Scott Sumner  

The Keynesian shell game

Klein's battle for "liberalism... More on the Modality of Monoga...

Ever since the spectacular implosion of Keynesian economics in 2013, I've seen increasingly desperate attempts to somehow salvage the model. In this post I'll outline some of the arguments that I run across, and explain why they are misleading. I hope to leave you better prepared for future debates at the water cooler.

During the course of 2013, real GDP growth was almost twice as fast as during 2012, and nominal GDP growth also accelerated. Today some Keynesian like to minimize the austerity program of 2013, acting like nothing of importance happened. Let's look at the record, starting with what Keynesians were saying in late 2012:

Responding to a letter by 80 CEOs published in the Wall Street Journal calling for budget cuts to reduce the deficit, 350 economists published a letter calling for job stimulus and growth instead.
That letter included this warning:
At the end of the year, we face a congressionally-created "fiscal cliff," with automatic "sequestration" spending cuts everyone agrees should be stopped to prevent a double-dip recession.
Only about 6 weeks later Congress raised the payroll tax by 2 percentage points. Income taxes were raised at the same time. A few months later spending was slashed under the "sequester." By April 2013, Keynesians like Mike Konczal and Paul Krugman were calling the austerity a "test" of the market monetarist proposition that fiscal austerity would be offset by monetary stimulus.

But the double-dip recession never happened; indeed growth sped up in 2013. I haven't seen so many Keynesians get things wrong since that 1981 letter attacking Thatcher. So what could they possibly say in response? Here are things I often see in my comment section:

1. People claim that growth didn't speed up in 2013, as compared to 2012. The trick here is that they use calendar year over calendar year growth rates, which means 2013 growth heavily reflects the very slow real GDP growth in the second half of 2012, before the austerity was imposed. Economists generally agree that when you have a shock that occurs at the beginning of a calendar year, you should look at growth over the course of the year (say Q4 to Q4), not the average GDP in 2013 compared with the average GDP in 2012. Using the correct method, RGDP growth accelerated substantially, from 1.60% in 2012 to 3.13% in 2013. That probably overstates things, but growth certainly didn't slow.

3. The second trick is to downplay the amount of austerity. One trick is to use total government spending, including state and local spending, and then argue that austerity began earlier. But state and local spending is no more relevant to federal decisions over fiscal policy than is corporate investment. From the perspective of federal government policymakers, state and local spending and private investment are equally endogenous. Don't be fooled by the term "G", what matters is federal spending. State and local spending may or may not affect growth, but it's not fiscal policy.

4. Another trick is to look at government output, ignoring taxes and transfers. This is often justified because the "G" in the GDP formula is output, not spending. Alternatively, some point to new Keynesian models featuring Ricardian equivalence to justify focusing on government output. But 99% of Keynesians don't believe in Ricardian equivalence, and they constantly complain that benefit cuts and payroll tax increases will slow aggregate demand. So it's a bit late in the day to suddenly argue that taxes and transfers don't matter.

5. Pick up any textbook and they'll tell you that Keynesian economics is about deficit spending. If you look at the official deficit figures you see an enormous drop in the deficit, from $1,087b in fiscal 2012 to only $680b in fiscal 2013. But it's even worse for the Keynesians. Fiscal years run from October 1st to September 30th. But the 2013 austerity did not begin until the sharp tax increases of January 1st 2013, 3 months into the 2013 fiscal year. I tried to estimate the deficit for the calendar 2013 from this source, and came up with $1061b in 2012 and only $561b in 2013, an astounding drop of $500 billion in just one year. That's austerity! Yes, the numbers that matter are the cyclically adjusted figures. But no one knows exactly where we are in the business cycle, and in any case the growth rate is not high enough relative to trend for the cyclically adjusted figures to be all that different from the unadjusted figures under any plausible corrections. There is no doubt that 2013 was a year of austerity (or at least dramatically less stimulus, if you are a conservative who cannot stomach calling $561b deficits "austerity.")

Don't be fooled by Keynesian excuses. They are the ones that warned about the effects of 2013 austerity. They are the ones who said it was a test of market monetarism. We should take them at their word.

