October 22, 2017

Naïve correlation

Published in The Ecologist Vol. 5 No. 1, January 1975.

One of the failings of the inductive principle which underlies the experi­mental method is that one is not provided with the means of distinguishing cause from effect. If (a) is associated with (b) purely on the basis of empirical correlation, we are making a gratuitous assumption in stating that (a) is causing (b), for (b) might easily have caused (a) or both (a) and (b) might have been caused by (c), or – what is most likely (if they are part of the same complex eco-system) – the relationship between the two, which has led them to appear together, is of so complicated and indirect a nature that it can only be understood by examining the inter-relationships between several thousand, or even millions of sub-systems, among which figure (a) and (b).

The only way to establish cause and effect relationships is by examining different hypotheses in the light of a model of the system involved, i.e. by making use of the deductive method, or, better still its modern version: simulation.

It is quite surprising to what extent totally uncritical correlations backed by no theoretical material of any kind have been allowed to influence public policy in this country. A striking example are the theories of Professor Nicholas Kaldor, developed to explain the slow rate of economic growth in the UK during the 1950s and 1960s, which were the basis of much of the economic policy of the Labour Government during this period.

In fact, they provided a justification for:

  1. the large scale immigration into this country of recent years;
  2. the now defunct Selective Employment Tax; and
  3. the encouragement of mergers in industry, in particular through the medium of the IRC set up by the Labour Government for this purpose.

Kaldor developed his theory in the following way: he looked through tables of statistics and found that countries whose economy had expanded rapidly over the last few years had certain things in common which they did not have with countries whose economy had not expanded. These were:

  1. Large scale production; an economist, Verdoorn, linked the rate of increased productivity with the rate of increased output. On the basis of this study, productivity, Kaldor maintained, was the result of large scale manufacturing, not the reverse. If companies were big enough, argued Kaldor, productivity would automatically be increased.
  2. Manpower; he showed that countries with a high economic growth recruited their labour force from agriculture. Employment in manufacturing increased to the detriment of that in agriculture. The UK, he argued has reached a stage of “premature maturity” and had no more cheap labour to attract from agriculture.
  3. Growth of manufacturing industries; Kaldor remarked that in countries whose economies had expanded, the manufacturing industries, what he called “secondary occupations”, had expanded faster than services, what he called “tertiary occupations”.

On the basis of these three sets of correlations, it became Government policy to create vast industrial giants to import cheap labour from abroad and to divert labour from service industries into the manufacturing ones.

Anyone vaguely in touch with reality will simply laugh at the naivety of Kaldor’s conclusions. Not so our Government, which swallowed it all – hook, line and sinker. Needless to say, in spite of all the measures taken in the direction indicated, there was no improvement in the rate of growth of British industry.

What is of interest to us here is the methodology used. What Kaldor did was simply to single out a few isolated factors which it was possible to associate empirically with a growing economy in a number of instances too small to constitute a s significant sample, and then to postulate that the introduction of a these factors into a non-growing economy would bring about the desired effects.

Needless to say, to build a model of an expanding economy would require taking into account a very large number of factors, most of which would probably not fall with-in the field of study of present day economics. If one did this, one could either find that the factors noted by Kaldor were irrelevant, alternatively that growth in the countries concerned occurred in spite of them, alternatively that they were the consequence rather than the causes of growth.

Similar uncritical correlations have led Professor Zuckerman to boast of the achievements of modern medicine. Thus he writes:

“The major killing and disabling diseases which used to plague the world – malaria, tuberculosis, poliomyelitis, typhoid, cholera and the simple fevers of childhood like measles and diphtheria – have been eliminated over the past two to three decades in many countries.”

Zuckerman assumes that these trends will last, without having built any model of the conditions that actually determine the outbreak of epidemics and the spread of diseases. If he were to do so, he would probably find that these diseases have not been eliminated at all. Technological solutions have in cases obtained for us a temporary respite, at a very considerable long term cost. Thus a short while ago we were all convinced that venereal diseases had been stamped out in industrial countries; not only is it recurring, but it has become an epidemic with 2.2 million cases in the US last year, and 150,000 in the UK.

