Comment on How do central banks write? An evaluation of inflation reports by inflation targeting central banks
Workshop in Norges Bank May 5 and 6 2003
by Peter Andrews, Bank of England
I am both honoured and a little embarrassed to have been invited to make this comment. Honoured because I am filling Mervyn King's shoes for a day; embarrassed because the paper is so kind about our Inflation Report that to criticise it seems churlish and to praise it seems immodest.
First, thanks to our hosts for commissioning this paper. Central banks are, by definition, unique in their own country. And although it should be easy to measure whether or not they are hitting a quantifiable objective, such as an inflation target, it is much more difficult to quantify how they are performing with respect to their other objectives. Yet both our stakeholders - the government and the public, to whom we are accountable for the value for money we provide and the competence with which we do our job - and we ourselves, looking to improve what we do, can benefit from measuring how we are doing and comparing ourselves with best practice elsewhere, which often can only mean in other central banks. For these reasons the Bank welcomed both Don Kohn's report, in 2000, on the procedures for providing briefing and analysis for MPC, and Adrian Pagan's 2003 report on where our macroeconomic models stand in relation to international best practice, and published these reports. Inflation reporting, or publication strategy more generally, is another important area, so we welcome a paper which sets quantifiable benchmarks from international experience.
To illustrate the difficulty of quantifying success in this field, this slide shows you one of the Bank's ten strategic objectives for 2002-03, published in our last Annual Report: the IR, and the rest of the publications programme, is intended to build public support. The Monetary Analysis area, which produces the report with MPC sign-off, is setting itself for 2003-04 a performance indicator which relies on outside feedback. But we have not previously had quantitative information of the kind in this paper to draw on, certainly not on any kind of comparative basis. (I understand we will in fact draw on this paper for the 2003 Annual Report! and I expect others here will wish to do the same).
Nevertheless, I'd like to suggest several ways in which this sort of work needs to be taken further. I'll look further at the macroeconomic objectives we can expect an Inflation Report to fulfil, and indeed our publications strategy more generally, and will then look at some alternative measures of success, at least in the UK.
Our Inflation Reports start by setting out two purposes. The first one listed is the contribution that preparing it makes to the quality of the policy process itself. The economic framework associated with the forecast process obliges MPC members to identify and discuss the key current issues as they reach their decisions. The explanatory objective is listed second. And I mention for completeness that publication of the IR is now a legal requirement, so that its part of the machinery by which the Bank is accountable for the way it carries out its remit.
That suggests that one criterion for the success of Inflation Reports is the success of policy itself - inflation close to target, and subject to that strong and sustainable growth. But the spirit of this discussion is to look at the explanatory purpose:
- how clearly understood is the Bank's policy?
- how credible is that policy, i.e. do financial market participants believe that we will carry it out.
The paper takes as a key criterion the extent to which monetary policy decisions are a surprise, measured by the same-day change in 3 month rates. That is a good measure of how much a decision has surprised market participants. But I question how much macroeconomic significance it has. It is rightly argued that rates longer than the policy rate are important for pricing assets. But the three-month horizon is too short to give much information on that front. So let's look at some recent examples of surprising monetary policy decisions. First a small surprise. This is the front end of the government yield curve on the day of the 50 basis point repo rate cut in November 2001. The three-month LIBOR rate, the measure used in the Genberg study, fell by 17bp that day, suggesting that the market had thought the 50bp cut we actually made was possible but less likely than a 25bp cut. The government curve only showed a material fall out to a year or so. Although the extent of the cut had some surprise value, it did not materially change their view of the future path of interest rates. And that was despite the fact that the ECB cut rates on the same day, and the Fed had cut the previous night - all of which should be picked in the same day's change in rates.
In the UK, we can also measure the effect of monetary policy announcements on the inflation expectations implied by the differential between the yields on index-linked and conventional gilts. I would warn against over-precise interpretation of these implied inflation rates - the indexed market is a bit less liquid than the conventional. On the day of that 2001 decision, qualitatively there was just a slight rise in inflation expectations as a result of this surprise rate cut decision. Overall, then, although the Bank's published information - the August IR, minutes of meeting in September and October, etc. - did not allow complete anticipation of the decision, the market's expectation of the key market variables, and the basic credibility of the strategy, were not much affected.
By way of contrast, here is what happened when UK interest rates last changed in February 2003. Measured by the change in 3-month LIBOR, the "surprise" on the day was 20 basis points or so, a bit larger than the November 2001 surprise, but in the same ballpark. But if we look at the front of the spot curve, we see the fall in market rates was much the same at a year as at three months, and the surprise extended along the curve.
And perhaps as you'd expect, the reaction of implied forward inflation expectations was much more marked than in the previous case: up to 13 basis points 5 years out, taking inflation expectations closer to target, though still slightly below it at all horizons. So although the "surprise" element of the decision by the Genberg criterion was not much larger than in November 2001, the effect in terms of the things we care about and write about in the IR was a lot larger. Reactions in the equity market and in sterling were greater, too.
While we look at these market data I should make one further point. Inflation Reports, and other publications such as the MPC minutes, may reduce the surprise entailed by monetary policy decisions. But the corollary is that the publication of that information can itself be an event with the capacity to move markets. Here, for example, is what happened when the minutes of the March 2003 MPC meeting were published. Nominal spot yields rose along the curve by up to 20 basis points.