Comments and Sharing

CATEGORIES: Fiscal Policy

COMMENTS (42 to date)
roystgnr writes:
State and local spending may or may not affect growth, but it's not fiscal policy.

This seems like a slightly arbitrary distinction. State and local governments get controlled by much the same party apparatus as the national government. If we were to discover that we could effectively improve the economy using "iscalfay policy", where all government levels borrow or repay debt in sync, we could probably achieve that, and only historians would care that the more narrowly defined "fiscal policy" wasn't effective.

On the other hand, the choice of G seems very arbitrary to begin with.

Scott Sumner writes:

roystner, Yes, you can imagine worlds where S&L spending is fiscal policy. In China perhaps investment is part of fiscal policy. But not the US. Obama and Congress don't really have any way to boost S&L spending except direct aid, which shows up as Federal spending.

Quinn writes:

One thing I don't quite get though, and please correct me if I am wrong, but if S&L spending did increase would it not have still have the same fiscal effect. What I mean is that if federal spending decreased but S&L increased, could the S&L pick up the slack, even if partial, and if it did so, would that not take away from the argument that it was monetary policy?

SG writes:

It's interesting, because the longer the debate goes on, the clearer it is that the key difference between Keynsians (new and old) and market monetarism is not technical, but empirical.

The reason MMs have been correct in their predictions is that they have correctly diagnosed institutional dysfunction at central banks (or psychopathy, in the case of the ECB) as resulting in an effective policy quite different than the one central bankers claim to be pursuing. Bernanke/Yellen go in front of Congress and talk about "extraodinarily accomodative policy" and "monetary policy isn't a panacea" and "fiscal policymakers need to do all they can" and blah blah blah, and Keynsians took the (reasonable, in my view) stance that *of course* the central bank wouldn't purposefully tighten policy to neutralize the additional inflation/NGDP growth generated by fiscal stimulus.

Well, with central banking, actions (sometimes) speak louder than words (and all words are not equal), and MM's, rather than taking congressional testimony of the Fed chair at face value, pointed to hawkish statements on inflation from other FOMC members, hints that the Fed had already priced in the fiscal cliff/sequester, and (most importantly) MARKET REACTIONS to the above, and correctly realized that, intentionally or not, the Fed was most likely running a monetary policy targeting inflation in the 1.5% range, with a hard cap on 2%.

And, given the results, it's pretty clear who was correct in their evaluation of what the Fed's policy really was. It shouldn't be surprising though, because (as Scott has so frequently pointed out) Keynsians kept assuming not just that the Fed was incompetent, but that it was incompetent in a very specific way, with tolerance for inflation depending on a complex evaluation of where political responsibility for such inflation might ultimately lie.

And maybe there was (or is, or could be) some truth to that. After all, the Fed (to say nothing of the BOJ, ECB, and now the SNB!) has been truly awful at communicating what their policy is any how they intend to evaluate and change their behavior to carry out their policy. So the only way we can even attempt to figure out what's going on is by rengaging in the same kabuki over and over again - reading press conference tea leaves and reasoning from price changes.

Maybe, Scott, you should focus some of your research dollars on some Organizational Behavior/Psychology research around group vs. individual policymakers in the realm of monetary policy.

PS If your Hawtrey program manager job didn't require "higher ed" experience and an econ degree, I'd be very interesting in ferreting out such research myself!

Scott Sumner writes:

Quinn, Yes, but I'd same the same about investment (i.e. reverse crowding out.)

SG, You sound far more educated than many people I know who do have "higher ed" and an econ degree.

Mark Bahner writes:
One thing I don't quite get though, and please correct me if I am wrong, but if S&L spending did increase would it not have still have the same fiscal effect.

One thing that I pointed out at Simon Wren Lewis' blog is that almost all states are forbidden by the laws of their state to deficit-spend. So if they increased spending, they'd also have to increase taxes by the same amount as their increased spending.

That should make a signficant difference in the economic effect, versus the federal government that can deficit-spend.

ThomasH writes:

I take the "Keynesian" position on government spending during recessions as first that governments ought not to behave like credit- constrained household. So they do NOT need to "tighten our belt" and reduce spending to reduce the deficit (because they do not think that monetary policy will always make sure that the private sector will take up the slack).

