And a post post post script.
The case against NGDP targeting is actually even stronger theoretically than I let on in that post. I organised the last post around the simplest possible sticky price model, with no saving, capital, only one type of consumer, no sticky nominal wages, no credit frictions, a closed economy, so no trade…
If we relax these restrictions, optimal policy becomes a much more complicated beast. It involves [actually this is an informed conjecture not an assertion of analytical fact] a weighted sum of deviations of inflation, nominal wages, consumption by borrowers and lenders (entering separately), interest rate spreads, the capital stock, the real exchange rate… and with weights on inflation of prices and nominal wages an order of magnitude greater than the rest.
It would be reasonable to scoff at this and argue for nominal GDP targeting on grounds of simplicity. But then, as I said in the last post, on grounds of simplicity I’d argue for sticking with the status quo, with plenty of communication about how the central bank also cares about nominal wage growth, the real exchange rate, spreads and unemployment.
longandvariable 
“on grounds of simplicity I’d argue for sticking with the status quo”
How is “We sort of want the inflation rate to be as close to 2% as possible, but we kind of care a little more if it’s a bit too high than if it’s a bit too low, and we also care about things like employment, which we sort of want to maximize, but we can’t really say exactly how or why” anywhere near as simple as “We intend NGDP at time T to be as close to the value X as we can get it.”
A few resposes.
First, asymmetry in the inflation target is not something I advocate or is prescibed by the literature. A wording like that is in the ECB self-defined mandate. I doubt any more whether it’s a material part of what they do.
Second, you aren’t really comparing like with like. You’ve picked the hardest line NGDP level target possible. Ie no room for contingencies. And you’ve compared it to the woolly variety of inflation targeting actually pursued.
Well, the version of NGDP targeting you’ve described would be miles away from optimal policy. So easily rejected on those grounds alone.
Third, since it’s not optimal, the NGDP target is an intermediate target. So there’s complexity built in there. You are not saying that you really want nominal GDP to be such and such. You are saying that this is what you will do as the best means to achieving the ultimate aims of monetary policy, which is to improve social welfare. Unless you try to be dishonest, you have to explain that you don’t actually, literally, want to do what the intermediate says you do. It goes without saying that just as with other intermediate targets, like exchange rate and monetary targets, the NGDP target, complexities communicated or not, will not be credible because of the gap between policies that achieve it and policies that will actually be in the interests of the central bank and society.
1. The asymmetry is explicitly rejected by the Fed, but it seems clear to many observers that it is there: in other words, the Fed’s commitment to symmetry, if it really exists, in not the least bit credible; so the actual policy is a very complicated mush which seems to be both symmetric and asymmetric at the same time. The ECB’s policy, absurd though it is, is simpler in that it is clearly specified and not widely doubted by observers.
2. Well, I don’t know. It depends on your model. Woodford came up with something fairly close to NGDP level targeting as being at least provisionally optimal in his Jackson Hole paper (2012, was it?). My priors are strongly inclined to think that models in which NGDPLT is far from optimal are missing what’s really important and keying in on things that aren’t. But I suppose I’m too lazy to read and understand all the relevant papers and argue the specifics. If I were younger…
3. If you say you’re going to target something simple, and you say you’re absolutely going to do it even though it’s not perfectly optimal, it has a lot more credibility than if you say you’re trying to do the optimal thing and you have some intermediate targets that aren’t absolute but are definitely basically approximately where you want to go in the long run but you’re not going to specify to what extent and for how long and under what conditions you’re willing to deviate from them.
The thing about level targets is that, if they are actually pursued, they have a self-reinforcing credibility: if you miss one, you automatically get punished for missing it by having to take on a more onerous target in the next round. Of course there’s always a risk that you’ll discard the regime altogether as a way of avoiding the punishment — which I guess is essentially what you’re arguing, eventually it will become so suboptimal that you’ll just say screw it, the way we did with gold in the early 70s. But everyone knows what happened next in the 70s, so I think the bar for that kind of “regime change to avoid a bad outcome in the short-to-intermediate run” is quite high and will remain so.
