Published 26th February 1969

Sir,-Your editorial of February 5 argues convincingly that various influences on the money supply are unpredictable. You conclude, however, that monetary policy is therefore an art rather than a science. May I describe how engineering science does in fact control the unpredictable - and how it could do so in monetary policy?

To be stable an engine must be able to operate steadily at very close to a specified speed. It must be able to do this when the load on the engine is steady; but it must also be able to do this when the load on the engine is fluctuating rapidly and unpredictably. The control system achieves stability by receiving a continuous feedback of the right data, with almost zero time lag, and then by putting in immediately a continuous adjustment to the fuel supply. The control system characteristics, which determine the adjustment to be made in the light of the information in the feedback, are determined in advance either by "physical argument" or by mathematical analysis.

Similar principles can be applied to the economy. In the economy there are several variables to be controlled, but fortunately they may be considered one at a time. In the present discussion we are concerned with the average of retail prices, as the variable to be controlled, and the supply of money as the variable with which to control it. In this perspective the effects which you mention, that is, unpredictable government borrowing in order to finance movements in the gold reserve, and variable preference of the public to be liquid, merely represent a future unpredictable and varying load.

In order to take such effects in its stride the control system must have the following five features. 1. It must receive a feed back of data (from the economy) that is almost continuous. 2. The data feedback must have an almost zero time lag. 3. The data to be fed back must be the right choice of data. 4. The internal characteristics of the control system must be inherently stable. 5. The control output must be almost continuously adjusting the money supply.

In scientific terms the first two items in the above list would be termed "instrumentation," or "high response rate instrumentation". These are scientific subjects in their own right. Instruments have to be specially designed to give a high response rate, that is, to eliminate almost all the time lag, but for control systems the problem is usually not too difficult because a "relative" rather than an "absolute" feedback is required. The third item in the list requires some analysis of the economy as well as some understanding of the theory of control or of oscillations. If, in the economy, an "incorrect" amount of money is in circulation then the stock of goods available for purchase will begin to increase or decrease at a rate which is different from the "correct" rate. When stocks have become significantly greater or significantly less than their "correct" level, rates of production may be changed and prices may be changed. The important point here is that prices are likely to have a time lag in their response to an incorrect money supply and may not therefore be adequate as a feed back. On the other hand, the stock of goods available for purchase, minus the outstanding order for goods, will show a fast response to an incorrect money supply. In item 3, above, therefore, the right data must include both the average price level and the net stock of goods.

In order to obtain inherently stable internal characteristics for the control system, as in item 4, each change specified by the control would probably contain four contributions. The first contribution would be designed as a "spring" type mechanism always tending to bring the average price back towards the correct level, the second would be a spring type mechanism to bring net stocks back towards the correct level, the third a "damping" mechanism to dampen oscillations and reduce overshoot by opposing rates of change of price, and the fourth a damping mechanism to dampen oscillations and reduce overshoot by opposing rates of change of stocks.

Item 5 reinforces the argument for using the money supply, rather than taxation, as the short term control. The money supply can readily be adjusted each working day. The adjustment would usually be very small. With the above control system it should be practicable to keep the average price level almost rock steady.

Your conclusion that monetary control is an art rather than a science therefore seems to be a result of sharing the widespread lack of appreciation of the power of scientific and engineering control systems. Much more than just a monetary control system is required in our economy, but a monetary control system would be a very good start.