Second, as monetary policy reduces interest rates, especially if it reduces long term interest rates, this change (and the existence of unemployed resources) will make many activities which have present costs and future benefits more attractive and governments using conventional cost benefit analysis will increase (deficit) spending on these. (Keynesians assume the monetary authority will not perversely offset these.)

In this regard, I do not see any reason in principle why states and local governments should behave any differently. They should not act as if they are credit constrained in carrying out projects with positive net present values. It was a lot more efficient to build the Hover Dan during the recession than it would have been during WWII.

In a regime in which the monetary authority is attempting to be stimulative, but is politically constrained from doing so -- not a bad description of the case in the US since 2008, the Keynesian position makes a lot of sense. Of course if a monetary authority was always on target with it's price level or NGDP level there would be no occasions for governments to run higher deficits during recessions. Keynesian prescriptions are for for epochs when monetary policy is not being conducted properly.

Brian Donohue writes:

Another great post, Scott.

Thank you for the 'dramatically less stimulus' description of recent fiscal policy. You're the first economist I've seen make this obvious point.

Ray Lopez writes:

Don't be fooled by the monetarists either. If, as Sumner would surely agree, monetary stimulus was cut back after 2010, and the US did not slip back into recession, then surely monetary stimulus has no real effect on the long term performance of the economy?

Nick writes:

I think I remember reading a comment where prof sumner talked about the idea of the fiscal multiplier as a measure of central bank error. Since our CBs are so often in error, it should be possible for a fiscal authority to make a difference if the errors are predictable. Since in almost all recent cases the error is rate targeting a bad inflation variable with an asymmetric target, it shouldn't even be that hard for fiscal decision makers to manipulate CB error.
If we had a benevolent fiscal dictator in 2007, he would have diagnosed the FED as being too tight and too focused on the price of oil. So he would have seen a negative multipler ... And an opportunity to open the SPR I guess. A carbon tax would have been very nice. Perhaps an onerous tax on the financial industry would also have been in order.
So it seems like fiscal policy can make a difference to me. But I don't think 'shovel ready projects' are ever going to be a part of it.

Scott Sumner writes:

Ray, I can't understand where that came from. Perhaps money tightened slightly in 2010, but I don't recall any dramatic changes from 2009. How are you measuring the stance of monetary policy?

Nick, It's theoretically possible that a fiscal authority could manipulate central bank incompetence, but very unlikely, as we saw in 2013.

Gary writes:
People claim that growth didn't speed up in 2013, as compared to 2012.

For the lay-person, can you explain in detail what you think is the valid way of comparing GDP growth and why?
If you're saying we should just look at the fourth quarter of 2012 vs fourth quarter of 2013, that seems extremely susceptible to cherry-picking quarters that make your point, no?

The world bank seems to show
2012: 2.3%
2013: 2.2%

And why would the world bank be using an invalid way of measuring GDP growth?

james in london writes:


Even if trend RGDP growth was flattish 2013 over 2012, or even end point 11, 12 and 13, it's hard to spot a double dip recession - as many Keynesians forecast would happen. Sure, the 1q2014 figure means some wierd stuff is going on the GDP numbers, but the decent, continuing, trend is clear enough.

Have you looked at the US jobless claims, unemployment and total employment numbers as a sort of cross check on the GDP numbers? Or a composite PMI like the Markit one? Or the Conference Board Leading Economic Indicators index? Or the Bloomberg Weekly Consumer Comfort index? Hard to spot even a slow down in growth in any of these. Many show an acceleration.

Nitpicking on RGDP growth 13 vs 12 somewhat misses the big picture.

Ram writes:

Re: #3. I think Jason Smith is right. If you want to claim that growth during 2013 was faster than during 2012, you need a model to estimate growth during each of those periods which you can use to compare them. Simply computing the implied cumulative growth between the endpoints is going to give you very noisy estimates, such that the difference is very unlikely to be significant.

Mark writes:

I think if one plotted out the annual change in the fiscal deficit as a % of GDP against the % change in real GDP a year hence, for every OECD nation since 2008, one would be hard pressed to find any correlation.

I also suggest the term "fiscal multiplier" be amended to read "fiscal divisor".

