This post is for David, but it's mostly about me. I used to think that inflation targeting was probably roughly the best monetary policy to follow. Then, a couple of years ago, I began to think that NGDP level targeting was probably roughly the best monetary policy to follow. This post is about why I changed my mind.
I could say "it's because the facts changed", which would be true, but incomplete. Because we always interpret those facts with theory. But it's mostly because the facts changed.
Here's David: "As a price-level (PL) target is equivalent to a nominal GDP (NGDP) target in a wide class of macroeconomic models (especially under the assumption of constant productivity growth), then what more does the NGDP crowd expect from an official NGDP target? Seems to me that they are just asking for more price inflation and wishfully hoping that some of the subsequent rise in NGDP will take the form of real income.
Tell me I'm wrong (and why). "
The first bit of "theory" is this: imagine a world with no shocks, either real nor nominal, in which the central bank never makes any mistakes in hitting its target. We can imagine a world in which NGDP grows at 5% per year every year, the price level grows at 2% per year every year, and so inflation is at 2% per year every year. It's not just that it wouldn't matter in such a world whether the central bank chose NGDPLT, PLT, or IT; those three monetary policies would be observationally equivalent. No outside observer could tell which of those three policies the central bank was following.
We can only tell the difference between those three policies when there are real (supply-side) shocks, or when the central bank makes a mistake, and misses its target in one period.
The real world (OK Canada) looked very much like that imaginary world from around 1992 to 2008, the golden years of inflation targeting. We can't tell which of the three policies the Bank of Canada was following, just from looking at the data. (Actually, that isn't strictly true, because the data suggest that the Bank was leaning more towards a price level target rather than a true inflation target, and maybe (I haven't checked) an NGDPLT would fit the data slightly better than either.)
I wasn't an advocate of inflation targeting when it was first introduced. Hardly any economist was, IIRC; because the policy we now know as "inflation targeting" was something dreamed up by the Bank of Canada and the Reserve Bank of New Zealand largely in response to outside pressure to "tell us what it is you are actually trying to do so, so we know what to expect and can plan accordingly, and so we can hold you accountable for doing it". We didn't adopt IT because theory said IT was best.
Inflation targeting seemed to work pretty well, and the theory about why it worked pretty well came after, not before, the policy itself. But any outside observer, who looked at the data, but didn't know that the Bank of Canada said it was doing IT, could equally well have asked: "Why does PLT work pretty well?" or "Why does NGDPLT work pretty well?". But nobody asked those questions, of course. Since the Bank of Canada said it was doing IT, it must be IT, rather than PLT or NGDPLT, that seemed to work pretty well. With hindsight, we were daft, because we let the Bank of Canada frame the question for us.
All we know from 1992-2008 data is that either IT or PLT or NGDPLT seemed to work pretty well, but we don't know which. It took a shock to let us see the difference.
Since IT seemed to work pretty well, and I never thought that maybe it was PLT or NGDPLT that seemed to work pretty well, I became a supporter of IT.
Then the facts changed.
(I put a lot of emphasis on the Canadian facts, and not just because I am Canadian and so more aware of those facts. It's because the Bank of Canada had a clear and credible inflation target that had been in place long enough for everyone to get used to it. And because Canada didn't really have much of a financial crisis. And because the Bank of Canada made no significant mistakes in hitting its inflation target. So if IT failed in Canada, which it did, then IT really failed, and there are no excuses.)
The first fact was that the Fed (and others) hit the ZLB hard. To my mind it is debatable whether the ZLB was actually a binding constraint on the Bank of Canada, or just on the border between slack and binding with lambda=0, but it definitely came close to binding. Like everyone else, I knew of the ZLB as a theoretical possibility, but seeing a theoretical possibility in front of your eyes does tend to change how you view the world. (Japan hit the ZLB earlier, but Japan doesn't count, because Japan wasn't targeting inflation.)
Theoretically, both PLT and NGDPLT have stronger automatic stabiliser properties than IT, and would reduce the chance of hitting the ZLB, simply because they are level targets. So that first fact pushed me away from IT towards either PLT or NGDPLT. (I'm going to return later to the question of PLT vs NGDPLT as automatic stabilisers).
The second fact was three graphs that Stephen Gordon drew for me: Canadian CPI, core CPI, and NGDP. You can see those graphs, and my reactions to them, in this old post. I came down off the fence and switched from IT to NGDPLT. Let me explain why.
