An Asset Price Decline IS a Negative Productivity Shock

So the Bullard speech caused a stir.  I don't much care what the man meant, if he did intend to say something sensible it doesn't appear he said it very well and from the snippets of the speech that I've read, (I haven't read the whole thing), I can see how some people got the idea that he was saying something as simple as - lower household wealth will lead people to spend less and save more thus permanently lowering aggregate demand.  If that's what Bullard meant then it was as stupid as everyone says. 

Not quite everyone said Bullard's speech was stupid however, David Andolfatto here and here provides an interpretation that actually make sense (and links to the rest of the discussion).  Of course, even if Bullard meant what Andolfatto seems to think it's quite absurd for a Fed president to be out there making speeches that require such interpretation and so in a sense Bullard deserves to be dragged over the coals a bit.  That said though, while Bullard may or may not have been saying something completely idiotic Andolfatto most certainly isn't.

Limited Commitment and the Costs of Credit Intermediation

It is perhaps an unfortunate fact of life but the reality is that we live in a world of limited commitment and that implies that credit intermediation is an astonishingly costly endeavour, costly in terms of requiring very large inputs of both labour and capital.

To illustrate let's think through an example that most people are familiar with, buying a house with a 30 year loan.  Now, let's suppose at first that the bank can't seize your house if you fail to repay the loan.  In what circumstances would the bank be willing to loan you the required amount of money?

Actually, it's hard to imagine any circumstance where a bank would loan the average person that kind of money without collateral, but why not?  Why can't the borrower just say to the bank that "I commit to pay back the loan out of my future labour income"?  After all, that is what the borrower is saying in the normal case of a standard mortgage, but why isn't that enough, why does the bank also want collateral?

Well, it's because from the bank's point of view you aren't really committed.  One problem is costly state verification, the bank doesn't necessarily know that state of your finances or of your labour prospects.  You might get out of paying back the loan by hiding your income, then you keep the house, you keep your money and when the bank calls you just tell them "leave me alone, there's no money here".  Think a bit about the amount of resources governments spend collecting income taxes and finding hidden income and you see how expensive this would be for the bank.

I could go on but I'm getting away from the main point which is that for this and a host of related problems having the loan sufficiently collateralized is the cure for what ails you.  If you post the house as collateral on the loan than the bank saves a lot on monitoring you and trying to enforce payment on the loan, so much so that in the bubble when everyone expected house prices to keep rising the people who invested in MBS where basically willing to finance mortgages with essentially no monitoring and enforcement at all.  If the borrower defaulted they were just as happy to seize the house, sell it and get all their money back.

Most importantly of all, the basic point that collateral is a cure all for otherwise nearly prohibitive credit intermediation costs is ubiquitous.  It applies everywhere that credit is extended and the costs in the absence of collateralization truly are prohibitive, just think how much resource is expanded on intermediating credit even with a high degree of collateralizaion, banks employ a lot of highly educated labour.

The Cost of Credit Intermediation and Aggregate Productivity

Well, an obvious effect of higher credit intermediation costs on aggregate productivity is the diversion of resources.  To the extent that more resource is directed at intermediating credit that leaves less for other uses but there's a further point.  In all probability the result will be a certain amount of credit rationing Stiglitz-Weiss style.

Stiglitz and Weiss made the point that if it's very expensive (or impossible) for a lender to distinguish good lenders from bad ones then the market may clear by quantity rationing rather than price adjustment, that is the interest rate may not clear the market.  The reason is that the guy who offers to pay a high interest rate may be the guy with great investment project but he also may be the guy who knows he's not gone pay anything back anyway.  It may be prohibitively expensive for the bank to tell the difference and so the bank may instead simply ration credit as a result.  Notice once again that sufficient collateralization is the solution to this problem.

So what if a fall in the value of collateral leads to an equilibrium rationing of credit?  Well, the equilibrium capital/labour ratio falls, a notion that Tim Duy and Noah Smith found rediculous.

In sum, in a world of limited commitment where credit intermediation is costly a fall in asset prices increases those costs and consequently reduces aggregate productivity.

A Permanent Fall in Asset Prices or a Monetary Policy Solution?

Of course the idea that the fall in asset prices is permanent is pretty debatable, I think instead that it implies that monetary policy is the answer.  Of course this would essentially entail the Fed raising asset prices back to their "bubble" levels and one can reasonably debate whether or not that is either possible or desirable but provided the Fed could do it there is a case to be made that the Fed should do it.

