Scott Sumner  

The Keynesian information bubble

For Crying Out Loud, Dineen: D... The Minimum Wage vs. Welfare: ...

Paul Krugman likes to brag that he doesn't read conservative bloggers, as they have nothing useful to say. And certainly the world is full of conservatives making sloppy arguments. But the downside of that strategy is that when he makes silly mistakes, he never finds out about them. So Krugman keeps repeating them over and over again. His friends certainly aren't going to tell him; they often get their ideas from him, and make the same mistakes.

Over the past few years he has repeatedly cited a cross sectional study using eurozone countries as evidence for a positive fiscal multiplier. Here he goes again:

Second, while there have been a number of studies using various approaches to estimate the impact of fiscal policy in a depressed economy, I think the really decisive evidence comes from differential austerity in Europe. Here's a crude picture, simply comparing the change in IMF estimates of the cyclically adjusted budget balance with growth from 2009 to 2013:


You can try to explain this correlation away -- but it's a steep climb. The prima facie evidence is that austerity is contractionary, with a multiplier more than 1.

This is nonsense, as all 11 of the countries are within the eurozone. It would be like arguing that fiscal stimulus works by looking at the effects of government spending in individual states in the US---in other words it completely ignores the issue of monetary offset. I've pointed this out every time he cites this meaningless regression, as have numerous other bloggers that he doesn't read.

Commenter Mark Sadowski did the correct regression and discussed the results in the comment section of an earlier post:

In this particular case Krugman uses the change in the cyclically adjusted balance (CAB) from the IMF World Economic Outlook as his measure of fiscal policy stance, and the change in RGDP between calendar years 2009 and 2013. There are 11 countries in his graph, and every single one of them is from the Euro Area. . . .

There are 35 countries in the IMF Advanced Country group of which there are CAB estimates for 33. Sixteen of those are Euro Area members. Of the remaining 17, all have monetary policies independent from each other except Hong Kong which is pegged to the US dollar. Removing Hong Kong leaves 16 nations in the non-Euro Area group.

Yes, regressing RGDP growth on CAB change for the Euro Area we find that the R-squared value is 74.1% and the slope coefficient (essentially the fiscal multiplier) is 1.49 and it is statistically significant at the 1% level. So spending less in Fargo, North Dakota lowers GDP in Fargo, North Dakota.

But, regressing RGDP growth on CAB change for the non-Euro Area countries, we find the R-squared value is 0.0% (you can't make this stuff up) with a slope coefficient of 0.05 and a p-value of 94.0%. The p-value means that if the null hypothesis of a zero slope coefficient is true, there is a 94.0% probability of the test statistic being further away from zero than it is.

Or, in plain English, when countries each have a unique monetary policy, THE FISCAL MULTIPLIER IS ZERO!

And it's even worse. Krugman claims the fiscal multiplier model applies when a country is at the zero bound. But the eurozone was not even at the zero bound for the vast majority of the period being studied. Indeed the ECB tightened monetary policy sharply in 2011, driving the eurozone into a double-dip recession. The US did as much fiscal austerity, or more, and avoided a double-dip recession because it had a less contractionary monetary policy.

The Keynesian information bubble is becoming increasingly obvious over time. Driving home from work the other day I heard NPR declare that GDP growth had slowed in 2013 as compared to 2012, due to fiscal austerity. Actually GDP increased faster in 2013 than 2012. But if you take it as a matter of faith that fiscal stimulus works, I guess the next step is to start rewriting history.

COMMENTS (32 to date)
Nick Bradley writes:

Yeah, Keynesians and their peer reviewed papers showing that fiscal stimulus is effective at the ZLB when there is an output gap IS NO MATCH for commenter Mark Sadowski

Nick Bradley writes:

You see, fiscal stimulus has to work at ZLB and a bunch of unemployed resources...unless you used the money to buy everyone heroin or something.

Higher G increases private income. This is not debatable. Recipients of G then spend the money on goods and services.

At full employment, this would just increase inflation.

With slack, more supply is brought online, preventing inflation.

...this is the point where I expect unsupported blatherings about "structural issues", implying that NAIRU is permanently much higher, etc.

David R. Henderson writes:

Nice post, Scott--and Mark Sadowski. :-)

Don Geddis writes:

@Nick Bradley: You seem to have missed the critical objection. It isn't structural or NAIRU. It's monetary offset. In particular: (1) monetary policy is still potent at the ZLB (via so-called "unconventional" policy); and (2) as Scott says, "the eurozone was not even at the zero bound for the vast majority of the period being studied".