1. It’s interesting that you think there’s asymmetry in the Fed, and that this is a widely held view. I haven’t encountered it, and I’m not of aware of any empirical evidence showing it. It would be pretty hard to prove, with all the many things the Fed could be responding to symmetrically at any point in time. Personally I think that their commitment to symmetry is credible. For 2 loose reasons. First, the FOMC is stuffed full of PhDs who read Woodford and others closely, and read the briefs of their staff who read him and others like him too. And what they read there tells them that symmetry is good for the economy. In fact, if they read closely, they’ll see arguments to be asymmetric in their concerns, but oppositely so from what you suspect. Deviations below the target are to be more strenuously avoided because they carry with them the threat of being trapped at the zero bound. Second reason I think their commitment is credible is because they were so bold with their QE and other interventions.
2. The model you are referring to is probably the simplest sticky price model with only 1 friction (sticky prices) that he uses to exposit many issues in monetary and fiscal policy. He commented informally that NGDP targeting could, by imparting the ‘levels concern’, act as a device to replicate the commitment like policy one ultimately hopes for under optimal policy (optimal policy being inflation targeting with a small weight on output stabilisation in that same stylised model). First, NGDP involves far too high a weight on output to be particularly good. Second, one can’t be confident that the commitment replicating properties of NGDP or any other levels target will operate, as I said in my posts: a) rational expectations is way to extreme an assumption, and it’s needed for this to work. b) there are multiple equilibria anyway in more realistic versions of the RE model with state variables.
3. Your point here is more arguable. I don’t agree, but I see the case you are making. It’s not enough to persuade me to ditch an already entrenched framework that a new one would not improve substantially [and is definitely not a cure for the ZLB problem that got me started on this occasion].
I don’t think it would be so unbelievable for a price level or NGDP levels target to be backslided on. At least not in the UK. First off, we are talking about a period after we have just ditched an inflation target. So regime change is not extraordinary at that point. Then, after an unavoidable overshoot, the government is going to stand by and let the central bank engineer a recession to make up for it, because of something that holds in a rational expectations real business cycle model with sticky prices? Or rather, holds in some versions of it?
Anyway, thanks for these comments. It’s good to get perspectives from outside modelworld.
Just quickly
“Then, after an unavoidable overshoot, the government is going to stand by and let the central bank engineer a recession to make up for it, because of something that holds in a rational expectations real business cycle model with sticky prices? Or rather, holds in some versions of it?”
But it’s not just the models. It’s the experience of the 1970’s (which was in fact the main motivation for those models). Overshooting on the upside has (or appears to have, which is what matters) an ugly history. (And in the 1970’s the problem existed even for those who, at the time, believed in adaptive expectations: there was no punishment for central banks that repeatedly overshot on the same side.) I think it would be easy in our time to find tough-guy parliamentarians who will decry the evil of inflation and say that something drastic needs to be done, and then quite convenient that it’s exactly what the central bank plans to do anyway. “Restore the value of money” has a better ring to it than “make sure money doesn’t keep losing value quite as quickly as it recently has.”
I’d be more concerned about credibility in the other direction. When monetary policy has temporarily lost traction and had some serious undershoots, level targeting means you’re promising high inflation, and the inflation will likely come mostly after the recovery, when people will say it’s no longer needed. But it’s also a chance to build credibility for the new regime.
Raghuman Rajan, president of the Reserve Bank of India, has pledged hmself to (flexible) inflation targeting. A year ago he stated, invoking the ideas of ‘Nobel price winners’, (see his inaugural speech) 8% of January 2015 and 6% for January 2016. In november, 2014, CPI inflation in India was… 4%. And the Reserve Bank did not act… The point is not if CPI inflation targeting should be symetric, the point is that silly rules (in this case: CPI targeting when the basket of consumer goods which is used to calculate the CPI consists to quite sme extent of food and energy items, with their very volatile prices) will lead to willy policies. A broader approach is needed. https://rwer.wordpress.com/author/merijnknibbe/
I encourage Tony Yates to post this comment* at theMoneyIllusion, run by the arch-priest of targeting NGDP. I have tried to make similar arguments but not being a skilled economist like Mr. Yates, I get shot down a lot by that blog’s volatile author. I am not against radical solutions to pressing problems, but I do have a problem with the certainty that the targeting NGDP disciples present their solution. If they simply said “we don’t know if targeting NGDP will work, but it’s worth a try during tough times” I would actually feel better. Instead they pretend it’s a cure-all.