Paul Krugman goes a bit bonkers

So here's Paul Krugman on the competitive model:

Yet textbook economics says that in a competitive economy, the contribution any individual (or for that matter any factor of production) makes to the economy at the margin is what that individual earns — period. What a worker contributes to GDP with an additional hour of work is that worker’s hourly wage, whether that hourly wage is $6 or $60,000 an hour. This in turn means that the effect on everyone else’s income if a worker chooses to work one hour less is precisely zero. If a hedge fund manager gets $60,000 an hour, and he works one hour less, he reduces GDP by $60,000 — but he also reduces his pay by $60,000, so the net effect on other peoples’ incomes is zip.

Say what?  Yeah, I'm gonna have to disagree.

Well, the first two sentences are fine, nothing to complain about there, but the rest?  I mean, as the man said in the competitive model you get paid the value of your output!  If this guy works an hour less then that means his clients either have to do that hour of work themselves or not get the benefit of a service that they valued so much they'd pay $60,000 per hour for it.  In so far as they paid that hourly rate voluntarily it follows that they are necessarily worse off.


Furthermore, there are gains from trade.  In the competitive model where you get paid the value of your output that guy only gets that wage if he's really good at his job, if his client has to do that hour himself he may only value his own hour spent investing at, say, $50,000.

[Nick Rowe corrects me on those two paragraphs so I'll try again.

Yes, at the margin the hedge fund manager is paid just the right amount so that his clients are indifferent to having that last hour of his labour and paying $60,000 for it or not having it at all.  However, it's also true that the manager is indifferent to working the hour and getting the pay or taking the hour as leisure, at the margin. 

The distortion from taxation arises because the hedge fund guy necessarily reduces his labour by more than that one hour, it must be the case because he needs to reduce his labour until he's indifferent between supplying his last hour worked and the new, lower, after tax hourly income.  His elasticity of substitution approximates how many hours he'll cut back.  Now, for every additional hour, after the first, that he reduces his labour supply there is a deadweight loss to society (basically his clients lose their consumer surplus).  For each of those hours his clients were not indifferent to having the service or not, for each of those hours they strictly preferred to have the service and pay $60,000 rather than not have the service, in fact for each of those hours his clients would have willingly paid a bit more that $60,000 (basically they lose their consumer's surplus).]

It appears that in using the example of the hedge fund manager Krugman is really just thinking of a guy who can't possibly be really worth what he gets paid and who knows, maybe in the real world that's even correct.  But in the competitive model it most certainly is not correct.

I think Krugman has broken one of his own precepts, he's made an economic argument into a morality play and that just doesn't work.  I mean, forget the hedge fund guy, look at Tiger Woods.  If ever a man didn't "deserve" his millions it seems like Tiger is him but in our society there are enough people who want to watch the best golfer and few enough top golfers that the market values Tiger highly regardless what kind of guy he is.  How do you argue with that?  Tiger is paid to play good golf not be a good guy and whatever you want to say about him I don't believe you can reasonably argue that Tiger Woods isn't good a golf.

Now, of course Krugman is saying this because what he really wants to do is tax Tiger and the hedge fund guy more and there surely are good arguments for doing that.  It seems the Diamond and Saez paper he links to probably makes some of them.  Krugman doesn't.

The Magic of 5: NGDP Targeting and the Natural Rate of Unemployment

I'm a bit late writing this post, but I think there's more to say on the NGDP targeting argument that I've been having with some of the "market monetarists".  (Actually, there will be more still to say.  To their credit Nick Rowe and David Beckworth actually give reasoned arguments instead of simply asserting their correctness.  I'm not convinced but to give a decent response will entail some long posts, which I hope to find time to do.)

Actually, since I last posted Bill Woolsey has actually come out against stable NGDP growth!  Well, actually not but he does agree that NGDP growth in the 70s was very stable around its roughly 10% trend, yet because real output gaps were large during the period he actually concludes that this was a case of "irresponsible out-of-control monetary policy".  Apparently it's all down to that magic number again, stable NGDP growth isn't really what we want.  It's stable NGDP growth at exactly the right number, and the correct number has been shown by divine revelation to Scott Sumner, it's 5%.