Bill Woolsey writes:

Nick Bradley:

Deep in that Keynesian echo chamber.

Your entire analysis was based upon theory. And upon the theory that monetary policy _is_ changes in a target interest rate.

Monetary offset has nothing to do with the slope of the short run aggregate supply curve.

Scott Sumner writes:

Nick, As Bill noted, you seem to think this is about the slope of the SRAS curve. It's not, I agree with Krugman that more AD will boost output.

I've talked to one of the authors of one of the "peer-reviewed papers" that Krugman cites, and he acknowledges that the multiplier estimates ignored monetary offset.

But thanks for proving the point of this post.

Thanks David.

lxdr1f7 writes:


Could it be true that the CAB is not indicative of the stance of fiscal policy in the same way that rates are not indicative of the MP stance?

An igyt writes:

Remember that Krugman has foes less enlightened than the market monetarists. He keeps making that point because it is a counter to claims of perfect Ricardian equivalence (or other flavors of flexible-price Policy Ineffectiveness).

Because in fact it is not at all obvious that spending more in Fargo increases real GDP in Fargo, if the bonds are to be paid by future revenues raised from Fargo taxpayers -- which is the right analogy for the European states. Not obvious, and maybe not true, and surely not a large impact (with such a small open economy). So it is useful to have evidence that fiscal policy has power if deployed by the the EU states.

The only problem is with transporting the conclusion to an economy with its own money, which is rather like expecting chemotherapy to extend the life of those without cancer.

Scott Sumner writes:

Lxdr, It's not true that they might be misleading in the same way, the issues are very different, but clearly the CAB might be misleading.

An igyt, I agree that Krugman must battle some pretty foolish people on the right. But using bad arguments is not the best way to do so. Those uninformed critics will then read posts like mine and become even more convinced that Krugman doesn't know what he is talking about. I've done this sort of post a number of times over at TheMoneyIllusion, and they've never been rebutted by Keynesians. Thus Krugman is losing the battle; people are increasingly skeptical of fiscal stimulus.

Nick Bradley writes:


Yeah, I appreciate the market monetarist view, but I dont see how buying equities from wealthy people boosts AD, but G buying goods and services does not.

They're both going to increase income. And as long as we don't have supply constraints (we don't!), there.shouldn't be any issues

Nick Bradley writes:

@Don Geddis,

- I didn't miss the objection.

1. I question the potency of monetary policy at the ZLB. the marginal propensity to consume for people at the top of the income distribution is much lower -- they're probably just going to buy more assets with the capital gains, more asset inflation. * that is not to say that monetary offsets at the ZLB are completely ineffective (they're not!)

2. yes - the Eurozone was not at the ZLB; but Krugman's entire post is about the US. He brings in European data, which may not be a good fit. after all, European states are not monetarily sovereign and have to pay back what they borrow. The US (or the UK or Canada or Australia,...,...) don't have to pay it all back in real terms.

Kevin Donoghue writes:

"This is nonsense, as all 11 of the countries are within the eurozone."

No, the fact that they all have their monetary policy set for them by the ECB is precisely the point. You isolate the effect of fiscal policy by holding monetary policy constant.

That's the experiment which the EU has constructed. It's a ghastly thing for Europe's unemployed, but if you were an alien investigator from an unsympathetic but highly scientific civilization studying Earthling economics, it's just what you'd want.

Nick Bradley writes:

@Kevin Donoghue, good point, but European states aren't really fiscally sovereign. its like comparing deficit spending across US state governments.

MarkusR writes:
I heard NPR declare that GDP growth had slowed in 2013 as compared to 2012, due to fiscal austerity. Actually GDP increased faster in 2013 than 2012.

Sumners fail.

Ed writes:

"Actually GDP increased faster in 2013 than 2012"

Am I missing something?

David Cushman writes:

@MarkusR: The table you link to gives nominal GDP growth rates, gives the growth % from quarter X of one year to quarter X of the next, and does not include the 4th quarter of 2013. Basically, it is incomplete for the question at hand and, conceptually much more important, it includes inflation.

The correct approach is to average the bar heights for each year in the graph that @Ed has linked to, which concerns real GDP. If you do the averaging, you find that the average quarterly growth (at annual rates) was 1.94% in 2012 and 2.72% in 2013. (From BEA data downloaded minutes ago.)

This is what Scott Sumner is talking about. It's not a fail.

J Mann writes:

This Bank of Canada explanation page discusses the difference between the different measures, but my head still hurts.

