Alex Haberis, Richard Harrison and Matt Waldron.
In textbook models of monetary policy, a promise to hold interest rates lower in the future has very powerful effects on economic activity and inflation today. This result relies on: a) a strong link between expected future policy rates and current activity; b) a belief that the policymaker will make good on the promise. We draw on analysis from our Staff Working Paper and show that there is a tension between (a) and (b) that creates a paradox: the stronger the expectations channel, the less likely it is that people will believe the promise in the first place. As a result, forward guidance promises in these models are much less powerful than standard analysis suggests.
Introduction
Economists use the term ‘forward guidance’ to describe central bank communications about the likely future path of policy rates. Forward guidance is not new: central banks have been providing guidance for decades in the form of forecasts for inflation and GDP growth. What is new is that, since policy rates have been stuck at or near their lower bounds, central banks have made increasing use of explicit statements about their intentions regarding future policy.
Why have they been doing this? It is crucial to draw a distinction between two different motivations (see also Simon Wren-Lewis). The first is to convey information about how the central bank intends to set policy to meet its objectives (its ‘reaction function’). The aim is not to stimulate the economy (though that could be a by-product), but rather to make policy more effective by guiding people’s expectations to be more consistent with the central bank’s intentions.
A second motivation is to stimulate the economy by promising to keep rates lower in future as a substitute for cuts today, which may not be possible due to a lower bound. This differs from the ’reaction function clarification’ motive because it requires the central bank to promise above-target inflation in the future, which is a promise to deviate from its usual reaction function.
We focus on this second ‘promise’ motivation for forward guidance. In our view, forward guidance announcements made by central banks to date have been the former “reaction function clarification” type rather than the latter “promise to deviate” type. Accordingly, our post is a comment on the consequences of implementing such promises in textbook monetary models rather than a commentary on what central banks have actually done.
What is the forward-guidance puzzle?
Textbook models deliver unreasonably large macroeconomic responses to forward-guidance promises, a phenomenon described as the ‘forward guidance puzzle’ by Del Negro, Giannoni, and Patterson.
To illustrate the puzzle, we draw on analysis from our paper. We use a textbook model of the output gap, inflation, and the policy rate in which expectations about the future play a large role in households’ and firms’ decision making. Policy is set optimally under discretion: the policy rate is set to minimise welfare costs arising from (current and future) deviations of output from potential and inflation from target; (ordinarily) the policymaker cannot commit to future policy actions that would be inconsistent with minimising welfare costs at that time.
We construct a scenario in which a large shock forces the policy rate to the zero lower bound (ZLB). The shock creates a deep recession because policy cannot be eased enough to offset it, given the ZLB constraint, and the presumed absence of other policy measures, like asset purchases. We then examine attempts by the policymaker to improve outcomes using forward-guidance promises to hold rates at the ZLB for longer than implied by their usual reaction function.
In Figure 1, the solid black lines show the baseline scenario, in which the policy rate is constrained at the ZLB for 7 quarters following the shock. The dotted blue and dashed red lines show the effects of promises to hold rates at the ZLB for an additional 3 and 4 quarters respectively.
Figure 1: effects of fully-credible forward-guidance policies
Notes: The model starts in its steady state and is hit by a large recessionary shock. In the baseline (black lines), interest rates are lowered to the ZLB, where they remain until quarter 7. The other two lines show the effects of two fully credible forward-guidance policies: a promise to hold interest rates at the ZLB for an extra 3 quarters (blue lines); a promise to hold interest rates at the ZLB for an extra 4 quarters (red lines).
The simulations illustrate that forward guidance in standard models can reduce the recession resulting from the combination of a bad shock and the ZLB. They also imply that a marginal extension in the duration of the promise (from 3 to 4 quarters) can have a large effect.
A policy that could successfully lower long-term real interest rates could stimulate activity and inflation. The question is whether the effect is sufficiently powerful to constitute a puzzle? Del Negro, Giannoni, and Patterson use the Federal Reserve Bank of New York’s model – a more empirically-reasonable model than the textbook model used above – and show that its responses to policies of this sort are much stronger than equivalent empirical estimates. Given that the transmission mechanism underlying this result is common to most monetary-policy models, the forward-guidance puzzle is also common to most models in that class (see e.g. John Cochrane).
This puzzle has prompted economists to develop models that deliver more empirically reasonable responses to forward-guidance promises – see, for example, John Cochrane’s discussion of a recent paper by Xavier Gabaix. We take the complementary approach of retaining the textbook model, but challenging the assumptions underpinning the experiment. We argue that the puzzle contains a paradox, which our analysis illuminates.
