Simon’s defence of NGDP targeting makes many good points – in contrast to what he calls the ‘faith-based’ argumentation of the market monetarists – and this blog responds to some of them.
He rightly challenges the emphasis I place on the extremely stark result in the modern sticky price macro model that optimal policy involves stabilising a weighted sum of inflation and output (and other things in more complex models) with a weight an order of magnitude greater attached to inflation than other things.
That is fair. This weight is controversial. Many participants in the literature I have spoken to privately would say that they doubt its literal truth. I would make a few points by way of a reply.
First, the same models would also put very high weights on nominal wage inflation. And zero weight on sectors with flexible prices. The main point is that this weight has to come from a grasp of the frictions in the model. And, repeating the old Neil Wallace dictum: it takes a model to beat a model.
Second, whilst the survey results asking people what concerns them (inflation, unemployment, the weather?) are interesting, I don’t think they are that informative about the underlying frictions that characterise the macro economy and therefore which should guide macro policy. People may not have a clue what is good for the macro economy, and, therefore, for them. Ultimately, I would only be convinced by a competing model that does a more convincing job of accounting for price-stickiness (taking that to be a macroeconomic fact, controversial for some, I know) and casts a different light on optimal policy. There are one or two already, but it is early days for that literature.
To dig into the intuition of what leads to the high weight on inflation. It’s not any distaste for inflation on the part of firms or consumers, or that it makes it hard for them to tell what’s what in the prices that confront them. It’s that unplanned inflation erodes their real wage or relative price that they would ideally prefer, and, as a result, firms and consumers end up experiencing volatility in the hours that they work, or the amount they have to produce, and therefore volatility in the wages/ profits that they recoup and the amount they can consume. In fact, if you were to tell me that people don’t like the idea that unemployment might vary a lot from one period to the next, throwing them out of their jobs, I’d say: well, the corollary of such fluctuations is unplanned inflation in the model, so though they are telling you that they dislike unemployment changes, part of the solution to tackling those, is curiously, to eliminate inflation changes.
I don’t want to pin myself to that mechanism so literally. These conclusions drop out of microeconomic evidence on the elasticity of demand with respect to changes in relative price, and also the particular resting point of the literature regarding what to do about how or if to clear markets when desired prices don’t prevail. Neither of these foundations are particularly strong and may not last. But the main take-away is the difficulty of inferring anything from what people tell you they don’t like in a macro economy.
That said, in a democracy, one can only give so much weight to opinion of a bunch of math-loving technocrats, and, ultimately, policy choices of all kinds, central bank mandates included, have to be anchored to what people want.
A final point on the high weight on inflation in optimal policy in these models: note that the models I referred to before leave out things we suspect may be costly about inflation. Its corrosion of liquid stores of value, its confusion of the process of determining real prices, and, typically, the costs associated with resetting prices.
Moving on: in my last post, I made the point, based on my joint work with Blake and Kirsanova, that delegating levels (NGDP, price level, or whatever in between) targets to central banks that can’t commit isn’t as likely to generate benefits as we thought. SWL worries that this is not relevant. I think it is. In the context of debating the commitment-mimicry benefits to come from NGDP targeting we are asking the following question:
‘Suppose we have a central bank that can’t commit and instead does discretionary policy each period. Will we get better outcomes if we give it the socially optimal policy mandate, or one modified so that it follows a levels target?’ Some, Woodford included, have suggested that perhaps we might.
My paper (BKY) says: not necessarily. You can end up making things no better, and, potentially even worse. Worse or no better, that is, than giving the original optimal policy mandate [a flexible inflation target] to the central bank and telling them to get on with it. Why worse or no better? Because there are at least two equilibria under every scheme, including the original optimal policy mandate, and all possible levels targets one might think of. So, in order to figure out if you are going to make life better by delegating, you need to know where you are starting from, and where you will jump to. (Unless all the start points are worse than all the end points).
Without an argument that can knock out the inconvenient equilibria – and one that leaves the rest of the model intact – we can’t put much faith in the benefits of delegation. If you believe BKY, then you would put more weight on leaving central banks doing what they are doing (inflation targeting under discretion) than trying to squeeze out potentially non-existent benefits from levels targets.
Simon disputes my argument that supporting a levels target – PL or NGDP – requires believing in rational expectations. He notes that RE is a routine assumption in central bank models and argumentation. This is partly true, but caricatures. Note that the Fed’s FRB/US model had the facility to be run under less-than-rational-expectations. Policymakers at the BoE used an RE model, but when they thought it was relevant, would often adjust forecast profiles by hand afterwards to try to offset what they thought were the effects of rational expectations. (Highly unscientific and hopelessly imprecise in doing it this way, but well-intended). However, whether central banks assume RE or not, it’s not a good defence of a regime that it works well in a false world that central banks happen to believe in.
SWL also makes the point that ‘a key argue for inflation targets is that agents are forward-looking’. From what I recall, flexible inflation targeting would be appropriate for most commonly used non-rational expectations assumptions, like adaptive learning, sticky-information, rational inattention [mentions that hesitantly], heuristic-based inflation expectations. Levels targets would not: since there is no point in using the expectation of correcting to a level to stabilise the impact of an initial shock, since there is no such widely held expectation.
It’s true that the forward-lookingness of expectations is often used in discussion of the merits of inflation targeting. But I think that is only done in good faith in the context of arguing against the existence of a long-run trade-off between inflation and unemployment, say in countering the idea that the effects of the financial crisis could be permanently ameliorated by a permanently higher inflation rate. It’s not an argument that should be made to defend inflation targeting against alternatives like price level targeting. The Bank of Canada’s research into this issue, and their deliberations on it, I would take as a good example of the state of the art.