*”The case against NGDP targeting is actually even stronger theoretically than I let on …”
Thanks for your encouragement. I’ve tried engaging before, with David Beckworth, and also there were exchanges prompted by my catch-all post on market monetarism. But I don’t think it’s worth pursuing: as far as I can tell there is no model behind their assertions, and they are unwilling to do anything but discuss using words, where there is endless scope for ambiguity and misunderstanding. The modern macro canon, and it’s policy prescriptions [a taste of which you got in my three posts] is out there, and it needs to be refuted with alternative models, and empirical evidence suggesting that these new models are more appropriate.
Higher profile economists like Krugman and Christina Romer have said it’s worth a shot. Woodford too I believe. We’ve had inflation targeting for many years and it hasn’t worked. Why not try something new. If the Central Bank is failing at NGDP path level targetting it’s a lot more obvious than how they’ve been failing at inflation targeting.
Krugman advocated for a higher inflation target as well but now has given up on both apparently viewing them as politically impossible.
Why not? Because I don’t think it would fix anything that is problematic about inflation targets. It won’t help us escape the zero bound, and NGDP targeting would not have helped avoid the financial crisis. That was caused not by bad monetary policy, but by bad financial regulation policy.
OK, I don’t get this:
“Because I don’t think it would fix anything that is problematic about inflation targets. It won’t help us escape the zero bound”
One thing to say it’s not the best solution, which we can argue about, but surely it does help fix what is problematic about the 2% inflation target and does help escape the zero bound (assuming the central bank is committed to hitting later targets after it misses early ones). NGDP level path targeting essentially says that, when a severe recession hits, we automatically temporarily increase the inflation target (assuming the hit to output is partly permanent) and promise to keep increasing it if the new target doesn’t get hit and/or the recovery doesn’t happen quickly enough. Now you can argue that rational expectations are completely irrelevant, or that the target will have zero credibility, in either of which cases it won’t be any help in escaping the zero bound, but those are pretty extreme cases. As long as the NGDP target has any effect on expectations, it makes it easier to exit the zero bound. You can also argue that it’s not worth the cost, as there won’t be perfect credibility, and that it is likely, if you continue to pursue target, that you’ll need a period of very high inflation afterward. Saying NGDPLPT causes a worse problem, though, is not the same as saying it won’t fix, or won’t help fix, the original problem.
The great thing about an NGDP level path target is that it allows the higher inflation target to be temporary (and automatic), so even if you generally don’t like high inflation (and most people apparently don’t) you can still take advantage of it to exit the zero bound. What is “problematic about inflation targets” in general is that, in order to avoid the zero bound, they have to be unacceptably high. (I mean “unacceptably” in the sense that actual people, irrational as they are, will not be willing to accept it, not that it’s necessarily bad from a welfare point of view.) NGDPLPT would fix that. Now again, you can say NGDPLT will cause a worse problem, especially if you think low variance in the inflation rate is important, but that’s a different argument.
“NGDP targeting would not have helped avoid the financial crisis. That was caused not by bad monetary policy, but by bad financial regulation policy.”
I disagree with this (unless by “helped avoid” you mean “allowed us perhaps to completely avoid”). Bad financial regulation policy was clearly a big problem, but the severity of the crisis also had to do with the fact that, once it started, people rationally adjusted their medium-term NGDP expectations downward (which, in addition to directly affecting creditworthiness, also meant that they further adjusted downward their medium-term home price expectations, which further reduced the quality of outstanding loans). If you have a credible NGDP level path target, you avoid those particular feedback loops.
