Subscribe to BrookesNews’ Bulletin
The validity of statistical methods in economics
Dr Frank Shostak
In an attempt to be seen as scientific modern economics in addition to sophisticated mathematics also employs probability distributions. What is probability? The probability of an event is the proportion of times the event happens out of a large number of trials. For instance, the probability of obtaining heads when a coin is tossed is 50 per cent.
This does not mean that when a coin is tossed 10 times, 5 heads are always obtained. However, if the experiment is repeated a large number of times then it is likely that 50 per cent will be obtained. The greater the number of throws, the nearer the approximation is likely to be.
Or say it has been established that in a particular area, the probability of wooden houses catching fire is .01. This means that on the basis of experience, on average, 1 per cent of wooden houses will catch fire. This does not mean that this year or the following year the percentage of houses catching fire will be exactly 1 per cent. The percentage might be 1 per cent or not each year. However, over time, the average of these percentages will be 1 per cent.
This information, in turn, can be converted into the cost of fire damages thereby establishing the case for insuring against the risk of fire. Owners of wooden houses might decide to pool their risk, i.e., spread the risk by setting up a fund. In other words, every owner of a wooden house will contribute according to a certain proportion to the total amount of money that is required in order to cover the damages of those owners whose houses will be damaged by the fire.
Note that insurance against the fire risk can only take place because we know its probability distribution and because there are enough owners of wooden houses to spread the cost of fire damage among them so that the premium will not be excessive. In this regard, these owners of wooden houses are all members of a particular group or class that will be affected in a similar way by a fire.
We know that, on average, 1 per cent of the members of this group will be affected by fire. However, we don’t know exactly who it will be. The important thing for insurance is that members of a group must be homogeneous as far as a particular event is concerned.
In economics however, we don’t deal with homogeneous cases. Each observation is a unique, non-repeatable event caused by a particular individual response. Consequently no probability distribution can be established. Again, probability distribution rests on the assumption that we are dealing with a particular repeatable event. Let us take for instance entrepreneurial activities. If these activities were repeatable with known probability distributions,
then we would not need entrepreneurs. After all, an entrepreneur is an individual who arranges his activities toward finding out consumers’ future requirements. People’s requirements however, are never constant with respect to a particular good. The assumption that the main stream economics makes that probability distribution exists and can be quantified leads to absurd results. For it describes not a world of human beings who exercise their minds in making choices, but machines that are subject to random errors and breakdowns of known types and characteristics.
Because human activity is unique and specific it cannot be homogenized ,as the main stream economics does, by forming historical data into a time series. Each historical observation is a unique, non-repeatable event caused by a particular individual response in a particular context. For instance in year one, entrepreneurial activity yielded 10 per cent return on investment.
In year two the return was 15 per cent. In year three it was 1 per cent, and in year four it was 2 per cent. The average of this distribution is 7 per cent. What we have here is a historical average of returns. By no means, however, does it imply that we can establish a probability distribution on the basis as one can establish for the risk of fire, or for obtaining heads in tossing a coin.
Human activities cannot be analyzed in the same way that one would analyze objects. To make sense of historical data one must scrutinize them not by means of statistical methods but by means of trying to grasp and understand how it emerged.
Dr Frank Shostak is a former professor of economics who now works in the private sector
BrookesNews.Com
Monday 17 April 2006