The time-inconsistency of forward-guidance promises
Our simulations show that the cost of using forward guidance to reduce the size of the recession is an overshoot of the inflation target after the recession has ended. At the time of the announcement, this cost appears worth paying because the policy ameliorates the recession. As time passes, however, the benefits of the smaller recession become history, leaving only the cost of the overshoot. At this point, the policymaker would prefer to tighten policy to bring inflation back to target.
This well-known ’time-inconsistency’ effect is shown in the lowermost panel of Figure 1. The promises to hold rates at the ZLB for 3 and 4 additional quarters both improve welfare immediately after the announcements. But, shortly thereafter, both promises deliver worse outcomes than in the baseline scenario in which there was no forward guidance policy. At this point, the policymaker has an incentive to renege on their original promise.
This is sometimes called ’Odyssean’ Forward Guidance. The policymaker needs to commit to stay the course, just as Odysseus resisted the sirens’ calls by having himself bound to the mast of his ship. In the real world, central banks do not have the ability to tie themselves to the proverbial mast, so the assumption that the policymaker would definitely follow through with the promise is questionable.
Imperfectly-credible forward guidance promises
We can incorporate scepticism about the policymaker’s ability to commit into our simulations by assuming that the private sector attributes some probability to the policymaker reneging whenever they have an incentive to do so (as measured by welfare). We assume that the probability is increasing in the size of the renege benefit Figure 2 shows the results.
Figure 2: effects of imperfectly-credible forward-guidance policies
Notes: The model starts in its steady state and is hit by a large recessionary shock. In the baseline (black lines), interest rates are lowered to the ZLB, where they remain until quarter 7. The other two lines show the effects of two imperfectly-credible forward-guidance policies: a promise to hold interest rates at the ZLB for an extra 3 quarters (blue lines); a promise to hold interest rates at the ZLB for an extra 4 quarters (red lines). The lowermost panel shows the per-period probability that the policymaker reneges on the announced promise.
Relative to the perfect-credibility case (Figure 1), the most striking difference is that the two forward-guidance policies now have very similar economic effects. The attenuation of stimulus is larger in the case of the longer-duration promise because the larger overshoot associated with that promise increases the policymaker’s temptation to renege on it, making it less credible (lowermost panel). Attempts to add more stimulus by increasing the duration of the promise yet further would be futile in the absence of more credibility – see our paper for further discussion.
This imperfect-credibility mechanism is not just a theoretical possibility; it has also been recognised by policymakers (see Nakata for a review).
Imperfect credibility and the forward guidance puzzle
What does this mean for the forward-guidance puzzle? Figure 3 shows simulations of the same 4 period lower-for-longer policy assuming imperfect credibility in two alternative versions of the model: the one used above; and a version calibrated so that real interest rates have a larger effect on activity.
By construction, the forward-guidance promise is much more powerful under perfect credibility in the more interest rate sensitive model. But the effects of the policy are very similar in the two models when we relax the perfect-credibility assumption. There is a link from the potential power of forward-guidance promises to their credibility: the greater the potential power of the promise, the less likely it is to be credible. The result is that a forward-guidance puzzle may no longer exist.
Figure 3: effects of imperfectly-credible forward-guidance policies in two alternative models
Notes: The model starts in its steady state and is hit by a large recessionary shock. In the baseline (black lines), interest rates are lowered to the ZLB, where they remain until quarter 7. The effects of a forward-guidance announcement to hold interest rates at the ZLB for an extra 4 quarters are shown for the base model (green lines) and an alternative model in which demand is more interest sensitive (red lines), under full credibility (dashed lines) and imperfect credibility (solid lines). The lowermost panel shows the per-period probability that the policymaker reneges on the announced promise in the imperfect credibility case.
Conclusions
The forward-guidance puzzle contains a paradox. It is the combination of a strong effect of expected future real interest rates on current activity and the assumption of perfect credibility that gives forward-guidance promises such power in textbook models. We have highlighted an internal inconsistency at the heart of the puzzle because a more powerful transmission channel is likely to go hand in hand with a less credible promise.
Using forward-guidance promises to stimulate the economy is an idea dating back to Krugman’s 1998 paper. The likelihood that a forward guidance promise of this type would lack full credibility in the absence of an appropriate commitment mechanism is one reason why central banks have not attempted to use forward guidance in this way.
Alex Haberis, Richard Harrison and Matt Waldron works in the Bank’s Monetary Analysis Division.
If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.
Comments will only appear once approved by a moderator, and are only published where a full name is supplied.
Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.