The question I'm asking is really quite simple, Sumner says that historically periods of stable 5% NGDP growth have been considered to be good times while times when NGDP growth fell below that have tended to be recessions, well all that is certainly true.  However, does it follow as a logical consequence that during those times that NGDP growth fell below 5% the outcome would have been better if the Fed had kept NGDP on target by pushing inflation higher?

Sumner thinks the answer to this question is obviously yes but that is a non-sequitor, the conclusion does not follow from the facts as a point of logic.  You need some theoretical and empirical work to give a good answer to that question and the rigorous work that's been done invariably says the answer is no.

So, when Scott Sumner says "I don’t work with toy models; I try to stay grounded in the real world" he's really saying that he can't actually answer the question. 

Inflation Volatility and Optimal Policy

Generally speaking models that give any useful role for monetary policy tend to have some sort of nominal rigidity, price stickiness and/or wage stickiness.  The thing about it is that when you ask, within the context of the model, what is the optimal monetary policy none of the models return NGDP targeting.  Perhaps this explains all the appeals to pragmatism and simplicity on the part of the market monetarists.

One thing these models do say however is that if prices are sticky then all else equal you want to minimize price volatility.  In fact many such specifications imply the "divine coincidence" where stabilizing inflation stabilizes the output gap.  Now the divine coincidence is itself a special result that you get from price stickiness in the absence of wage stickiness but there is a general principle at work here.  If a negative shock to output is amenable to a monetary solution then it should show up in lower inflation, if it doesn't then there's nothing monetary policy can do to help and any attempt to help will be counter-productive.  This principle carries over to other situations, when wages are assumed sticky then optimal policy tries to stabilize wage growth as well as inflation.  Nowhere from rigorous theory do you get NGDP targeting as optimal policy.

Natural Rates

One of the points I've been making is that the 1970's are a period that seems to refute the idea that NGDP targeting would be a good idea.  After all, in the first few years of the decade after the 1970 recession NGDP grows at a stable 10% and the economic result in terms of real output and unemployment is just fine.  Then of course we get the oil shock, real output plummets but the fed responds by pushing up inflation to keep NGDP around it's trend path.  Sumner himself acknowledges that in the face of such a shock stable NGDP growth won't prevent a recession.  We agree on that but my question is this, is pushing inflation up to keep NGDP growth on trend a good idea?  One of the market monetarist's primary arguments in favour of NGDP targeting over inflation targeting is that pushing up inflation in response to an adverse supply shock is a good idea.

My point was simply that in the 1970's this advice was followed and it didn't seem to work out to well.  So, how about this question:  suppose that in the early 70s NGDP growth had been 5% per year when the oil shock hit.  The oil shock sent real GDP growth to -2.5%, in that case would it have been a good idea for the Fed to push inflation to 7.5% to keep NGDP growth at 5%?

The market monetarists say yes.

Finally responding to the poor labour market performance of the 70s Sumner says:

It’s also widely acknowledged that the natural rate of unemployment rose sharply during the 1970s—market monetarist policies have no control over that variable.

 Indeed.  So what about those natural rate models?

Some International Evidence on Output-Inflation Tradeoffs

Here's a really great paper with a natural rate model in it.  In the paper Robert Lucas runs time series regressions of the deviation from trend of real output on nominal income growth.  He also runs time series regressions of inflation on NGDP growth.  (Both regressions include some lagged dependent variables to improve the fit.)

He then compares the estimated coefficients with the variance of inflation and finds that the countries with high inflation variance (only 2 of the 18 countries) have coefficients on output that are a tenth as large as as the countries with low inflation volatility. 

Here's Lucas:

In a stable price country like the United States, then, policies which increase nominal income tend to have a large initial effect on real output, together with a small, positive initial effect on the rate of inflation. Thus the apparent short-term trade-off is favorable, as long as it remains unused. In contrast, in a volatile price country like Argentina, nominal income changes are associated with equal, contemporaneous price movements with no discernible effect on real output.

 It's unfortunate for Lucas that he as only 2 countries in his sample that have highly variable inflation rates while the rest have very stable inflation rates, however his results and the theory he develops in the first part of the paper imply strongly that this is not unfortunate for the residents of those other 16 countries.

The point though is that Lucas' data show that, historically, volatile inflation has been associated with monetary policy losing its ability to increase real output, this implies that keeping NGDP stable at the expense of volatile inflation is a bad idea. 

