JT claims that the Taylor Rule emerged from ‘two decades of research on optimal policy’. I think it’s important that John does not allow this to be read by WSJ readers, who might not have encountered any of the actual research, as ‘after two decades, we figured out that the Taylor Rule was optimal policy’. Because that would not be right.
As I wrote in my last John Taylor piece, Taylor Rules do a pretty good job, in a narrow range of DSGE models, but they fall short of optimal policy even there. Moreover, it’s worth re-emphasising just how much the financial crisis ought to shake our faith in this pre-crisis research, which almost exclusively ignored banking and finance. We have no idea, really, in model terms, how Taylor Rules do in a financial crisis. There’s an inkling that models that progress is made by augmenting TRs with a term in spreads, a modification that can take the presciption a long way from the original. But given our lack of confidence in the tools we have so far to articulate crises [most of them don’t and can’t articulate crises], this is only an inkling.
Even if we leave aside the banking and finance problem, there were some interesting connundrums that JT’s one sentence research summary leaves out. For example, if you take a modern macro model and work out what is the optimal Taylor Rule – tune the coefficients so that they maximise social welfare, properly defined in model terms, you will get very large coefficients on the term in inflation. Perhaps an order of magnitude greater than JT’s. This same result is manifest in ‘pure’ optimal policies, where we don’t try to calculate the best Taylor Rule, but we calculate the best interest rate scheme in general. In such a model, interest rates are ludicrously volatile. This lead to the common practice of including terms in interest rate volatility in the criterion function that we used to judge policy. Doing that dials down interest rate volatility. Or, in the exercise where we try to find the best Taylor Rule, it dials down the inflation coefficient to something reasonable. This pointed to a huge disconnect between what the models were suggesting should happen, and what central banks were actually doing to tame inflation [and what John Taylor was saying they should do]. JT points out that most agree that the response to inflation should be greater than one for one. But should it be less than 20? Without an entirely arbitary term penalising interest rate volatility, it’s possible to get that answer.
JT recommends in his WSJ article that rather than flipping to quantitative easing, central banks adopt a ‘fixed money growth rule’. WSJ readers might well like the sound of that. Because it has a ring of ‘sound money’ about it.
But those working in monetary policy research would be perplexed to hear such a rule recommended. What does it even mean, anyhow? If we were in a situation where there was a pure liquidity trap, with interest rates not only at but expected to be zero indefinitely, we know [from the same models JT invokes to support his other cases] that money growth at whatever level has no effect on anything. Flipping to a fixed or variable rate money growth rule does nothing at all. Suppose, differently, we are in a situation of a temporary liquidity trap, where interest rates are expected to be positive again after a period of time. Presumably, JT, the master of interest rate rules, would like policy to resume adherence to the rule when later possible. In which case, the amount of money growth induced while at the zero bound is irrelevant. It only becomes relevant if the central bank, when conditions allow rates to lift off the zero bound, is expected to accommodate the extra money in terms of rates being lower than the TR prescription by the mount required by the nature of money demand. So money growth does something only if we make a credible commitment to ditch the Taylor Rule later. JT isn’t advocating that. What is he advocating? I am mystified.
JT also makes reference to the idea that the zero bound is not a problem for the Taylor Rule. He has stated this a few times. He refers to a 2000 paper by Reifschneider and Williams which offers ‘a rules-based approach to deal with the problem’. This reference is perplexing. The zero bound cannot be overcome by anything suggested in that paper. It’s a constraint, and policy has to live with it as best it can. If the natural rate falls to 7 or 8 percentage points below zero [a common guess in central bank circles in the heat of the crisis] some pre-crisis tinkers to the Taylor rule cannot ‘deal with’ this problem. Moreover, it deserves repeating that undue adherence to the Taylor Rule has been shown to open up the possibility of entering self-fulfilling traps at the zero bound. This could be just a theoretical curioso. But, to the extent that it is – to the extent that the model in which it occurs fails us – we should also regard the research advocating Taylor rules as a curioso too.
Let me return, finally, to the matter of the coefficient on the Taylor Rule, which JT says we can agree should be greater than 1. This really captures the sense in which for JT monetary policy design is done and dusted. Well, I don’t think things are like that any more. As I wrote before, John Cochrane explains that the equilibrium selection that is going on in rational expectations studies of alternative Taylor Rules may be a load of nonsense. What is JT’s position on that? Williamson, Wright and others think that the model of money in the background (of the model JT has in mind when he invokes the >1 coefficient result) is a load of nonsense. And then there’s the issue of how sticky prices are modelled. And heterogeneity….
Really, we are nowhere near agreeing that anything like the Taylor Rule is optimal policy. The crisis, and wave after wave of powerful critiques from other strands of mainstream macro, but outside the policy rule research agenda have revealed that these old received wisdoms are much less useful than we thought.