- K.A. Brownlee
Econometrics can be defined as the study in which the tools of economic theory, statistical inference and mathematics are systematically applied, using observed data, to the analysis of economic laws. It is therefore concerned with the 'empirical determination of economic laws', Brown (1991). Economic theories are written in mathematical form and are then analyzed using statistical methods. If the observed data are found to be incompatible with the predictions of the theory, it is rejected. Theories are accepted if the data are found to fit the theory.
Although this process may appear to be essentially simple and straightforward, there are many problems intrinsically related to every econometric study. The biggest problem econometricians must overcome involves the 'data generating process' (DGP). The data generating process is 'the underlying behavioral function that actually produces the data that are being analyzed'. In economics, our data is provided to us by nature. Often, there is insufficient data available, which means the theory cannot be tested in the first place. Other times, the available data is inaccurate which leads to flawed results, which have little or no value. Even assuming that accurate data is readily available, we still cannot be certain that we will be able to perform an analysis. Many economic theories may not be testable with econometrics. In the real world, many variables are inter-related, a consequence of this is the problem of multicollinearity. Multicollinearity is almost always present in multiple regression models, especially those using time-series data. This can also lead to flawed results as it is difficult, sometimes impossible to solve the problem by adjusting the data. Unlike multicollinearity, specification errors can lead to biased parameters, potentially a very serious problem. Other problems which econometricians encounter include those of identification and aggregation.
Econometrics therefore provides us with a quantitative basis to aid the formulation and simulation of economic policies and the provision of economic forecasts. Its limitations however must not be ignored.
The question has to be raised, during any discussion about the scientific status of economics, as to whether or not it is possible to have a scientific study in a field of human behavior. Before the development of econometrics, the natural scientists rejected claims by economists that economics was a scientific study, on the grounds that the ability to test empirically is a precondition for scientific status. What is at issue here is the whole inductive method of economics, of finding in Marshall's words the 'great uniformity' when facts are collected and arranged. Of course all modern economic analysis proceeds not only from the obvious facts of human behavior, but also from the mass of statistics which are available to us today. So, in theory, economics now has the necessary credentials to be truly 'scientific'. It is based on logic, theories can be empirically tested, and accurate predictions about future economic developments can be made. In practice, econometrics is not perfect. Economists are divided as to the usefulness of econometrics. Two schools of thought have emerged.
One contends that economic theory can never achieve true scientific status in the 'natural' or 'physical' sense of the word. Their argument is that all economic models will have omitted variables, because 'everything depends on everything else' in economics. The actual variables used will depend on what was believed beforehand, the result being that all parameter estimates are potentially biased. This in turn leads to biased predictions. In this light, economics can never overcome its inherent subjectivity and can never produce consistently accurate results. Scientific status, akin to the natural sciences can never be achieved.
The other argument claims that as econometrics is merely a process of 'number-crunching', it is an objective method for carrying out empirical testing. As long as there is no manipulation of the data to suit a particular theory or point of view, then consistently accurate results are obtainable and economics deserves full scientific status.
The ongoing development of econometrics has not solved these inherent problems. In fact, the converse has occurred. Econometricians are faced with a multitude of difficulties with every analysis. One of the consequences of these difficulties is that econometricians are wont to try different functional forms, lag structures, and so on until the desired results are obtained or at least results that are interesting enough to warrant publication. This practice, termed 'data mining' is common, although considered unethical. Data mining certainly does not contribute to the validity of econometrics as a science. Every result obtained is imperfect and widely open to subjective interpretation. It is logical that accurate predictions are unobtainable without accurate results. This analysis may seem to present a gloomy picture of econometrics but that the subject is exceedingly complicated does not entail that it is hopeless. Even with all its weaknesses econometrics is still a very useful component of the field of economics.
In my opinion, the fundamental use of econometrics is as an aid in the policy process. However it is essential to remember that econometric analysis is prone to error. Any predictions deduced from the results of an analysis must be seen only as guidelines and not definitive reflections of the future. In economics we are only certain of what has gone in the past. Econometrics is not a crystal ball that allows us to see every fluctuation in the economy before it happens. Policy makers have tended to forget this and in so doing contributed to economic crises such as the massive inflation of the early 80's or the stock market crash and recession of the late 80's. When used logically and correctly econometrics is an invaluable tool available to policy makers in formatting and evaluating economic policy.
Thiel, H., (1971) :Principles of Econometrics.