In fact, in Lucas' data Austria, Denmark, West Germany, Italy and the Netherlands all have average NGDP growth over 8% and yet all have coefficients on output that are of the same order of magnitude as the one for the US.  This suggests that perhaps, just perhaps, the poor response of real output and employment to higher NGDP after the oil shock actually was a result of monetary policy but that the average growth rate of NGDP being too high was not the problem.

More importantly for my purpose it shows that the arguments of the market monetarists that NGDP targeting is a good idea because it's better at handling supply shocks is not consistent with theory or history.

Misreading Graphs, Assuming Your Conclusion and Getting Your Own Argument Wrong

Scott Sumner replies to this post and catches me making a pretty embarrassing mistake.  I claimed that the unemployment rate in the 1960 recession stayed stable at 5% when actually it didn't, it rose to around 7%.  My eye had literally read right over the jump in unemployment without seeing it, suffice to say the "Misreading Graphs" part of the title of this post refers to me.  However, the thing is that it doesn't matter that much, the reference to the 1960's was an afterthought when the real point of the post was to point out that keeping NGDP growing didn't seem to do any good when faced with a supply shock.

So, what about the "Assuming Your Conclusion and Getting your Own Argument Wrong" part of the title?  That part applies to Sumner.

Assuming your conclusion

Sumner begins his discussion of the 1974 oil shock and subsequent economic performance with this:

NGDP growth rises sharply in the early 1970s, from barely over 5% to a peak of 12.5% in 1973. Then it falls to about 8% in the recession. Also recall that market monetarism is a natural rate model. That means even if NGDP growth had remained at 12.5% in 1974, and even if there had been no oil shock, we still would have expected a recession. The preceding demand-side growth certainly pushed unemployment below the natural rate, and hence a relapse was inevitable in 1974

Now, I know that I'm a bit challenged in reading these plots but nonetheless, it's clear that unemployment is above 5% from 1971 to the beginning of 1974 which is the same value it was for the expansionary periods of the 1960s up until about 1966.  From 1966 to 1970 the unemployment dips below 5% as nominal growth rises above 5% so one can actually make the argument that unemployment has been pushed to low but then the recession of 1970 was the one that would have restored equilibrium.

After the 1970 recession unemployment is something like 6% and gradually falls back to 5% over the years 1971 to 1973.  Is that below the natural rate?  Well, NGDP growth does average around 10% for those years and Sumner apparently takes that to imply that whatever level unemployment was at it must be below the natural rate.  However, since actual unemployment never falls below 5% that must mean that there has been an unobserved, and unexplained, shift up in the natural rate of unemployment.  He's effectively assuming his conclusion, unemployment didn't behave as expected so that must mean that the natural rate had changed.

Getting your own argument wrong

Now, let's suppose for the sake of argument that the Fed began following an NGDP level targeting scheme after the 1970 recession.  The scheme is as Sumner calls for now, you set a level target for NGDP that rises at a fixed rate per year.  This means that undershoots and overshoots are made up for.  However, suppose they'd chosen to raise the target level at 10% per year instead of Sumner's favoured 5%.

Well, in this case the path of NGDP growth might look just as the actual path did.  Coming out of the 1970 recession NGDP accelerates quickly to near 10% but the Fed doesn't tighten yet because it is still below target, the Fed needs to make up for the previous undershoot.  NGDP growth gets up over 12% and stays there for a bit more than a year by which time the Fed is at or a bit above target so the Fed begins to tighten, bring the growth rate temporarily below 10%.  The point is that in the first half at least of the 1974 recession NGDP would have been just about smack on the level target and yet RGDP would be plunging and unemployment spiking.

As I said before, the behaviour of NGDP during this episode is entirely consistent with the Fed following an NGDP level targeting regime, it's exactly as the market monetarists would hope, and yet the economic performance in the face of a supply shock is abysmal.

Arguing my point for me

Here's another quote from Scott:

Now add on the fact that there was a severe real shock to the economy, in the form of a dramatic reduction in OPEC oil production. Also recall that our economy was much more oil intensive back then. The easiest way to consider this case is to assume the AD curve is a hyperbola. Then if aggregate supply shifts left, we’ll have a recession, even with NGDP targeting.

Well yes, we agree that output would fall but Bill Woolsey had explicitly argued that one of the reasons that an NGDP target was preferred was that if a supply shock hit and wages are downwardly sticky then raising prices would maintain employment, this didn't happen.  Of course that's because the natural rate of unemployment went up and monetary policy can't prevent that! 