- The BEA article that Ed links to reports that "average annual growth" decreased from 12 to 13.

- "Average annual growth" is an average of the "year to year" numbers for each of the 4 quarters in a year.

- Apparently, the average annual growth of the US RGDP in 2013 is not actually the amount that US RGDP grew in 2013. If you want to compare the amount that the US RGDP grew in 2013 vs 2012, then 2013 would be bigger.

- Is there a good page someplace with examples I can work through? Khan academy or Purplemath or something?

Ed writes:

Why would we average annualized quarterly figures as opposed to simply calculating the growth rate from the end of each year?

In 2009 chained dollars:

YE2011: 15,052.4
YE2012: 16,244.6 (2.78% growth from YE2011)
YE2013: 15,767.1 (1.92% growth from YE2012)

I realize this is not what the debate or column was originally about, but some clarity would me much appreciated.

Mark A. Sadowski writes:

"Why would we average annualized quarterly figures as opposed to simply calculating the growth rate from the end of each year?"

Well, for one reason, none of the forecasts concerning the effects of fiscal consolidation by either the CBO or the major private forecasters referred to annual 2013 RGDP. They all referred to quarterly RGDP performance in 2013 or to year on year RGDP in 2013Q4.

For another, the weak performance in annual 2013 RGDP growth was already 3/8ths baked into the cake before even a single act of the federal fiscal consolidation went into effect on January 1, 2013.

A careful reading of the CBO’s estimates from November 2012 indicates that the fiscal consolidation (the 2% payroll tax increase, the high income tax increase and the sequester) should have subtracted 1.3% from year on year RGDP growth through 2013Q4. Prior to the beginning of 2013, the CBO’s forecast without any fiscal consolidation was for 2.3% year on year RGDP growth in 2013Q4. Thus the CBO forecast with fiscal consolidation was for 1.0% year on year RGDP growth in 2013Q4.

A similar thing applies to the major private forecasters. The effect of the fiscal consolidation (again, the 2% payroll tax increase, the high income tax increase and the sequester) according to Bank of America, IHS Global Insight, Moody’s, Goldman Sachs, Morgan Stanley, Macroeconomic Advisers and Credit Suisse ranged from 1.0% to 2.0% of GDP, with the average estimate being about 1.6%. The baseline forecast prior to the beginning of 2013 of these same seven private forecasters was for year on year RGDP growth of 2.0% to 3.5% in 2013Q4 with the average forecast being about 2.7%. Thus the average forecasted year on year RGDP growth in 2013Q4 adjusted for fiscal consolidation was about 1.1%.

In short, year on year RGDP growth in 2013Q4 ended up as high, or higher, than what was forecasted without fiscal consolidation.

P.S. Or, in plain English, the fiscal multiplier is zero.

David Cushman writes:

First, Ed, you have a typo: "16,244.6" is the nominal value for 2012. The real GDP value is 15470.7. But your growth rates are right, using the annual real GDP values (which are the numbers you give, subject to my correction).

The BEA's annual values, for which you have given the two latest growth rates, are not year-end values. They are the annualized average of, say, daily real GDP flow variables that accumulate over the entire year. (They can therefore be thought of as being centered on July 1.)

Here is a numerical example that may answer J Mann's request for clarity.

Suppose in 2011 real GDP was 100 in every quarter at annual rates. Suppose in 2012 it was 102 in every quarter at annual rates, and in 2013 it was 102 in the first 3 quarters and 106 in the 4th quarter at annual rates.

The three annual real GDPs are 100 102, and 103.

The annualized average of the 2012 quarterly growth rates is 2% and of the 2013 growth rates is 4%.

The annual growth rate from 2011 to 2012 is 2%, but from 2012 to 2013 it is only 1.5% (not 4%).

This example is a stylized version of what really happened: 2012's growth was concentrated nearer the beginning of the year, and 2013's nearer the end of the year. The average quarterly rate for 2012 more accurately indicates the relatively higher level of real GDP to be expected in 2014 (assuming no unusual shocks) than does the 2012 annual growth rate.

To see it another way, suppose that real GDP is the same in all 4 quarters of 2014 as it was in the 4th quarter of 2013. In my numerical example, the average of quarterly growth rates would be 0%, correctly indicating the stagnation. The average annual growth computation would show 3% growth! Which method would better indicate how we did in 2014?!

As Mark A. Sadowski would say, the 2014 annual growth rate would have the strong end-of-2013 growth "baked into the cake," just as the 2013 annual growth rate has slow 2012 year-end growth baked in. The effect occurs because, in effect, the annual growth rates are based on values centered in the middle of the various years. Quarterly growth rates are centered mid-quarter,and so involve much less of this distortion when applied to a calendar year.

