SNB's balance sheet growth coming to a halt in mid 2012, and actually falling recently, which is consistent with your position, and strongly supportive of the "commitment" thesis. Now I am really confused as to the SNB rationale. Their policy was working, and the balance sheet was shrinking, but they decided to change the policy??

Jim Rose writes:

Thomas Humphrey wrote an excellent 250-year long literature surveys of both the rules versus discretion debate and the cost-push theories of inflation in the 1998 and 1999 Richmond Fed Quarterly.

Humphrey wrote the reviews to see if economics was a progressive science in the sense that superior new ideas relentlessly supplant inferior old ones.

Humphrey showed that policy rules were popular in good times to contain inflation, and when unemployment was rising, discretionary policies returned to vogue. The policy debate keeps recycling because

people forget the lessons of the past; and
For better or worse, politicians and the public have tended to believe that central banks, the focus of his studies, have the power to boost output, employment, and growth permanently.
Mercantilists, with their fears of hoarding and scarcity of money together with their prescription of cheap (low interest rates) and plentiful cash as a stimulus to real activity, tend to gain the upper hand when unemployment is the dominant problem.

Classicals, chanting their mantra that inflation is always and everywhere a monetary phenomenon, tend to prevail when price stability is the chief policy concern.

Cost-push fallacies about inflation were even more resilient against repeated refutations.

There is nothing new under the sun in macroeconomics.

The same issues that divided twentieth-century monetarists and non-monetarists as well as current macroeconomists were discussed by everyone from David Hume (1752) to Knut Wicksell (1898) and in the Bullionist-Anti-Bullionist and the Currency School-Banking School controversies: rules v. discretion, inflation as a monetary v. real cost push phenomenon, direct v. inverse money-to-price causality, central bank-determined v. market demand-determined money stocks, exogenous v. endogenous money, and backing v. supply-and-demand theories of money’s value

Current macroeconomists and monetary economists often unaware of the eighteenth and nineteenth-century origins of the ideas they employ.

Scott Sumner writes:

Charlie, So far as I know all macro theories predict this will reduce prices and output. But I don't believe that forecasts are the right way to test a theory. Rather I look at how policies affect market forecasts. Does anyone know if the Swiss have something comparable to TIPS spreads? That's what you'd need to test the effect.

ChrisA, I've also been surprised by the confusing data. I'm also confused by the interest rate cut to 0.75%. Why didn't they try that first, to see if it would discourage people from holding SF, before taking this much more extreme step?

Jim, You said:

"Humphrey showed that policy rules were popular in good times to contain inflation, and when unemployment was rising, discretionary policies returned to vogue"

Interesting theory, but just the opposite occurred this time around. When unemployment rose in 2008-09 the inflation target got even more rigid and inflation fell, (the opposite of what Humphrey's model would predict.)

I certainly agree that modern macroeconomists would benefit from learning more about the old debates.

Ray Lopez writes:

Sorry Dr. Sumner, but IMO numbered paragraphs 1, 2, ("Suppose oil prices...") are compound questions that make it difficult to respond to.

A good place to understand the price signal distortions caused by even zero inflation in a period of rising productivity is the book by Selgin "Less Than Zero: The Case for a Falling Price
Level in a Growing Economy" (1997)

Comments for this entry have been closed
Return to top