Scott Sumner writes:

Gary, I didn't say there is anything invalid about the World Bank's data, as long as it's not misused. But if they used to to examine the impact of a program that only began at the beginning of 2013, then obviously that would be invalid. When you use calendar year over calendar year growth rates, they mostly reflect growth in the second half of one year and the first half of the next. But you want to look at growth during a single calendar year, not spread out over the course of 2 years.

James, Yes, employment growth also sped up in 2013.

Ram, I'm not sure what Jason Smith said but your second sentence isn't how I'd put it. You don't need a model, what you need is to recognize that the data is noisy and that a single observation proves nothing.

But if a single observation proves nothing then everything the Keynesians have been telling us for the past 6 years is wrong, as Krugman has done exactly what you criticize, and he's done it over and over again. He did it for the eurozone, he did it for the UK, and he tried to do it for the US. But it didn't work for the US. Which shows that when you try to predict coin flips you eventually run out of luck.

In my view the market test is all that matters. You set monetary policy to where the market expects on-target NGDP growth. Anything less is Stone Age macro, not even worth discussing.

Ram writes:

Sorry, I was referring to this:

You say:

"You don't need a model, what you need is to recognize that the data is noisy and that a single observation proves nothing."

That sounds like a dodge to me. Here is what you said in the post:

"People claim that growth didn't speed up in 2013, as compared to 2012. ... Using the correct method, RGDP growth accelerated substantially, ... That probably overstates things, but growth certainly didn't slow."

If some people claim that growth didn't speed up, while you claim that it did, presumably there is a fact of the matter, or else what is the debate about? We arrive at a fact of the matter by estimating how GDP evolved during 2012, how it evolved during 2013, and comparing the two. We do this with a model. Your analysis comes from a model, too--I'm just saying the assumptions on which it is based are dubious.

When we try to estimate the path of GDP in each year, using all of the available data, and compare the two, we see that there is no significant difference. This does not mean the growth rates were the same, but it does not mean that they were different either. Whether Krugman is guilty of this kind of reasoning too is beside the point. Either you stand behind your claim or you don't. If you do, I think the analysis is flawed. If you don't, then why make the claim?

Ray Lopez writes:

@Dr. Sumner: "Ray, I can't understand where that came from. Perhaps money tightened slightly in 2010, but I don't recall any dramatic changes from 2009. How are you measuring the stance of monetary policy?"

Evidence of monetary expansion impotence:

See the second graph. Inflation and 'world gdp' falling dramatically despite a record Fed balance sheet.

I'm not an expert on this topic so I leave any last word to you, but the graph seems compelling. One could argue that 'not enough' is being done today to expand the money supply, but at some point--maybe not now but later--this argument becomes indistinguishable from metaphysics, since in any endeavor once can do even more.

Paul Johnson writes:

364 signed the 1981 Thatcher letter. 350 signed the 2012 letter. Strange, eh?

Ram writes:

Out of curiosity, I downloaded the real GDP data series from FRED (series GDPC1--please correct me if I'm using the wrong one). I used quarterly levels of real GDP from 1/1/2012 through 1/1/2014. I fit a piecewise linear model to log GDP, with the knot at 1/1/2013. The model estimates different growth rates for 2012 and 2013, assuming these to be constant during a given year, while forcing the two paths to agree on the level of real GDP on 1/1/2013. Here is what I found:

2012: 1.7% +/- 0.9%
2013: 2.4% +/- 0.9%

Difference: 0.6% +/- 1.5%

p = 0.44

(The difference in the point estimates appears to be 0.7% rather than 0.6% due to rounding.)

In other words, growth appears to have accelerated slightly, but this difference was not significantly different from 0. Plausibly, growth may have slowed by as much as 0.9%, or may have accelerated by as much as 2.1%. The evidence leans slightly in favor of acceleration, but is essentially inconclusive.

Mark Bahner writes:
Don't be fooled by Keynesian excuses. They are the ones that warned about the effects of 2013 austerity. They are the ones who said it was a test of market monetarism. We should take them at their word.

Yes, I really don't understand Paul Krugman. In December 2013 he said the U.S. was engaged in "unprecedented austerity"

And if we look at this graph--that he presents--it looks pretty clear that government spending in the U.S. (all levels combined) has become even more "austere" since the end of 2013 (see "cyclically adjusted primary balances, percent of potential GDP).