But then what advantage was gained by keeping NGDP on target?  That's all I'm asking.

The magic number

Finally we have this from Scott:

To add insult to injury, Adam suggests that market monetarism is somehow to blame for the inflation overshoot of the late 1970s. Recall that NGDP grew at an 11% rate from 1972 to 1981. We recommend a 5% rate. Those extra six percentage points pushed inflation from 2% to 8%. There’s no mystery here, inflation rose because the Fed was ignoring the advice of market monetarists, and regular monetarists as well.

Well, I don't see where I ever suggested that market monetarism was to blame for the inflation overshoot. My point was that keeping NGDP on a level target apparently did no good.  This passage does however reveal pretty much the entire substance of Scott's argument, it was all because they chose the wrong growth rate for NGDP!  Had they grown it at 5% instead of 10% everything would have been fine? 

Well, except that unemployment still would have risen because that was the natural rate increasing.  Inflation wouldn't have overshot?  Oh, but real GDP growth was -2.5% in the depths of the recession and remember this was due to a supply shock so you can't blame the Fed for it.  Thus maintaining a 5% NGDP growth rate would have still required pushing inflation to 7.5%.  Oh, and Sumner can't save his story by saying the problem was with expectations because the whole problem in the 1970s was that the inflation was expected.

I still think this is an ideal test case, the behaviour of NGDP is perfectly consistent with a level targeting regime and yet the outcome is horrible.  Furthermore keeping NGDP on target, even Sumner's preferred target would have required a huge and useless rise in inflation.

This is better than low and stable inflation?

NGDP Targeting and Supply Shocks

It seems that one of the main arguments in favour of NGDP targeting is that its handling of supply shocks is supposed to be preferable to the outcome under inflation targeting.  David Beckworth makes that argument here and Ryan Avent does it here.

To take a more specific example, here's Bill Woolsey saying something that, to my reading appears entirely consistent with what Beckworth and Avent are saying:

The implications of "supply shocks" play a key role in market monetarist thinking. If potential output changes, then firms will need to adjust their prices to clear markets. Adam P. argues that this will create welfare reducing fluctuations in inflation.

Consider what Adam P.'s argument implies. If there is an adverse supply shock, with potential output falling or growing less than the expected amount, then nominal GDP targeting will imply excessive inflation. With nominal GDP remaining on target, the equilibrium level of the prices of output rises. (Nominal incomes, including nominal wage rates, continue to grow at trend.)

How would inflation targeting fix this problem? The monetary authority would slow nominal GDP growth when there is an adverse supply shock, so inflation rate would remain at trend.

Now, consider a specific example--a bad harvest for corn. According to Adam P., what should happen is that the monetary authority slow nominal expenditure in the economy, to keep the inflation implied by the higher price of corn from "reducing welfare" due to its arithmetic impact on the price level. What that really means is that all the other prices in the economy, including nominal incomes like wages, must grow more slowly. The price of corn, then, rises slightly less than it would with nominal GDP targeting, and all the other prices are slightly lower than they would otherwise be. If those other prices, for example, wages, are sticky, then the effort to reduce the "welfare loss" from inflation causes a recession...

To be clear, Bill's post is long with many more details so I risk quoting out of context here but I actually think the whole argument is confused nonsense so I'll leave it to any interested reader to go to Bill's post and read the whole thing.

Now, the thing about supply shocks is that there really isn't anything monetary policy can do to "fix" the problem, one of the reasons that inflation or price level targeting is better is exactly because there is no attempt to fix a problem that is not amenable to a monetary solution.

The part that's really interesting is the last bit that I've cited, "If those other prices, for example, wages, are sticky, then the effort to reduce the "welfare loss" from inflation causes a recession".  Bill is apparently claiming that under an NGDP target the recession would be avoided.

Now, the first question is in what sense Bill thinks NGDP targeting would avoid recession, by construction a drop in output is unavoidable since that's what "supply shock" means.  I suppose he means that a rise in unemployment would be prevented, that would be consistent with his reference to sticky nominal wages.

So what about that?  Would keeping NGDP stable prevent a rise in unemployment after a supply shock?  Well, fortunately we have some data available to us and we at least one episode that is nearly universally regarded as a supply shock, the oil shock of the 1970s.  This is the supply shock, the one that gave rise to real business cycle theory and the rational expectations revolution.

