Central Bank Interest Rate Forecasts: It’s not about accuracy

                Charles Goodhart writes (FT-£) that “Proposals that central bankers report their expectations of official rates, beyond some short future horizon, are retrograde, pushed forward by fashionable theory without reference to empirical reality.” By short horizon here I think he means three or six months: much shorter than the recent move by the US Fed. Charles is usually right about most things, and similar comments have also been made by other experienced ex-central bankers, so I think it is important to set out carefully one important counter argument.

                Charles writes that “whether the publication of central bank predictions of the future path of interest rates is likely to be beneficial depends on the relative accuracy of such forecasts.” I disagree. I’m happy to assume they are no better than those of forecasters in general. What is then gained by publication?

                The key point is that central banks, or members of a central bank committee, have inside knowledge. Not about how the economy works, or of statistics a few days before they are published, but about themselves. Their forecasts for interest rates tell us what they are likely to do if events (inflation, growth etc) turn out as they expect, and if they are being consistent. So the important question is whether this information is useful.

                Central to the academic case for delegation of interest rate decisions to central banks is that they are less susceptible to the temptations of time inconsistency. (Those familiar with what this means can skip this and the next paragraph.) A classic example involves the trade-off between inflation and output. Suppose the monetary authority wants zero inflation, but would like output above the level consistent with zero inflation (the ‘natural’ level of output). Suppose the public share those preferences. The monetary authority announces a policy of achieving zero inflation, and people form expectations on that basis. Once those expectations have been set, the monetary authority realises that outcomes would be better if output was higher than the natural rate. This will raise inflation above zero, but given their and the public’s views on output, everyone would be better off with a little bit more inflation and higher output.  This is sometimes called reneging or cheating by the policy maker, but notice that everyone is apparently better off when it changes its mind.

                Now, if people are smart, they will know the monetary authority will behave this way, so they will not believe the initial announcement of zero inflation. In fact the only situation in which expectations prove correct is when they are so high that it is no longer in the monetary authority’s interest to renege. So the outcome is high inflation with no gain in output. In this situation it would be much better if the monetary authority could commit to zero inflation and not renege on this commitment. It is generally thought that central banks are more likely to be able to do this than politicians, in part because politicians will be too tempted by short term gains, and will discount the longer term repercussions.

                Being an independent central bank may increase your ability to commit to a policy in the face of the time inconsistency temptation, but it does not guarantee that result. Central bankers would like to be able to establish their credibility for commitment. (I gave a particular example when it might matter a lot here.) Publishing interest rate forecasts allows them to do this. If events turn out roughly as expected, yet the central bank does not follow its own forecast for interest rates, it might be because they are exploiting these time inconsistency possibilities. Following your own forecasts when nothing changes does not prove that you are resisting such temptations, but it is consistent with doing so.

                Now Charles Goodhart might respond, in the spirit of his article, by saying that events never turn out roughly as expected, and so interest rate forecasts can never be used in this way. However the critical question is as follows. Is the fog of news so dense that we can never say anything useful using these forecasts? I think not. For a start, we not only have the bank’s interest rate forecasts, but also the inflation forecasts that go with them. We often have no problem saying that the balance of news about the economy over some period is – say – that activity is stronger than expected, and so inflation is expected to be higher. If, despite this, the central bank reduced interest rates compared to their own forecasts, we would be suspicious that they were being inconsistent. However, if we do not know what they were expecting to do with interest rates (because these forecasts were not published), then we are no wiser about whether they are being consistent or not.

                Charles talks about the importance of putting error bands around forecasts in general (the fan charts), and I could not agree more. However in this particular case, we do not need such bands. To the first approximation, we only need to know what their best guess was for future interest rates, and the associated best guesses for inflation and other variables. This is because we are using the interest rate forecast to judge consistency, and not as a forecast per se.

                Now perhaps the importance of time inconsistency problems in monetary policy is not as great as implied by the academic literature.  However, I think the chances of the public being misled by the publication of interest rate forecasts is so small that on this occasion it is worth taking this academic idea seriously. I believe that in ten years time, when more central banks publish interest rate forecasts in this way, we will wonder what all the fuss was about.