But U.S. GDP growth in the last two quarters of 2014 was pretty good. Yet he seems to be positively crowing about how he's called things correctly. I'm not asking this rhetorically...can anyone explain this?

Brian Donohue writes:

@Mark Bahner,

You two links provide a delicious and worthy addendum to Scott's fine post.

That's some top-notch commenting right there.

Nick writes:

Prof Sumner wasn't out there clamor in for austerity, insisting it would be expansionary, and calling 2013 a 'test'. He was sticking by the (previously widely held) view that the fed regularly offsets changes in fiscal policy.
By contrast, Krugman WAS loudly clamoring against austerity, claiming it would be contractionary, and that the next year would be a test.
Since prof Sumner's view (offset) actually calls for no change between years, it's hard to think of him getting a better single data point than he got. Dramatic acceleration would have made Krugman look even dumber, but would have been less of a vindication for prof Sumner. Can one years worth of data prove or disprove a theory like offset? No. But Krugman was the one claiming it could.

Boonton writes:

Regarding the 'excuses' (apologies if some of these have been covered in comments already)

1. Seems a fair point, but such a 'trick' only works for a year or two at most. Also most economic data is published using calendar years so unless you are digging into quarterly numbers that is where most of the Internet discussion will go.

3. (He seems to have forgotten point 2). I'm unclear why state and local spending is not an important factor. The essence of Keynesian economics is that you have a demand shortfall because the private sector essentially stashes cash in paper assets rather than actual Spending (I use Spending with a capital S to denote the purchase of a good or service that actually raises GDP. Buying a house from someone else does not, but the fees paid to brokers and agents does. Buying a share of Apple does not, but when Apple builds a new store or pays someone to work on designing te next iphone it does). Textbook "G" is actual gov't Spending. Things like unemployment and Social Security payments are not actually "G" but are transfer payments...but they may still increase Spending by transferring cash from people stashing it in bonds to people who will spend.

So yes even though the Fed. gov't doesn't control state and local gov'ts, it does matter what they do. S&L contraction can confound an expansionary policy at the Federal level and the reverse can happen.

4. I'm not sure what 'government output' is.

5. "Pick up any textbook and they'll tell you that Keynesian economics is about deficit spending." Actually any textbook will tell you about the 'balanced budget multiplier'. Raise taxes $100 and you decrease demand by some amount smaller than $100. If the gov't Spends $100 you raise demand by at least $100. (note Spends with capital 'S'. If the $100 is given to people as transfer payments then the effect is more muted) So yes you can pull off a pure Keynesian policy without ever letting the budget go out of balance.

To make it worse, if a Keynesian policy works and the economy recovers 'automatic stabilizers' will increase tax payments and decrease transfer payments thereby cutting the deficit. Is that reduction an austerity policy (which would imply you should see declining GDP) or the consquence of a successful Keynesian intervention (in which case improving GDP confirms the policy)?

And also, let's be clear exactly what Keynesian economics predicts. If a gov't does not stimuluate, if it prematurely inflicts austerity Keynes does not predict declining economic growth for the rest of time. The longer a recession drags on, the marginal return on Investment increases. The economy will return 'in the long run' but time is money and ten years of lost growth means a lot of suffering.

Which brings me to what is actually happening now? If you were getting a raise every year of 2%, but one year your boss gives you a 5% wage cut and then resumes your 2% yearly raise you haven't 'recovered'. Positive economic news, lower unemployment, this is all great news but what we had was a muddled implementation of Keynesian policy that began positive then turned negative. What we got was we avoided catastrophic collapse (we didn't hit 25% unemployment like we did in the Great Depression) and got a muddled rather than robust recovery.

Ram writes:

Nick, I understand. I'm not talking about Krugman, however. Scott has repeatedly made a claim about the correct way to estimate growth in 2012 and 2013, and I'm disputing that claim. I'm also observing that when this is done in a better way, we find no significant difference. Each time this is pointed out the subject gets changed--"we all agree the data is noisy; it was the Keynesians who claimed it was a test!" Maybe so, but that is a non-sequitur. Either Scott's claim is right or it is not. If we agree it is not, then he should stop making the claim.