The reasons I don’t think NGDPLT will fix anything are the original reasons set out in my earlier posts. It’s not optimal policy, generally, in monetary models. And the arguments for it delivering good outcomes as an intermediate target are to be set aside as very weak (as the Bank of Canada rightly did). Unfortunately, there’s no evidence in the current suite of macro models that the feedback-loop avoiding properties you describe can be delivered with NGDPLT, to any degree worth worrying about that is different from how IT is practiced. The essence of this is that in these models, the credit friction is a real problem and there is not much that monetary policy can do about it. Analogously, if you destroy productive capacity [just what the credit friction does] you can’t recreate it with lower nominal interest rates. The precise regime aimed at with interest rates is really not material.
… of course, you might say, well so much the worse for the current suite of models. But then, if you do say that, you need to set out your own competing model of the crisis and how it propagated through the macro economy. But even if you do, I’ll likely just scoff at the rational expectations assumption you invoke when defending the levels aspect of your regime and the lovely automatic stabilising qualities. I enjoyed your comment about maths and going jogging. I often feel the same way. If there were an easier way to keep track of stuff, I’d resort to it to.
TY: “[The financial crisis] was caused not by bad monetary policy, but by bad financial regulation policy.”
Do you have any evidence for this?
Friedman was 17 in 1929. He and Anna Schwartz published A monetary history of the United States in 1963, when Friedman was 51.
How do you think the cohort of economists born in 1991 will judge monetary policy makers actions during the great recession 2008-2014?
I’m curious about how you think a central bank goes about achieving a given NGDP target, inflation target, price level target, whatever. For example, suppose we are at the zero lower bound, and we think inflation is too low. What does the central bank do to make the inflation rate higher, permanently?
In suggesting higher inflation, I hope I was careful to point out it’s a preventative measure and not a cure; that while I would support a modest increase in the inflation target, that this could not be achieved until the state of the economy had evolved to the point where, in the absence of a regime change, there was enough clearance from the zero bound with interest rates to implement some monetary stimulus to deliver it. So, right now, there’s little that could be done, bar an attempt to talk down long rates, or extra QE, both of which might prove ineffectual. We might wonder, off-model, whether just announcing it might have some impact on expected inflation, and stimulate by that route. But we might equally worry that later down the road, if the zero bound meant the new higher target was simply undershot for longer (than the old lower one would have been) that this would undermine confidence in the Fed’s competence, or honesty. I wrote about this a while back here.
What I’m getting at is that the policy has to be a contingent rule for something the central bank can control more or less directly. Suppose we rule out QE, and think of the policy as a state contingent rule for the short-term nominal interest rate (leaving aside the question of which nominal interest rate). Now, suppose the nominal interest rate is currently zero, I am thinking in terms of inflation targets, and I think the inflation rate is too low. What is the policy rule for the nominal interest rate that I should implement from today foreword that will allow me to hit the inflation target? Similarly, if I thought I terms of NGDP targeting, what would I do, as a central banker?
Well, such a contingent schedule might exist and could be computed in some scenarios that line up with your comment.
For example, use the nonlinear form of the NK model. Take as initial conditions the endpoint of a 1st simulation of that model with the central bank responding to a natural rate shock using a Taylor rule, sufficiently large to drive rates to the zero bound. Then solve a Ramsey problem to find the interest rate that takes the economy as close as possible to a raised inflation target. [ie raised at the point the first simulation ends.] The solution would be a nonlinear mapping from the state [the natural rate shock, plus any lagged variables on account of dynamics built into the NK model] to the interest rate.
If we started out in the deflationary steady state with rates at zero identified in Schmitt-Grohe and Uribe, then I guess there would be no feasible action by the central bank to raise inflation. As I write this, I can’t remember what’s been deduced about this but there might be some chance that a large fiscal stimulus might be able to shift the economy up to the higher inflation steady state in that model.
Nick Rowe offered completely different reasons why NGDP target beats IT:
http://worthwhile.typepad.com/worthwhile_canadian_initi/2012/04/three-arguments-for-ngdp-targeting.html