The same is true of Zuckerman’s contention that we shall always be able to feed the world and otherwise accommodate the growing population: statements made entirely on the basis of the ridiculously short experience of industrial society. Also, during the “Limits to Growth” controversy, Maddox never tired of pointing out as proof of our ability to feed the world the fact that as a result of the Green Revolution, India was actually exporting wheat for the first time in its history. As it happened, the very next year it had to import vast quantities from America, while a year later it was faced with famine on a massive scale.

Let us consider a few ridiculous examples which might help reveal the inadequacy of current scientific methodology. Supposing we establish that most of the people driving Rolls Royces are chairmen of large companies; on the basis of a naive correlation one could conclude that it sufficed to buy a Rolls Royce in order to become the chairman of a large company. The behaviour pattern we would then adopt would be no more absurd than that adopted by our Government on the basis of the advice of its ‘experts’.

Consider yet another example. Imagine a man from Mars being invited to watch the Derby at Epsom. Being unable to communicate with his hosts, and never having seen a spectacle of this sort before, he might quite easily come to some very odd conclusions on the basis of empirical correlations.

Thus he would notice that a vast crowd of people was being entertained by 20-odd horses and 20 very small men in gaily coloured costumes. He might consequently infer that the English had a predilection for being entertained by very small people. If he did not make-the essential, but not all that obvious, connection between the smallness of the men, their lightness and the reduced strain imposed on the horse (a connection which cannot be made empirically, but only by building a model of the horse race), he might be led when arranging for the future entertainment of his hosts on Mars, to go to great lengths to find diminutive jugglers, conjurors, musicians and dancing girls. Our Martian visitor might easily come to other equally absurd conclusions on the basis of empirical correlations.

For instance, he might notice that those gamblers queuing after a race to obtain their winnings in front of a book-maker’s stall wore a very happy, if not jubilant expression on their faces. He might quite easily infer from this that the English had a particular predilection for queuing and that few occupations would be calculated to confer upon them a greater degree of happiness. This consideration might equally well affect the arrangements he might make for their entertainment.

Once more, if he had built a model of a horse race, and understood the mechanism of betting, he would be able to infer from the happy expression on the faces of the queuers that they had backed the right horse. He would come to the happy conclusion that it is not queuing that renders the Engish so jubilant, so much as making money on horses.

I think we have every reason to suppose that the model-building Martian would be capable of organising a more satisfactory stay for his English visitors than the empiricist. His model would enable him to understand the function of the various things he had seen (i.e. the variables of his models would be teleonomic classifications). He would thus be capable of finding substitute entertainment for his English visitors, which, although not empirically similar, would satisfy, e.g. would fulfil, the same psychological functions as the horse race which he may not be able to replicate.

The uselessness of correlations that are not supported by an adequate body of theory, i.e. of conclusions not based on the examination of a situation in the light of an interdisciplinary model, is pointed out by Norman Moore in an article in Nature. In considering the effects of organochlorine pesticides on bird life he writes:

“The decline in the bird of prey population is noted after the introduction of the chemical and is then attributed to it. If conclusions were based entirely on correlations of this kind, they would have very little value, but if other relevant information is taken fully into account – notably detailed consideration of alternative hypotheses – knowledge of the biology of the species and of the toxicology of the chemical to it or closely related species, one hypothesis can be shown to be greatly preferable to others. In the absence of the crucial experiment this is as far as the scientists can go, but the element of doubt can be extremely small. The situation is a familiar one in ecological research There is very good evidence that the reproductive capacity and population size of some birds of prey have been severely affected by organochlorine insecticides. The evidence is accepted by a wide range of workers in many disciplines, and by almost all who have first-hand knowledge of bird populations. No serious alternative hypothesis has been advanced to account for the remarkable phenomena observed in the field.”

Unfortunately, our Government experts do not proceed in this way, and the Government has made no move to ban these poisons any more than it has to deal with any of the other real problems which could be shown to face our society were they to be examined in the light of an objective non-disciplinary model of its relationships with its environment.


  • Twitter
  • Facebook
  • Digg
  • Reddit
  • StumbleUpon
  • Diaspora
  • Identi.ca
  • email
  • Add to favorites
Back to top