Scott Sumner writes:

Ram, That's a very strange way of looking at things. Suppose I say Obama got more votes than Romney in 2012. Would you say I need a model? After all, official vote totals are estimates.

Here's the rules of the game as I believe Krugman accepts them. When one claims that RGDP growth will slow in 2013, one tacitly accepts the official government estimates. I see Krugman as predicting that the official numbers would come in with slower growth. If you aren't going to accept the government figures, you need to say so. But elsewhere he always seems to accept the government RGDP data when talking about the effects of policy--if things go his way. Now I don't need a model to report that the official government data shows a speedup in growth. I understand these are measured with error, and indeed said I thought the official figures overstated the speedup in growth. (Job growth sped up, but by less.) But the official figures did show a speedup, which contradicts the claim that growth would slow. No matter how large the error term you aren't going to get an estimate of slowing growth in 2013, just less confidence that it sped up.

I have no idea what your model is doing, and I certainly don't accept the claim that the speed up was not statistically significant. If you have a model that shows that, I'd love to see it, but you have not provided it. To know whether the changes were statistically significant, we'd need to know the measurement error, which would require us to know the actual GDP. But we don't know the actual GDP! So how can you possibly say the growth rate is not statistically different? The numbers by themselves tell us nothing, without knowing measurement error.

Regarding your last comment, you are confusing two very different points. One is whether we should use Q4 over Q4 or annual over annual. I think any competent economist would say Q4 over Q4. That was my point. The other issue is statistical significance. Yes, there is measurement error, I've never denied that. But we start with the Q4 over Q4 data.

And just to be clear, I do NOT think 2013 was a good test of market monetarism. It's others who said it was a test. I said, "OK, if it was a test, we passed." Do you disagree?

Ray, You can't measure the stance of monetary policy by looking at balance sheets. The Australian balance sheet is really small (no QE) and they've had the most expansionary monetary policy of any developed country central bank. The whole point of market monetarism is that the monetary base is a poor indicator of the stance of monetary policy--you look at NGDP expectations.

Mark Bahner writes:

Hi Brian,


I do want to emphasize that I sincerely want to know how Paul Krugman would explain his comments of December 2013 and his comments in January 2015. (I guess I could email him, but I just figured he probably gets 1000 emails a day, so I didn't bother.)

So if anyone at all could come up with a "this is how Paul Krugman might explain" regarding my comments of January 24th, I'd appreciate it.

Steve Roth writes:

Still starting with the unstated assumption that 2012 is an appropriate and revealing counterfactual...

Barrow writes:

Can someone explain the methodological difference between Q4 over Q4 and year over year GDP?

Ram writes:

Scott, I don't follow. I told you the model I fitted: GDPC1 from FRED, quarterly between 1/1/2012 and 1/1/2014. Piecewise linear model of log GDP, with the knot at 1/1/2013. Any competent statistician can reproduce what I did from that information alone. Using the model, I estimated the growth rate during 2012, and the growth rate during 2013, and the difference between these. The latter was not significantly different from zero. Measurement error in the GDP data is absorbed into the residual.

Why GDPC1? It was the highest frequency series I could find, maximizing the effective sample size. Why a piecewise linear? Because the claim is that "2013 growth" was faster than "2012 growth". A linear model on the log scale estimates what constant growth rate best fits the data, so a piecewise model allows us to estimate a separate rate for the two years which we can compare. Additionally, the model forces the two pieces to agree where they meet, at 1/1/2013, so that we don't get a nonsensical prediction at the change point.

You can dispute the assumption of IID normal errors, or suggest a different series to use, or whatever, but you need to
make some such assumptions to be able to say growth differed between the two years. Otherwise you just have a point estimate based on 3 data points, which carries far more noise than signal.

Again, I don't really care what Krugman said, or whether you're just playing the game by his rules. You've said that there is a right way to decide whether growth rates were different. I've provided you with a fully specified model and data source that says they were not different (nor that they were the same). What say you?

