Podium: Forecasts aren't a map of the future

From the Sir Alec Cairncross Lecture given by the former chief economic adviser to the Treasury
Click to follow
The Independent Culture
MY TOPIC this evening is the use of forecasts in economic policy- making and their use by the Bank of England's Monetary Policy Committee, or MPC.

The context in which I shall be discussing economic forecasts is the control of inflation. I hope it is common ground that inflation responds to economic developments with a lag (although there are some changes, eg changes in indirect taxes, which can affect the price level immediately).

Thus policy actions taken today will affect inflation over a period of up to two years or more. It follows that anyone taking policy actions to control inflation must, at least implicitly, be thinking about the effects on inflation in the future; ie, they are forecasting.

I should emphasise that the following account represents my own views, which may or may not be shared by my colleagues. We are all individually responsible for our own decisions and are free to use the forecasts to inform our decisions in the way we think best.

The immediate point is that we do not rely wholly on formal methods to reach decisions. The econometric model that is used in the Bank to help produce the forecasts uses about 150 economic variables; but each month the MPC is presented with information on a thousand or so variables.

If we believe that the variables that are not included in the model are relevant to our decision, we obviously want to take them into account. There are considerable advantages to using the forecast as a framework for doing this.

For one thing, it gives us a language with which we are familiar. We are used to watching an economy unfold and thinking about how it will develop. We also have some ideas about the normal ranges within which an economy will behave (although, of course, we can all be surprised).

In fact, the Bank uses a suite of models, and we use our intuition and theoretical understanding to decide which particular model is appropriate to thinking about the current circumstances.

We can also develop ad hoc models to explore particular developments - for example, the implications of a possible credit crunch.

We go from one set of observations to a formally produced forecast, and we then adjust that forecast in the light of all other information that we believe is relevant.

The forecast represents the collective view of the MPC members. It is not necessarily anyone's individual forecast, nor is it the average of the individual forecasts.

If the MPC were told to go away until it had produced a single forecast, this is the forecast it would produce after discussion and compromise.

Does the forecast imply a decision? The practical answer is, clearly not. At the time of the February inflation forecast, the committee was divided four-four on whether interest rates should be raised. At the time of the August inflation forecast, seven members voted for no change, one voted for a cut and one voted for an increase. So it is clearly possible to produce a collective forecast and yet disagree about the interest rate decision.

The Government has given the MPC a function that is not precisely specified. The objective is clearly defined: the target "at all times" is a 12-month inflation rate of retail prices, excluding mortgage interest payments, of 2.5 per cent. But the Chancellor's letter setting out the remit also says:

"The framework takes into account that any economy, at some point, can suffer from external events or temporary difficulties, often beyond its control. The framework is based on the recognition that the real inflation rate will, on occasions, depart from its target as a result of shocks and disturbances. Attempts to keep inflation at the inflation target in these circumstances may cause undesirable volatility in output."

There is room for disagreement about "undesirable volatility". There is also room for disagreement about whether policy should be directed at the mean, the mode or the median of future inflation. So, even if there is agreement about the forecast, there is no simple mapping from the forecast to the policy decision.

Do we have an inflated view of our ability to control inflation? The Inflation Report shows the probability distribution of the inflation forecast. The August Report suggests that the 50 per cent confidence band for inflation in the third quarter of 2000 is about 1.9 per cent to 3.3 per cent.

That means there is a 50-50 chance that inflation will lie outside that range.

I do not know whether that is excessively modest or excessively boastful - but we are certainly not claiming that we know precisely what level inflation will be in two years' time.