Dustin writes:

Nick: "Yeah, I appreciate the market monetarist view, but I dont see how buying equities from wealthy people boosts AD, but G buying goods and services does not.

They're both going to increase income. And as long as we don't have supply constraints (we don't!), there.shouldn't be any issues"

The point is that the Fed has a target path in mind. The Fed implements policy decisions to achieve that target path. If / when an outside force (e.g., fiscal policy) influences the target path, the Fed adjusts policy to get back onto the target path = Monetary Offset

Of course, in a no central bank environment, fiscal policy could have tangible impacts, though more-so in the short run as fiscal policy is roughly symmetrical over time.

primedprimate writes:

What Kevin Donaghue said.

Europe is a good sample to study the role of fiscal policy since monetary policy was held constant and the countries were politically constrained in terms of fiscal policy as well due to various sovereign debt and political crises.

The non-European countries are a terrible sample. If fiscal policy is unconstrained and works perfectly and each of those countries had a given desired growth rate, then each would engage in fiscal policy proportional to the severity of their demand shock. So the coefficient on fiscal policy would be zero. Nick Rowe has a bunch of posts on this topic (look up Friedman's thermostat). Though Nick's post is on measuring the consequences of monetary policy, the logic is applicable to fiscal policy as well.

Mark Sadowski is very smart so I am sure he has used some instrument to correct for this tremendous endogeneity and I am eager to know what he used.

J Mann writes:

Kevin and Primedprimate, I read Krugman to arguing that his evidence shows that the stimulus and austerity that happened in the US were expansionary and contractionary, respectively. (Krugman is usually not 100% clear if the Obama tax increases count as austerity or just the spending cuts but he probably privately thinks both are.)

If that's the case, then Mark's point seems well taken. If the Fed has a target that it's going to hit regardless of fiscal activity, then the multiplier is going to be pretty close to 0, right?

Then the question for Krugman is probably if we look back at Fed activity over the last two years, does it look like the Fed was hitting its target or falling short?

Kevin Donoghue writes:

J Mann: "If the Fed has a target that it's going to hit regardless of fiscal activity, then the multiplier is going to be pretty close to 0, right?"

If I grant your premise then I guess I will have to accept your conclusion. However Ben Bernanke, for one, doesn't see that premise as being true:

"Although long-term fiscal sustainability is a critical objective, excessively tight near-term fiscal policies have likely been counterproductive. Most importantly, with fiscal and monetary policy working in opposite directions, the recovery is weaker than it otherwise would be. But the current policy mix is particularly problematic when interest rates are very low, as is the case today. Monetary policy has less room to maneuver when interest rates are close to zero, while expansionary fiscal policy is likely both more effective and less costly in terms of increased debt burden when interest rates are pinned at low levels. A more balanced policy mix might also avoid some of the costs of very low interest rates, such as potential risks to financial stability, without sacrificing jobs and growth."

primedprimate writes:

J Mann, similarly, if the fiscal authority has a target it wants to hit and fiscal policy is effective and unconstrained, the measured fiscal multiplier will be zero! So, even if fiscal policy had a true multiplier of ten Mark Sadowski's regression would have a coefficient of zero!

So unless Mark Sadowski has an instrument of some sort (which I am guessing he does and I'd like to know what it is), his regression result is irrelevant.

As regards the broader argument, my reading of Krugman is as follows: Unconventional monetary policy is effective but the Fed, for a variety of institutional factors, is unable to be expansionary enough to hit the target. In this case with an institutionally constrained Fed, fiscal and monetary policy are complements rather than substitutes.

Krugman usually takes the Fed's institutional constraint as a given (not entirely though, because he does occasionally write columns supporting unconventional monetary policy) and therefore advocates fiscal policy which can serve as a complement to inadequate monetary policy.

Sumner wants to change this institutional constraint (by implementing NGDP targeting for instance) and devotes almost all his blogging energy to this cause. Only after Sumner's proposed change is implemented would monetary policy 'offset' fiscal policy. Until then, fiscal policy and institutionally constrained monetary policy will remain complements in a liquidity trap.

Scott Sumner writes:

Nick, You are forgetting about monetary offset.

Kevin, Yes, but the study has no relevance for countries with independent monetary policies, such as the US. Krugman doesn't get that.

Ed, Growth between 2012 and 2013 is mostly growth that occurred DURING 2012. According to Krugman that's of no relevance for fiscal austerity in 2013.