Below is a time series plot of real GDP growth, nominal GDP growth and the unemployment rate during the 1970s.  On the eve of the supply shock NGDP growth is running about 10%, real GDP growth is running around 5% and unemployment is about 5%.  Now, 10% NGDP growth is a higher level than you'd typically choose to target but the important thing to notice is that in the preceding 4 years or so the unemployment rate is extremely stable right around 5%, inflation is also very stable as both real GDP growth and NGDP growth are accelerating virtually in parallel.  It certainly looks like an economy in equilibrium.

Now, when the shock hits real GDP growth of course plummets as is required for the episode to warrant the name "supply shock", but look at NGDP growth, it is remarkably stable!  This is exactly the response of monetary policy and NGDP growth that the market monetarists want, this is supposed to prevent unemployment from rising but if fails miserably!  Unemployment rises sharply and remains elevated for several years.

Furthermore, as we all know there was subsequently an acceleration in inflation that got somewhat out of hand, you might even say an inflation overshoot perhaps?

This is an amazingly pure test case, NGDP did exactly as the market monetarists would want and the whole thing is a total failure.  Monetary policy does not appeared to have fixed the situation.

An Example of Good Monetary Policy

To conclude lets take a look at another historical episode.  The following plot shows NGDP growth, real GDP growth, inflation and unemployment from 1957 to 1967 in the US.  Inflation is extremely stable while real output growth varies a lot and thus NGDP growth is very volatile as well (both inflation and NGDP are included because the inflation series is ex food and energy).

Unemployment stays stable around 5% the whole time.  So, which outcome do you prefer?

Scott Sumner says "I make no apologies for ignoring these little toy models, and having my policy analysis incorporate a complex mixture of politics, macroeconomic history, well-established basic economic principles, and logic."

I pointed out here that "well-established basic economic principles" shouldn't really be on that list, can we strike "macroeconomic history" as well?

Update: Bill Woolsey has responded with a post that completely misrepresents what I said, again.   This is getting tedious....

Here's Bill:

Adam P. shows some charts where he claims that the record of the seventies proves that a stable nominal GDP growth path would have adverse effects due to supply shocks.

Is that what I said?  Can someone point to where I said that stable nominal GDP gowth "would have" adverse effects? I'm not claiming to show that maintaining the growth rate of NGDP caused the poor outcome and I don't need to show that to make my argument.  I'm only trying to give a counter example in which maintaing NGDP growth failed to prevent the bad outcome.  (The comment about the subsequent inflation overshoot was also not a claim that NGDP targeting was its cause since the Fed wasn't actually targeting NGDP at the time, it was just a reference to our previous discussion.)

Bill again:

What is more incredible, is that he counts the sixties as a period of inflation targeting, where unemployment held stable at 5 percent due to the stable inflation rate.

Really?  I said that the sixties were a period of inlation targeting?  Can somene please show me where?  No, the point was that in 1960 there was a very large, sharp drop in NGDP growth, an NGDP shock that had absolutely no adverse effect on employment.  The Fed clearly didn't attempt to "fix" the situation and the outcome was just fine.

Now the fact that the better outcome did happen to coincide with low and stable inflation is suggestive but I never claimed that the Fed was targeting inflation at the time.  In any event, the real point of the examples is to show that maintaining NGDP growth appears to have nothing at all to do with the quality of the labour market outcome.

The Problems with NGDP Targeting in Theory and Reality

I've spent the weekend trying to get the market monetarists to actually explain why we should expect nominal GDP targeting, level or growth, to really be the panacea they claim it is (see my last several posts).  Thus far, still nothing but assertions of their correctness without much backing it up so I thought I'd do a post summarizing my case against it.

Theory

Any case in favour of NGDP targeting must start with the view that money is non-neutral and that monetary easing is at least partly reflected in increased real output instead of only raising prices.  This basically means you must believe in some variant of the Phillips curve.  Generally this means believing in some form of sticky prices and/or wages.  Scott Sumner for one frequently cites price/wage stickiness as the source of money's non-neutrality.

Well, there's a problem here.  When you actually work out the effects of sticky prices/wages you don't find that it implies targeting NGDP as being in any way optimal.  The first problem is that in any reasonable description of the world potential output must be allowed to vary, that is there are supply shocks. 

