A new Bank of England paper points to a dangerous pitfall for monetary policy in a low rates environment. “Speed-limit rules” stipulate that central banks adjust policies in accordance with the change in the economy (e.g. growth, the unemployment rate, or inflation), not its level, in order to avoid policy errors and anchor expectations. At the zero bound these rules can lead to self-fulfilling recessions, as a downshift in (growth and inflation) expectations triggers less hope for policy easing than fears for subsequent tightening (when the initial downshift is being reversed). In my view this dovetails recurrent fears of convexity or jump risk in low-yield bond markets, and may help explaining why global central banks at the zero bound have so far struggled to produce sustained recoveries.

The pitfalls of speed-limit interest rate rules at the zero lower bound 
Charles Brendon, Matthias Paustian and Tony Yates 
Bank of England Working Paper No. 473: 

“A…common approach…is to study interest rate set ting procedures such as Taylor rules, where the interest rate is raised if inflation exceeds target, or the output gap is positive. A speed-limit rule, by contrast, is a rule where it is decided how far to raise rates based not on the level, but the rate of change of some concept like the output gap.”

Speed limits rules have relied on two arguments:

  • Speed-limit rules seemed to provide a way to insulate central banks from policy errors that occurred through mis-measurement of key concepts like the output gap.
  • Speed-limit rules were shown to be a way for central banks to…stabilise inflation and the output gap…by using the power of inflation expectations to anchor inflation, through making commitments not to simply think afresh as each period and each new shock to the economy comes along.

“There is a chance that rates could end up pinned at the zero bound through self-fulfilling expectations… Normally…if rates followed a Taylor rule with interest rates sufficiently responsive to inflation, and the zero bound were not in play, self-fulfilling recessions would be ruled out…
However, under a speed-limit rule, and faced with the zero bound, agents in the economy would correctly surmise that things will be different. First, rates cannot fall so far to begin with to counter the fall in inflation. And second, agents would forecast that after the initial fall in inflation and opening up of the output gap, the central bank would tighten more quickly. This is because it would be concerned to make sure the output gap does not close too quickly (given its concern for the ‘speed limit’). This means people forecast tighter policy tomorrow, which validates the initial forecast of low inflation. Inflation and the output gap fall, and interest rates are pushed to the zero bound, simply because agents in the economy believe it will.”

A collapse in the value of a target variable today implies it will grow in the future, as part of a subsequent recovery. But if policy is scheduled to `lean against’ this growth then any current collapse will imply expectations that policy in the future will be relatively tight. This in turn places downwards pressure on current inflation expectations, raising the real interest rate. If nominal rates are constrained at the zero bound, these higher real rates cannot be resisted by the central bank, and may themselves be sufficient to support the initial collapse.”

“The main lesson we take from our exercise is that care should be taken in targeting the growth rate of any economic variable (aside from the price level) via a simple policy rule. In the context of the zero bound even rules that appear benign may have detrimental consequences.”

“Two features of the [New Keynesian] model [used to illustrate the problem] are essential for the existence of these crisis episodes: the zero bound on nominal interest rates and a high feedback parameter on output growth in the policy rule. Without a zero bound the policymaker would not permit real interest rates to rise…when both output and inflation are below steady state. Without a large coefficient on output growth, future policy would not be expected to be particularly restrictive as the economy recovers back to steady state, and so the contemporary real interest rate would not rise by enough for the conjectured output collapse to be consistent. “