Primed Primate, You are missing the point, as there is no monetary offset within the eurozone. So the data has no applicability to countries with independent monetary policies. I don't think it is even correct for the eurozone because of causality issues, but I certainly concede it might be correct for those countries.

Yes, the thermostat issue is valid, but why isn't that equally relevant to the graph Krugman showed?

primedprimate writes:

Scott, the thermostat argument applies to Krugman's graph as well, but potentially to a lesser extent because sovereign debt crises constrained fiscal policy for many of those countries. Of course, sovereign debt crises aren't truly random, but from a macro perspective may be they can, at a stretch, be considered random and therefore reduce the extent of endogeniety.

I am sure Mark Sadowski used an instrument (what is it?) to tackle this obvious endogeniety for the alternative sample but if he hasn't, his result is unsurprising and irrelevant and does not in any way bolster your argument against Krugman.

As regards the broader point, I think Kevin Donoghue's quote by Bernanke is on the money (ha!). The Fed did not think it could pull off a monetary offset and this in turn makes fiscal and monetary policy complements as suggested by the chair of the Fed himself.

Scott, I totally support your drive for NGDP targeting but until that gets implemented, I am cautiously supportive of expansionary fiscal policy in a depressed economy.

Mark A. Sadowski writes:

To be clear the results above are for simple OLS regressions which were part of a series that I estimated in response to a conversation I had with Menzie Chinn of Econbrowser about a month ago. You can learn more about this by clicking through the link to my comment and then to the Econbrowser post.

Other estimates involved using NGDP in place of RGDP, dropping outliers, and using White Standard Errors. But I did nothing more sophisticated than that, as it was simply an econblogosphere conversation, not a research project.

Yes, it's quite possible that there is a problem with endogeneity, and there are several ways such bias might be addressed econometrically. Unfortunately most of those techniques are not relevant to such a simple exercise. Did you have something specific in mind?

[no link was included--Econlib Ed.]

primedprimate writes:

Using the thermostat analogy to clarify why the European example could potentially (with a considerable stretching of assumptions) make sense:

Suppose house#1 has a busted thermostat that only goes to 55 degrees while house#2 has a busted thermostat that only goes to 60 degrees, house#3 has a busted thermostat than only goes to 70 degrees, etc. (This is analogous to saying fiscal policy is constrained and by assumption, random - i.e., the extent to which the thermostat can work is independent of the temperature outside.)

Assume all houses would like to be at 75 degrees and the outside temperature everywhere is 30 degrees.

Now if the first house ends up with a temperature of 55, the second with a temperature of 60, and the third with a temperature of 70, a regression will show the effectiveness of the heating system.

If growth rates are correlated to fiscal policy (which by assumption was constrained at different random levels in different countries) then fiscal policy is effective.

You could rightly argue (and you do!) that there was no monetary offset in the Euro area so the result from those countries is irrelevant to the US - except that Bernanke explicitly acknowledged the Fed's inability to do a full monetary offset. So assuming the absence of a full monetary offset (in line with Bernanke), the situation in the US is no longer that different from the situation of countries in the Euro area.

Mark A. Sadowski writes:

Econlib Ed.
The link to my comment is already included in this post but here it is again:

J Mann writes:

Thanks, Kevin and Primate, I really appreciate you guys taking the time.

This gets back to my personal beef with Krugman. As a non-economist who would like to learn more, there's no way I would read Krugman's description of that graph to mean "assuming that the US fed is constrained and can't or won't do monetary offset, then this graph is decisive evidence of fiscal policy's effect."

First, to the extent we all agree a less constrained monetary policy would be better, more clarity from Krugman would make it clear that's an option.

Second, with the question clarified, we could move into the evidence that the fed is constrained. (And if it's partially constrained, I assume that reduces but does not eliminate the multiplier.)

Scott's argument is that even if the central bankers don't think qe and similar unconventional monetary policy will work, they will (and did) try them in the absense of desired fiscal policy and that they will (and did) work. I don't have the knowledge to know who's right, but Krugman is begging the question.

If Bernanke proves that Krugman was right to beg the question, then Krugman should just have used Bernanke - the graph doesn't do anything that Bernanke doesn't, because if you don't accept Bernanke's argument from authority, the graph doesn't improve things.

J Mann writes:


Sorry to be a nerd, but I'm curious. Do you still argue that "Actually GDP increased faster in 2013 than 2012," and if so, could you post a link to the math?

Comments for this entry have been closed
Return to top