Also, the vote thing is cheap. Presumably we want to approach the question of whether the roughly simultaneous monetary easing and fiscal tightening led to a change in growth rates scientifically. I'm trying to hold you to higher standards here, but if you would rather traffic in the blogging equivalent of talking points (noisy point estimates), then fair enough.

Paul Mathis writes:

"Keynesian economics is about deficit spending."

For those who have not actually read what Keynes wrote, let me help you out:

"If it is impracticable materially to increase investment, obviously there is no means of securing a higher level of employment except by increasing consumption.

"I should support at the same time all sorts of policies for increasing the propensity to consume. For it is unlikely that full employment can be maintained, whatever we may do about investment, with the existing propensity to consume. There is room, therefore, for both policies to operate together; — to promote investment and, at the same time, to promote consumption, not merely to the level which with the existing propensity to consume would correspond to the increased investment, but to a higher level still."
The General Theory of Employment, Interest, and Money, p. 325.

Keynesian economics is not exclusively about deficit spending at all; personal consumption spending is equally important.

The Fed's $1.6 trillion QE3, which began in late 2012, increased PCE steadily throughout 2013 and thereby counteracted the federal austerity just as Keynes prescribed.

Mr. Sumner conveniently forgets about QE3 and its powerful effects on PCE that were well understood by Mr. Bernanke.

warren mosler writes:

First, I had been looking for 4% growth for 2013 and scaled back to 2% due to the tax increases and sequesters, and I thought it would continue to weaken until deficit spending increased.

Turns out there was an increase in private sector deficit spending/credit expansion on oil and gas exploration and production that offset the 2013 fiscal adjustments and further expanded in 2014 to further support GDP growth.

The increase in energy related deficit spending is now over due to the Saudi price cut, and unless deficit spending elsewhere accelerates seems to me GDP growth will quickly evaporate.

pgl writes:

[Comment removed for supplying false email address. Email the to discuss restoring your comment privileges. A valid email address is required to post comments on EconLog and EconTalk.--Econlib Ed.]

Ram writes:

Apologies for any tone in my last few comments--didn't mean for it to come across that way. As you know, I respect your work, Scott, and I don't want your larger, valid point to be clouded by a less than fully convincing claim about noisy data.

Tom Brown writes:

@Ram, what is your opinion of the arguments in the two links I provide in my comment? Thanks.

Mark Bahner writes:
The Fed's $1.6 trillion QE3, which began in late 2012, increased PCE steadily throughout 2013 and thereby counteracted the federal austerity just as Keynes prescribed.

So why, in December 2013, did Paul Krugman make such a big deal about the "austerity"? Why didn't he write, in December 2013, words to the effect of "everything is going just as Keynes prescribed"?

Nick writes:

Paul Mathis,
You wrote:
'The Fed's $1.6 trillion QE3, which began in late 2012, increased PCE steadily throughout 2013 and thereby counteracted the federal austerity just as Keynes prescribed.

Mr. Sumner conveniently forgets about QE3 and its powerful effects on PCE that were well understood by Mr. Bernanke.'

Actually, as I understand it, this is prof sumner's opinion. I don't know if he'd say say PCE was the only channel, but the fed adjusting the size of QE 3 to offset the fiscal austerity in 2013 is his entire point.
If you think Keynes himself would have agreed all the better. But several prominent 'new Keynesians' (including parts of the FOMC statements themselves) argued the opposite side of this: that at the ZLB the bank could not or was unwilling to offset.

Scott Sumner writes:

Steve Roth. Fine, then take the average RGDP growth rate of the previous 2 years, or the previous 3 years, or the previous 4 years, or the previous 5 years, or the previous 6 years, or the previous 7 years, or the previous 8 years. . . .

See my point.

Ram, There's no reason to insult me, I honestly don't know what you are doing. Let me ask the question a different way. Let's suppose we had some independent evidence that the GDP figures were 100% accurate. Would your statistical technique show a statistically significant change in the growth rate after the beginning of 2013?

In other words, I can't tell whether you are arguing:

1. Because GDP growth is measured with error, we don't know if it sped up between 2012 and 2013.


2. Even if GDP is measured 100% accurately, the speed-up is within the normal year to year variation we observe in GDP growth rates. Nothing significant.

Those are two radically different claims.

