'...the difference between facts which are what they are independent of human desire and endeavour and facts which are what they are because of human interest and purpose and which alter with alteration in the latter, cannot be got rid of by any methodology. In the degree in which we ignore this difference, social science becomes pseudo-science.'
- John Dewey: 'The Public and Its Problems'
Science is to be defined as a human cognitive enterprise, the aim of which is to produce falsifiable and predictively successful theories. A scientific theory is not so much a way of discovering 'truth' but rather a means of organising our empirical observations and thoughts in a useful way. This, according to Robert Aumann, is the sole criteria by which a theory, and therefore a scientific discipline, is to be judged. It follows that a theory which facilitates the organising of our thoughts in a useful and directed way must necessarily be a predictively successful theory.
This relates to the definition of economics, or at least positive economics, which is a system of generalised theories purporting to explain the behaviour of economic agents, often in a deterministic way. However, any inspection of actual economic behaviour reveals it to be a variegated and complex phenomenon, involving non-deterministic relationships. This introduces us to econometrics which attempts to measure in an empirical manner the nature of the association between stochastic economic phenomena.
On first glance, it would appear that econometrics has an extremely important role in making economics 'scientific'. If the true field of economic analysis is non-deterministic relationships, econometrics would appear to be an ideal way to examine the congruence between the postulates of economic theory and the requirements of a scientific discipline. In this essay I will argue that the impact of econometrics on economic theory has been minimal and far from being the only way of examining the scientific credentials of economic theory, there are in fact several other ways of developing our understanding of the structures of a real economy.
Aristotle's (Aristotle 1972) precept will be observed throughout the essay: 'Our discussion will be adequate if it admits of enough clarity as the subject matter allows'. But why is it necessary to make this qualification for the subjects of economics and econometrics?
We use Aristotle to qualify our discussion because of the reflexivity, lack of historical constancy, and conditionality of economics as a scientific discipline. These are the facts which provide the main difference between testing in economics and the natural sciences. The data that econometricians work with inhibit the precise estimation of empirical relationships. In the context of the restrictions discussed below, we must examine whether econometrics does allow us a discussion of the data which is as adequate and instructive as the data itself allows.
The main difference between testing in economics versus the natural sciences is that the phenomena/data under consideration are liable to historical change. For example, the CPI measures something different from the CPI of 40 years ago. An atom of carbon has the same structures it did 40 years ago. Not only do structural relationships lack the historical constancy that the materials of the natural sciences possess but shocks to the parameters of the economic system are definitionally unpredictable in their timing and consequences. A quote by Robbins (Robbins, 1981) vividly illustrates the difficulty econometrics has in estimating structural parameters. 'The influence of the Reformation made no impact on the forces of gravity. But it certainly must have changed the demand for fish on Fridays'. As Johnston (Johnston, 1991) has reflected 'It is surely too much to expect the econometrician to develop a super demand curve for fish, which contains within itself an explanation for the Reformation'.
The problem of the reflexivity of economic behaviour has been discussed by Rosenberg (1992):
'When all agents are informed about these plans and predictions and know the relevant theory, the reactions of at least some agents to these plans and predictions will result in some of the predictions being falsified and some of the plans going awry'.
Another unique aspect of economic data was pointed out by Kamarck (Kamarck, 1983a)'No electron profits from deceiving a physicist but the source of economic statistics may have a direct interest in reporting inaccurately or falsifying economic data'-to avoid tax liability for example.
These facts need to be borne in mind when one looks at the success of econometrics in providing a firm empirical base for economic theory. I will argue that econometrics provides an unsatisfactory analysis of the problems even given the restrictions discussed above, because it is plagued by problems of weak data, ideology affecting the outcome of 'empirical' tests, and misdirected effort. I will suggest that forms of economic analysis which do not necessarily rely on the crutch of econometrics are more useful in examining the structures of a real economy generally and the scientific credibility of economic theory in particular.
For example, econometrics is often characterised by data mining and a prioristic conclusions; researchers massage results so as to produce an outcome that accords with personal opinion. This means that the theory has not been tested and therefore has not been subjected to the falsification principle. Econometricians rarely try to find out if there is another fit to the data, 'acting as if the data admitted only a unique inference'. This was described by Solow (Blaug, 1992), in discussing the significance of econometrics for economics as 'the biggest sin of all'. In Kenen's (Blaug, 1992) words 'It is not enough to show that our favourite theory does as well as-or better than-some other theory when it comes to accounting retrospectively for the available evidence'. It is difficult to disagree with Mayer (Mayer, 1993), who reports that the practice of running 30 regressions and only publishing the one that confirms a hypothesis is widespread, when he concludes that 'this shouldn't be done in hard testing'.