The second problem is that variability in inflation is itself welfare reducing because with sticky prices it implies that transactions are taking place at prices that are different from the level that would imply the optimal allocation.  Thus extraneous variability in inflation is welfare reducing.

These two points already show that NGDP targeting is suboptimal because keeping NGDP on target while potential output varies implies unnecessary inflation volatility.  However, we can go even farther.  In such models what you hope to do is stabilize the output gap and inflation separately, however since any variation in the output gap feeds into the Phillips curve and shows up in actual inflation it generally turns out that stabilizing inflation is the way to stabilize the output gap.

So, from the theory there's no support for NGDP targeting.  Of course, the market monetarists also tend to appeal to pragmatism in arguing for an NGDP target.  So what about that?

Reality

In a paper that has been central to the discussion Laurence Ball showed that if the Phillips curve has any weight on lagged inflation then NGDP targeting has another problem, an overshooting problem.  Basically stabilizing NGDP in response to a shock entails inflation and output oscillating inefficiently around their trend values.  Again, this extraneous volatility is welfare reducing relative to a first best policy.

The responses to this result have tended to either ignore the optimality issue entirely or to critique Ball's specification of the Phillips curve.  I'm not defending Ball's specific model but the important point is that the overshooting problem applies to any specification with any weight on the backward looking inflation term.  Now, the theoretical models discussed above don't usually have any weight on lagged inflation and this point featured in prominently in Scott Sumner's response.

Well, here's where reality gets in the way of a good story.  The thing is, empirically the Phillips curve is very well studied and it is pretty much universal that in order to get anything like an acceptable fit the lagged inflation term has to be there.  That pesky data!  There have even been several theoretical attempts to generate a Phillips curve that depends on lagged inflation to be consistent with this fact, (most theoretical models imply that the Phillips curve depends only on expectations of future inflation).

None of the responses to Ball's paper addressed the overshooting issue yet the data demands that they do if they want to claim NGDP targeting to be the cure all for the economy.

It seems to me that NGDP targeting has flaws both in theory and reality.

Monetarists Misunderstanding Other People's Arguments (and misquoting them): Bill Woolsey edition

So Bill Woolsey also has a response the my Laurence Ball post.  It's not any better.

He begins by quoting me out of context, the part where I work out the two period ahead nominal income stabilization rule in Ball's model was in direct response to what Nick Rowe had said, it wasn't central to the point of whether NGDP targeting is a good idea or not.  There I was mainly critiquing Nick for failing to at least take the question seriously.

Then Bill goes directly to the dynamic instability point:

Why would any economist take seriously the claim that a constant nominal GDP could possibly cause prices to veer off to infinity and output to zero, and then have output veer off to infinity and prices fall to zero?

Why do all the monetarists want to focus on this instead of the relevant question?  I gather it's because to the instability point they have a response while nobody actually has a rigorous argument that NGDP targeting is in any sense optimal.

But seriously, what if an NGDP target causes oscillations of output and inflation, in response to a shock, that die out over time but are still unnecessary?  Isn't that almost as bad?  And again, doesn't the fact that your rule would consider this situation a success say something about your rule?

It doesn't matter though, the fact is that output variation and inflation variation are different, they have different causes and consequences and it's not a great idea to confound them.  As I keep saying, one of the reasons an nominal income target is inferior is because some output fluctuation is efficient and this immediately implies that hitting the target would necessitate inefficient volatility in inflation, and that's something you get out of a completely forward looking model with no backward looking terms in the Phillips curve.

Bill ends with this:

With nominal GDP targeting, the equilibrium price level is the target for nominal GDP (if it actually will be reached,) divided by potential output. If the system is expected to work, the expected equilibrium price level is the target for nominal GDP divided by expected potential output. And, as usual, the equilibrium level of output is potential output. If the price level is set at the expected equilibrium price level, and nominal GDP is on target and potential income is at the expected level, then real expenditure will equal potential output. The price level will clear markets and real expenditures will purchase potential output.

That has to be the where any model begins.

I certainly don't disagree but the problem is that his argument actually implies that an inflation target is superior.  The reason is that potential output is neither constant nor observable, as it varies around the value of NGDP that is optimal changes as well unless you want inefficient, and welfare reducing, volatility in inflation.  On the other hand, if you stabilize relative prices so that price/wage stickiness doesn't distort the real outcome then you get exactly the outcome Bill describes.  If you target NGDP with any variability in all in potential output then you don't get the outcome Bill describes.