My second point is that even if everything you say is 100% true, your criticism should be directed at Krugman, not me. He's the one that claims these year to year changes in GDP growth are meaningful, not me.

It would have been helpful if you told me what was wrong with my vote analogy, instead of assuming it was a cheap shot. I honestly don't know what you are talking about.

Again, I've never claimed the speed up in GDP growth in 2013 proves anything, except that the Keynesian test failed. They set the rules, not me. To show it proved anything I'd have to show that GDP growth did not speed up for some other reason. I've consistently said the Keynesian "test" failed, that's all. They set the ground rules.

Paul, Yes, Keynes wanted to increase C and I, but viewed the deficit as a tool for doing so, that was my point.

You said:

"The Fed's $1.6 trillion QE3, which began in late 2012, increased PCE steadily throughout 2013 and thereby counteracted the federal austerity just as Keynes prescribed."

That's true but modern Keynesians don't seem to understand that point.

Scott Sumner writes:

Sorry Ram, I did see your apology and was going to acknowledge it, but forgot before posting. No hard feelings.

Ram writes:

Hi Scott,

Again, I apologize for my tone--I did not mean for it to come out that way. To answer your questions:

Even if the real GDP figures were not subject to any measurement error, there would still remain a question of what we mean by "growth in 2012" and "growth in 2013". It could mean the constant growth rate that would account for cumulative growth between the beginning and end of each year, or it could mean various other things.

If we use the former interpretation, and there is no measurement error, then the question of statistical significance does not arise. This case corresponds to your computation.

A different interpretation is that "underlying" growth is a constant that is subject to various disturbances over time. In this case, figuring out whether there was a change in that constant due to the policy change involves separating out between-year variation from within year variation. This is like trying to find a needle in a haystack, and is harder the smaller the needle is, and the bigger the haystack is.

What my model suggests is that between-year variation is swamped by within-year variation, such that we cannot establish up to the usual standards that there was any between-year variation (nor that there wasn't any).

In plain English, the difference between growth in the two years was so small relative to quarterly ups and downs that, with our data, we cannot rule out beyond a reasonable doubt that there was no difference at all. These quarterly ups and downs could be measurement error, or they could be genuine disturbances to underlying growth (e.g., Black Friday).

The problem I have with your calculation is that, since it reduces everything to a few data points, there is no way to determine whether the difference you've computed represents a change in underlying growth, or just randomly more positive disturbances during 2013 than during 2012. As a result, the calculation seems to have nothing interesting to say about the question of monetary offset.

I know that, in making such a comparison, you are playing Krugman's game by his own rules. I'm not criticizing Krugman because I think he is often careless about things like this. I'm suggesting that instead of saying things turned out the opposite of what he predicted, you just say that the "test" he laid down was a poor test, was inconclusive in any case, and leave it at that. Anything stronger seems to me to undermine your case.

My point about voting is that when we're counting votes in an election, we're not trying to come to a scientific conclusion about which candidate had more votes. In this case, we are trying to come to a scientific conclusion, since we're evaluating the merits of two scientific approaches, so I was suggesting we set a higher bar for evidence here. Again, apologies for any tone earlier.

Scott Sumner writes:

Ram, That helps a little, but I think you and I use language in a different way, which clouds understanding. Thus I would argue that in the case of no measurement error, growth clearly sped up in 2013, with 100% certainty. You don't seem to agree, and I think that's because you differentiate between growth as it actually occurred and some sort of underlying trend.

Here's an analogy. Suppose you do 10 coin flips with a fair coin and get 5 heads. Then you do 10 more and get 6 heads. There is no measurement error (by assumption.) I'd say the second round definitely saw more heads, you'd say there was no meaningful change in heads, because it might have been random noise.

In any case, none of this affects our dispute about the implications of 2013, where I think we agree. We both agree it was not a good test. To me, that means the Keynesians have NOTHING. Because all their arguments in favor of their model are just as bogus as this "test." So I continue to say, "If you buy Keynesian methodology, then 2013 refutes Keynesianism."

And if you don't buy it, they why pay any attention to Keynesians? If someone wants to spend a trillion dollars of your money, and can't come up with a good reason why, I tune them out.

Comments for this entry have been closed
Return to top