And inflation expectations rose at the short end and fell further out. Anecdote suggested that the market was influenced by a discussion in the minutes on the pros and cons of a rate rise, even though no member voted for a rise. So here is a case where what the central bank said looks to have had as much or more impact as its actions.
So to summarise: short rates may not be the clearest way of establishing how surprised the market was by a monetary policy decision; and longer-term interest rates and inflation expectations are also of greater macroeconomic significance.
I'll raise some other points for discussion on the empirical study in a moment. But I suggest a completely different measure of the effectiveness of central bank publications strategy: opinion polling.
For over three years, the Bank has had an opinion poll conducted each quarter. We publish the results on our website. We look at the inflation expectations of the general public, but also at their general understanding of monetary policy, which is arguably an important test of our information strategy.
For example, we test understanding of the proposition that higher interest rates bring about lower inflation. We get a positive balance of responses to that, although there is a significant proportion of don't knows. Given that interest rates tend to be high when inflation is high, so that a member of the public observing the casual correlation would come to the wrong conclusion, the result is quite encouraging about basic understanding of how monetary policy works.
It is helpful for the monetary authority to enjoy public acceptance that it is doing at least a reasonable job. That helps to reinforce confidence in the inflation targeting framework, and probably also confidence that inflation will be maintained close to target, which in turn must be helpful for example in keeping pay expectations and pricing behaviour in line. This slide shows that the Bank has enjoyed a generally high level of support over the last four years; the balance of those satisfied against those dissatisfied has always been positive. Support picked up when the Bank was very visibly cutting rates in the autumn of 2001, but the balance was still solidly positive when interest rates were at their high point for the last cycle. As a matter of fact, the Governor typically receives more letters of complaint when interest rates are low, perhaps because the elderly savers have more time to write letters than those with mortgages.
As a light-hearted warning against complacency - and as a way of helping you calibrate the reliability of the previous information - this slide shows that typically at least half the population, asked the question "each month, a group of people meets to set Britain's basic interest rate level. Do you know what this group is?" answer that they don't know. Most people who pick one of the options do at least pick the Bank, or the MPC.
So to summarise, direct polls perhaps get us closer to measuring what we would like to influence through our communications strategy: public understanding of what monetary policy is trying to achieve. But we have to be realistic about the meaning of the results, given the large proportion of the public whose understanding is hazy. It would be interesting to compare notes with any other central banks who undertake comparable studies.
An inevitable limitation of today's paper is that each central bank has a different range of publications, each of which may take on some of the functions suggested for an Inflation Report in today's paper.
By way of example, here's a list of the Bank of England's publication programme on monetary policy issues. The Inflation Report presents a description of current economic conditions, and the outlook, including the range of uncertainty; it briefly summarises the minutes of three previous meetings; and it periodically summarises the forecast record. But it does not debate monetary strategy issues (research on monetary strategy appears in the WPs and QB, and debates in MPC meetings about how strategic principles should be applied of course appear in the minutes). Objections to Committee decisions might well be dealt with in speeches or in the Inflation Report press conference, which is an important event in its own right; Mervyn King and colleagues taking questions for up to an hour. Assuming that other central banks are similar in their diversity, it means, as the authors recognise, that a study which judges only Inflation Reports by criteria which are appropriate for the overall publications strategy, may produce misleading results.
One of the many benefits from this paper is that it will help to spark discussion on what Inflation Reports should cover. Here's a contribution on some other issues arising from the paper.
- It's important that we recognise the limitations of the projections described in IRs. We stress that our projections are probabilistic. We don't publish point forecasts and prefer to spend time on the balance of risks around the projections. That may make it harder for a reader to follow the Report.
- I should stress that the Bank of England's inflation target is continuous through time. We don't wish to become over-focused on the forecast at the two-year horizon; two years is simply a useful horizon, because we believe that most of the effect of an interest rate change today will have come through to inflation by then. But different shocks hitting the economy today will have their impact over a shorter or longer period than that - so policy may have to trade off deviations from target at shorter and longer horizons.
- More generally, the Committee has to exercise degrees of judgement both in constructing the projections and in setting interest rates having made its projections. Neither is a mechanical exercise. The IR should describe the broad forecasting judgements, while the second set of judgements, on the rate decisions, is more appropriate for the minutes, month by month, though the first section of the IR summarises the reasons for the decision that accompanies it. Again, all of this can make the IR less than simple and clear, especially since committee members may differ among themselves on these judgements.
- In order to keep the length of the IR within bounds, there is some information which we don't repeat each time, so that your readers, looking at only one issue, may have felt our coverage incomplete. For example, we look at the forecast record once a year. And the November 1999 IR gave some detail on the convention for projecting the exchange rate, which changed at that time, but we haven't repeated that discussion each time.
- And in the "surprises" exercise, I noted that the range of macro economic control variables is quite limited - just inflation. Given that several inflation-targeting central banks take account of the variability of activity as well, it might have been useful to include such a measure in the exercise.
So my main conclusions would be: this is a very welcome and innovative paper. The exercise could be extended by looking at other measures of expectations - though I recognise not every country has indexed bonds, or opinion polls of the kind the Bank of England pays for. I noted the qualification that Inflation Reports are only part of a central bank's communication on monetary policy, so their contents may not be completely comparable. And there are points at which the message cannot be too clear and simple, lest it be misunderstood: the judgements involved in making economic forecasts and monetary policy are sometimes difficult.