Horizons, symmetry, Kocherlakota and Wren-Lewis

This is prompted by SWL’s mainlymacro post on Kocherlakota’s comments about the Fed’s objectives.

Kocherlakota and Simon want clarification that the target is symmetric. Simon suspects that the asymmetry in the ECB’s objective implied by the words ‘close to but below’ 2 per cent is partly behind the sluggishness of that institution to loosen policy in the face of weakening inflation and activity. If I were allowed to amend the ECB’s statement of its own objective, I would certainly remove the ‘close to but below’ words.  But my sense is that they were a textual sop to monetary conservatives, and not any longer a significant influence on monetary policy.

However, whether symmetry is desirable is debatable. From an academic perspective, symmetry follows only from the fact that we choose to approximate what would be in the interests of agents in the model with a function that is symmetric.  From a practical perspective, it’s clear to me that deviations below target pose a greater risk to monetary stability than deviations above, because they involve the potential for the conventional policy instrument to get trapped at the zero bound, and require costly, unreliable discretionary fiscal stimulus as a helping hand, and recourse to unconventional measures of still uncertain merit.

Simon comments on the proposal that the objective be clarified such that it’s understood that the Fed targets two year ahead forecast inflation. Simon characterises the BoE as having ‘operated until quite recently what appeared to be exactly this two year benchmark’, plotting a chart that shows the two year ahead forecast pretty close to target through the years of the great moderation.

In my view Simon overstates the influence of the two year horizon.  Policy was conceived of like this in the early days of independence.  Partly because of an exaggeration of the generality of an early paper of Svensson’s that assumed there was a two year lag between a policy change and a change in inflation.  But quite soon afterwards, and particularly once senior people became aware of the unrepresentativeness of that Svensson paper, communication began to stress that the horizon was elastic, and, in fact, depended on the shocks hitting the economy.  During the great moderation years the horizon, if there was one, was less binding.  Shocks were smaller, so the economy would be projected to return to base quicker; and there were fewer trade-off-inducing shocks, requiring fewer conscious forecast deviations from target. It took the crisis to make use of the elasticity always stressed in the horizon.

Simon worries that ‘the 2 year horizon came quite close to having a very damaging impact in the UK’. It might have looked like that from the outside, but inside the Bank of England, I don’t anyone worried about this at all.

Simon also overstates - I think greatly - the importance of the published forecast in the policy process.  That policy follows from the forecast would seem natural, and is required for Simon's chart of the time series of forecasts to be useful in diagnosing anything about the regime.  But it didn't work like this.

To explain:  very frequently, one could note MPC members operating with the following logic during the forecast round.  ‘For better or worse, we have got ourselves into the situation where we have to publish a forecast, and people are focusing on the two-year ahead forecast relative to target.  However, I don't believe the forecast coming out of the model, and I don't always understand it anyway.  So, I will lean on the forecast judgements to ensure that whatever policy decision I prefer the forecast will be equal to target at that horizon.'  If this were always true, the two-year horizon would not be a constraint on anything, ever.

I think encoding a horizon is problematic, for two reasons.

First, I speculate that the influence of the two-year horizon, such as it was, was to contaminate a forecast process [what's going to happen to the economy and what should we do about it?] and reduce it to a reverse-engineering process [what are we going to do, and how can we build a forecast so that it looks like that is what we should do?]

Second, in principle, the horizon should depend on what is driving the economy away from target/base. And our understanding of the appropriate horizon is very incomplete.  A crude summary of the key problem is this: the models we have that are used to describe the economy have a much stronger mean-reversion than the actual economy, and imply fast returns to base under optimal policy.  Most policymakers and model practitioners operate with the hunch that this property should be overridden, meaning that time away from target should be more protracted.  But how much more?  I don't think we are ready to fix a horizon in our monetary frameworks.