Strangely, for many people, that title qualifies as clickbait.
It’s actually the title used by this paper, and has attracted quite a bit of attention in the popular econ media, eg on Twitter, appearing in my timeline several times. Partly because of the viral interest in nominal GDP targeting spread by the ‘market monetarists’.
I want to emphasise only that this paper does not show that nominal GDP targeting beats how the Fed or the BoE’s MPC, or the BoCanada would interpret ‘inflation targeting’. They would view their mandates as providing them with a mandate to do optimal policy, as best they see it, with the quantitative target for inflation defining the expected rate of inflation over the long term. In the UK, a Treasury review of the BoE’s mandate in 2013 interpreted things in just this way, which in the profession we’d call ‘flexible inflation targeting’.
Nominal GDP targeting can beat other constrained policies [not least because it’s not that dissimilar to ‘flexible inflation targeting’, which is optimal in the New Keynesian model] but rarely wins in general. Then again, neither does it lose by much. My beef with the NGDP lot was never that this was a dumb policy, just that it’s not right to think it would change much about the world, in particular that it would magically solve problems we experienced during the crisis, or cure booms and busts semi-automatically, forever.
As a parting shot, this paper should serve as a reminder to NGDP/MMT magpies whose googling finds it of just how policy evaluation should be done. It’s a quantitative thing, involving setting out a model and trying policies out, and scoring them somehow. If this isn’t too self-contradictory, it’s not done by repeatedly bashing readers over the heads with wordy blogs.
longandvariable 
“New Keynesian was exposed as a completely futile paradigm by its complete failure to anticipate the worst macro environment since the 1930s.”
this is a bit silly. New Keynesian models are driven by exogenous shocks, so obviously were never going to anticipate anything – they are about how to respond to shocks, not about predicting crises.
fine, if you want to say “we needed economists to predict crises, not merely think about how to respond to them” but that does not tell us the models used to think about responding to crises are bad
Well, true, but actually one reason why Smets-Wouters took hold was that its unconditional forecasting wasn’t that bad compared to a VAR.
right … but shocks are still shocks. These are not models that were ever going to say “we are about to have a crisis” … or at least not beyond the properties of whatever stochastic processes pin down how often shocks of a certain magnitude arrive in expectation.
James is complaining that models cannot anticipate trouble coming down the road. Neither Smets-Wouters or a VAR, run in 2005, is going to say, watch out something nasty is coming in 2007 – whereas some of these model of deterministic crises might in principle say, if you calibrated/estimated them in 2005, watch out trouble ahead.
True.
“As a parting shot, this paper should serve as a reminder to NGDP/MMT magpies whose googling finds it of just how policy evaluation should be done. It’s a quantitative thing, involving setting out a model and trying policies out, and scoring them somehow. If this isn’t too self-contradictory, it’s not done by repeatedly bashing readers over the heads with wordy blogs.”
It’s more a marketing thing – you guys are better at selling your snake oil.
The last 40 years in economics is a huge waste of time.
Yes, if you thought central banks faced measurement error in real time, or changes in committee membership that meant changes in the preferences of the median committee voter. Have a look. Large literature on just that.
I’d call these just ‘questions’. Most of them are discussed at length in the applied monetary econ literature. Many not resolved conclusively. If someone paid me to do this, I would take you through it all, but it would take a couple of terms to take you through it. But don’t fire them thinking that somehow these are great mysteries central bank economists aren’t already thinking about, and that aren’t already dealt with in frameworks they are given and how they are applied. They are.
On Mathematical Modelling
Privatisation:
“It’s an asset swap. QE in reverse.
Royal Mail shares are given to the purchasers and the purchaser’s cash accounts are debited which then goes to the Treasury and reduces the amount of Gilts the Treasury has to issue that week.
If you treat it like a tax payment where you get a Royal Mail share as a receipt then you’ll get the picture. ”
Nationalisation costs nothing and Privatisation raises nothing!
Tony: but when I did (sort of) build you a model to explain the desirability of NGDP targeting, you didn’t respond! I’m hurt, sniff, because I (sort of) built it specially for you.
http://worthwhile.typepad.com/worthwhile_canadian_initi/2015/01/a-model-where-ngdplt-beats-it.html
Yes, quantitative models are useful, but the historical data are also useful. The Bank of Canada kept (core) inflation very close to trend, but we still had a recession. And the fact that NGDP fell well below trend during that recession is very suggestive that NGDPLT would have done better.
And the paper you link to looks rather similar to what Scott Sumner has been saying in words.
But never mind all that. We’re a broad church, and we welcome you as a member (whether you want to be or not)!
This made me smile! I took the point of your model, though I can’t recall whether you’d demonstrated superiority of NGDP beyond doubt even there, which was, we can build a model to support the idea. Actually, I knew that – there’s Charlie Bean’s work; and Kevin Sheedy… they are out there. They just don’t command as much empirical support as the front and centre NK model with bells and whistles. In which optimal policy is a pretty complicated beast, but can be described as ‘flexible inflation targeting’. My main beef with NGDP is i) it wouldn’t make as much difference as many often claim, ii) the LEVELs version relies [just like it’s price level targeting cousin] on RE for its stabilising benefits, to be desired around the ZLB. RE I view as not to be taken so literally, and unrealistic. I find it too hard to parse Scott mostly, but if he says what Eric Sims says, in words, that’s fine by me. In which case he agrees: flexible inflation targeting rules, but NGDP isn’t much worse, so why not leave alone….
One point I forgot to make: yes, I agree that quantitative models with empirical support are useful and important. But not all of us are equally good at everything — comparative advantage and all that — so we do the things we are least relatively bad at, and hope someone else does the rest. (I think there’s something empirically wrong with how the NK Phillips Curve handles “supply shocks”, and I strongly suspect the purely random flight of Calvo’s fairy is to blame, and my little model sketch was trying to show that.)
Communication/Credibility (RE) is important for NGDPLT, but the “flexibility” bit in flexible inflation targeting is very difficult to communicate, and very difficult to implement, since we don’t observe potential output (Simon van Norden has great fun with comparing the Bank of Canada’s real time estimates of the output gap with its final estimates).
I too was a small-c conservative sitting on the fence between NGDPLT and flexible IT (“it ain’t broke badly enough to be worth fixing, so let’s stick with what we’ve got”) until I looked at the Canadian data. A hypothetical NGDPLT would have told us that AD was too low during the recession, like a guard dog barking loudly (or a magpie screeching loudly?).
As Nick remembers:
What’s wrong with economics:
“As soon as you investigate something from a slightly different angle there is huge resistance – because of course it may undermine somebody’s favourite pet theory for which they get lots of invites to conferences. Even though falsifiability is a prerequisite of any scientific statement.
Most of this stuff has more to do with formation of religious cults and tribes than anything else. That’s why people continue to use words to describe it – even though words are utterly useless at describing concurrent asynchronous dynamic processes. Or they use relatively primitive mathematics – which has nothing to do with being scientific and actually has the same purpose as keeping the Bible in Latin. Ensuring only a small elite has access to the actual meaning. “