Econometrics' credibility is damaged by prizing statistical pyrotechnics -'Physicists do not compete to find more and more elaborate ways of observing falling apples'- while ignoring the problem that the data is weak. In the natural sciences, as opposed to almost all economic data collection, great care is taken by highly skilled people to produce accurate and representative data. Kamarck (Kamarck, 1993b) in discussing econometrics has noted how: 'More prestige is acquired for applying the latest techniques to good, bad or indifferent data than arriving at valid, verifiable and useful results'.
The irrelevance of econometrics for economic theory was taken up by Leamer (Hendry et al, 1990) who argued 'We don't take empirical work seriously in economics. It is not the source by which economists accumulate their opinions, by and large'. Klamer and Colander (in Mayer, 1993) report that : '[Econometricians] will confess, usually at unguarded moments, that their highly sophisticated research produces ultimately meaningless results' since, their results and conclusions don't describe the operation of any economy that has ever existed or will ever exist. This cannot be said to contribute to the scientific status of economic theory.
Summers (Poirier, 1994) has taken up this argument by convincingly asserting that non-econometric models have had a far greater impact on economic theory than econometric studies.
'Surely A Monetary History of the United States (1963) had a greater impact in highlighting the role of money than any particular econometric study or combination of studies... data were presented in a straightforward way to buttress verbal theoretical arguments and emphasis was placed on natural experiments in assessing directions of causality'.
Walters has claimed that the most significant development in post war empirical economics was the development of the permanent income hypothesis by Friedman, which did not rely on econometrics. If it is pointed out that these works are also affected by the ideological alignment of the author, at least this is implicitly acknowledged. To recall the quote by Dewey at the start of the essay, it is only when we ignore ideological positions, as opposed to acknowledging them, that the social sciences become pseudo-science. Patinkin (in Latsis, 1976) was distressed by '...the high correlation between the policy views of a researcher ...and his empirical findings'. Surely in this case, econometrics is becoming the instrument of ideology and not science, thus diminishing the credibility of economic theory.
Leontief (Leontief, 1984) has been especially critical of econometrics: 'In no other field of empirical enquiry has so massive and sophisticated a statistical machinery been used with such indifferent results', but this may be ultimately too harsh an assessment since if we can say anything in defence of econometrics is that it has at least raised the wider recognition of the importance of the principle of testability and thus making economics theory more amenable to methodological appraisal. Furthermore, despite all that has been said against econometrics in this short essay, it could best be considered a type of 'weak testing', which shows, if nothing else, some sort of relationship exists between the variables under discussion.
Aristotle (1972) The Nicomachean Ethics translated by Sir David Ross, Oxford University Press : London ,p 163.
Aumann, Robert (1985) 'What is game theory trying to accomplish' in Kenneth Arrow and Seppo Honkapojha (Eds.) Frontiers of Economics, Basil Blackwell: Oxford, p31.
Gilbert, Christopher (1986) 'Professor Hendry's Econometric Methodology' in Oxford Bulletin of Economics and Statistics, vol. 48, no 3, p284.
Hausman, Daniel (1992) The Inexact and Separate Science of Economics, Cambridge University Press : Cambridge, p 281.
Hendry, David (1990) 'The ET Dialogue: Conversations on Econometric Methodology', in Econometric Theory, vol. 6, p181.
Johnston, J (1991) 'Econometrics: Retrospect and Prospect', in Economic Journal, 101, January, pp 51-56.
Kamarck, Andrew (1983a) Economics and the Real World, Blackwell: Oxford, p 15.
Kamarck, Andrew (1983b), op.cit., p9.
Kenen, Peter quoted by Mark Blaug, op.cit, p 241.
Latsis, Spiro (ed.) 1976 Method and Appraisal in Economics, Cambridge University Press: Cambridge, p202.
Leamer, Edward (1985) 'Sensitivity Analysis Would Help', in American Economic Review, vol. 75, no 3. p 324.
Leontief, Wassily (1984) Essays in Economics: Theory and Theorising, Vol. 1, Basil Blackwell :Oxford, 1984: p 27.
Mayer, Thomas (1993a) Truth Versus Precision in Economics, Elgar: Aldershot, p 141.
Mayer, Thomas (1993b), op.cit, p 2.
Ormerod, Paul (1994) The Death of Economics, Faber and Faber: London, p 42.
Poirier, Dale (1994 ) The Methodology of Econometrics II, Elgar: Aldershot, p 530.
Robbins, Lord (1981) 'Economics and Political Economy', in American Economic Review, 71, pp1-10.
Rosenberg, Alexander (1992) Economics: Mathematical Politics or Science of Diminishing Returns, University of Chicago Press: Chicago-London, p 53.
Solow, R quoted by Mark Blaug (1992) The Methodology of Economics-or How Economists Explain, Cambridge University Press: Cambridge, p 242.
Walters, Alan (1986) 'The rise and fall of econometrics' in Martin J. Anderson (ed.) The Unfinished Agenda, Institute of Economic Affairs: London, p 119.