Econometrics

Peter O’Brien

Jan 10th 2000

How does Econometrics contribute, if at all, to the scientific status of Economics?

Table of Contents

In examining this question, I am going to look at three principle areas. Firstly, under what criteria do we judge Economics to be scientific? Secondly, very simply, what is Econometrics? Finally we will examine the core of the argument by investigating the effect of Econometrics on the status of Economics as a science. (assuming that we have accepted Economics as a science, of course!!)

Definitions abound. Encyclopaedia Britannica defines science as "any mode of investigation by which impartial and systematic knowledge is acquired" (in Kearns: 1995). For a statement to be scientific it must be amenable to verification; i.e. that it is possible for others to repeat and experiment and arrive at exactly the same result given a particular test, ceteris paribus. This brings us to one of our key issues when it comes to Economics and the Social Sciences more generally, that of ceteris paribus. In the 'real' social world, no two social, economic or political situations are ever exactly identical.

Two quotes taken form Kearns (1995) are of particular use when applying the issue of scientific methods to Economics. The first comes from Mises:

This gives us a better intuitive understanding of what we expect scientific endeavours to entail, namely a rigourous examination of some given experiment. At the mention of experiments, the gap between the 'Social' and 'Natural' sciences grows even wider. In the laboratories of the 'Natural' scientists, all experiments are precisely controlled and it is possible to repeat such experiments, assuming all the conditions, and to arrive at identical results and that this exact same result would be achieved time after time. In other words, it is possible to verify such controllable experiments. However, in certain disciplines, namely the Social Sciences and other 'renegades' like meteorology (see Gujarati, p.3 for example) it is not possible to repeat and/or directly control any experiments in any two situations. In conclusion, just because the social world is unscientific in its nature, it does not necessarily imply that those disciplines that attempt to observe, measure and quantify its behaviour and actions are themselves unscientific in their approaches. So the main point to take from this section to section 4 is that while we must assume that Economics is a (social) science, what we are really being asked is whether or not Econometrics makes Economics more or less scientific.

According to Brown (in Falk) Econometrics deals with "[the] empirical determination of economic laws", whereas Koutsoyiannis (1977) believes that Econometrics deals with the measurement of economic relationships. According to her it is a combination of Economic Theory, Mathematical Economics and Statistics. As a relatively new discipline, there exist many variations on this theme. For example Kennedy (1992) makes no distinction between Economic Theory and Mathematical Economics. Such (minor?) inconsistencies aside, I am satisfied with the description of the four disciplines as outlined by Koutsoyiannis (1977) and further supported by Gujarati (1995). According to Koutsoyiannis (1977: Taken from Section 1.2) there are three main 'goals' of Econometrics:

1. Analysis: The testing of given Economic Theory.
2. Policy-making: Supplying numerical estimates of the coefficients of economic relationships which may then be used for decision making.
3. Forecasting: Perhaps the most controversial. Using numerical estimates of the coefficients in order to forecast the future values of economic magnitudes.

Some important aspects of Econometrics are specification errors, sampling technique and data compositions. There are innumerable problems involved in dealing with the social world, such as Multi-collinearity, but it was to tackle such issues that Econometrics was applied to the Social World in the first place. There may be also be collection problems; an insufficient sample, ones data may be biased, or just untrue, or perhaps ones initial theory has been mis-specified.

Things to briefly note before they are examined in more detail include reflexivity and the lack of historical consistency in econometrical data.

It is crucial to understand that we re not trying to show that Economics can be as consistent as any of the natural sciences. To do so would be to ignore Economics as a Social Science. So an alternative is to illustrate that without Econometrics that Economics would be less scientific. In looking at the Social Sciences, we must think of the key issue of reflexivity. That is, the researcher examining the social world is an active part of that which s/he is researching, and her/his actions have an influence over the result.

The other major issue that I pointed to in section 3 was that of historical constancy. In the 'social world', data is prone to changes in measurement over time, whereas this is not normally the case for the 'natural sciences'. A good example is given in Dixon (1998), the CPI (Consumer Price Index). The CPI does not measure the same things today as it did twenty years ago.

Zellner (1984) has a very good critique extolling the merits of Econometrics in bolstering the credibility of Economics as a science. He believed that without a good understanding of the philosophies and objectives of Econometrics that that we really do not know what we are doing in Economic research, and that a "strong interaction" between theory and economic science is most beneficial for the progress of economic science. He states that:

This points to clear indications that Economics would be considered less scientific without Econometrics as a supporting tool. Dixon (1998) points out that Econometrics is an important tool in making Economics more amenable to methodological appraisal.

Zellner also gives us very good counter examples of what can go wrong if Econometrics is taken out of the 'marriage' referred to in the Introduction. He gives examples of wonderfully elaborate mathematical economic theories that yield no verifiable predictions. He points out that while these may make contributions to the world of mathematics in their own rights, they may not necessarily have any useful contributions to make to the world of Economic Science. Likewise, Econometric models with a high R² (e.g. R² =.99) but that don't predict the data very well are not exactly making a very useful contribution to Economic Science. It is such illustrations which indicate to us the extent to which all of these fields have become interdependent and interreliant.

In summing up his arguments as to why Econometrics does not contribute to the scientific status of Economics, Falk (1995) says: "natural scientists ... are generally able to acquire more consistent and accurate results than those acquired from the research of Econometricians". Now while this is undoubtedly true, it almost serves to prove my own point as has been discussed. At no point do we in the Social Sciences ever try to present exact formulas with the same degree of precision and certainty of those proposed by Natural Sciences. However, I think it has become clear through this discussion that Econometrics does have an important role to play in enhancing Economics' scientific status.

6. Bibliography