In brief, it's because deviations of inflation from target, or deviations of the price level from the implied target, were the guard dogs that didn't bark right when we needed them to bark. And deviations of NGDP from the implied target was a guard dog that barked loud and clear.
Yes, it would be nice if people could be more certain about what the future inflation rate or price level would be. And it would also be nice if people could be more certain about what the future level of nominal income would be. We could argue about which of those would be more nice. But that's not the main reason I do not want a totally random and unpredictable monetary policy. A totally random and unpredictable monetary policy, where the central bank just tosses a coin every period to decide what to do, would be bad because in a world of nominal rigidities a monetary policy like that would cause undesirable real fluctuations.
What we are looking for in monetary policy is some target variable, and some numerical target for that target variable, so that deviations of the target variable from target act as a good guard dog that barks loudly and clearly when, and only when, there are undesirable real effects from the interaction between monetary policy, nominal rigidities, and whatever other shocks happen to be hitting the economy. (And that target variable has to be a nominal variable, of course, because of long run neutrality.)
We do not understand nominal rigidities very well at all. We do not understand the shocks hitting the economy very well at all. We do not understand the interaction between nominal rigidities, those shocks, and monetary policy, very well at all. Our hopes of solving theoretically for the optimal guard dog, with any degree of confidence, are very small.
We had three guard dogs, named IT, PLT, and NGDPLT, that were all saying about the same thing from 1992 to 2008. The Bank of Canada listened to the first of those three dogs, and things seemed to go pretty well. So we figured IT was a good guard dog. In 2009 things changed. The Bank of Canada kept on listening to the IT dog, and did what was needed to make sure the IT dog stayed fairly quiet. The PLT dog stayed fairly quiet too. But the NGDP dog started barking loudly, telling us that monetary policy was too tight. And when I looked out the window, I saw exactly those symptoms that I normally associate with a random tightening of monetary policy: people having greater difficulty selling things for money; greater ease buying things for money; and a fall in the quantities of things sold for money. I saw exactly the same sort of recession I would expect to see if monetary policy suddenly at random became too tight. The NGDP dog was right to start barking loudly; the IT and PLT dogs failed to warn us of the recessionary burglar.
So I say: get rid of the IT dog and start listening to the NGDPLT dog instead.
[Update: and raising the inflation target is like giving the IT dog a hearing aid in the hope it will do better; and switching to a temporary NGDPLT but only during a recession is like having the NGDPLT dog temporarily replace the IT dog when you already know there's a burglar in the house.]
And back to the question of automatic stabilisers keeping you off the ZLB. If the price level had fallen below the implied PLT path when the Canadian economy grazed the ZLB, then PLT would have had the desired automatic stabiliser effect; people would have expected a temporarily higher inflation rate, which would have increased aggregate demand for any given nominal interest rate. But it didn't fall below the implied target path. But NGDP did fall well below the implied NGDPLT path, so NGDPLT would have had the desired automatic stabiliser effect: people would have expected a temporarily higher NGDP growth rate, which would have increased aggregate demand for any given nominal interest rate.
You want a theoretical model which shows that NGDPLT beats PLT? I expect I could rig one up. But it would be fake. It wouldn't explain why I prefer NGDPLT over PLT. It was never a model that caused me to like IT before 2008. I liked IT because it seemed to work, until it didn't. Nobody liked IT because their model said it was best. Their models were ex post rationalisations of why IT seemed to work pretty well. Given the state of our theoretical knowledge, all our models can tell us with any confidence is that some monetary policies will work pretty badly.
Put it another way: I do not know why the IT and PLT dogs didn't bark when they should have barked. And the main reason I don't know is that I do not understand nominal rigidities very well at all. I can make some guesses, but that's about it. All I do know is that they didn't bark when they should have barked.
Sure, Scott Sumner had an influence on my thinking too, mostly by telling me about the NGDP dog. But if the Canadian NGDPLT dog hadn't barked at the right time, while the other two Canadian dogs failed, I would probably still be sitting on the fence.
There are also some long run reasons as well to support NGDPLT over IT or PLT, but they aren't the important reasons why I switched. (NGDPLT means inflation is higher when long run real growth is lower and the natural rate of interest is probably lower, which gives better insurance against hitting the ZLB when you need that insurance most. NGDPLT probably gives better risk sharing between borrowers and lenders in the face of uncertain long run growth.)