12.2 Real concrete dgps
What econometricians call ‘data generation process’ is nothing less than a system or phenomenon taken as a target of scientific research, out of which empirical data in a numerical form are expected to flow, or be painstakingly obtained.
From an ontological point of view, it is a real and concrete economic system or phenomenon that behaves in a random way. Spanos’ working definition of DGP is “the mechanism underlying the observable phenomena of interest” (Spanos 1986: 20). The term is alsoused to designate the phenomenon of interest which a theory purports to explain. The concept is used in order to emphasise the intended scope of the theory as well as the source of the observable data. Defined this way, the concept of the actual DGP might be a real observable phenomenon or an experimental-like situation depending on the intended scope of the theory.
For example, in the case of the demand schedule discussed in Chapter 1, if the intended scope of the theory is to determine a relationship between a hypothetical range of prices and the corresponding intentions to buy by a group of economic agents, at some particular point in time, the actual DGP might be used to designate the experimental-like situation where such data could be generated. On the other hand, if the intended scope of the theory is to explain observed quantity or/and price changes over time, then the actual DGP should refer to the actual market process giving rise to the observed data. It should be noted that the intended scope of the theory is used to determine the choice of the observable data to be used.
(Spanos 1986: 661-2)
In the language of Chapter 4, the dgp is a real concrete subjekt that remains outside the head, independent of it and of any theory; it is a “chance mechanism” giving rise to observable data. Pretheoretical thought can grasp its existence and some of its properties and effects, providing the starting point for a description of the same in a language that contains general but not idealized terms.
This description is sufficient in order to refer to the DGP and start probing into it. Hence, the rather strange (idealistic) doctrine of the “theory dependence” of observation is rejected as unnecessary and confusing. This is not to deny that theories can lead to discover new aspects of the process or entity; only that its existence does not depend of any conceptualization. It is just not true that we do not have access to the world independently of our theoretical systematizations; if that were the case we would never be able to notice that the axioms of an idealized theory are actually false or inaccurate.For econometrics the dgp is a random ‘mechanism’ but randomness must be taken as an objetive property of real things and not as a subjective measure of certainty. As Patrick Suppes (1984: 10) puts it: “the fundamental laws of natural phenomena are essentially probabilistic rather than deterministic in character”. This same proposition can be extended to include social, and not only natural phenomena. In adopting the frequency interpretation of probability, Spanos actually accepts this proposition, that I will take also here for granted.
Spanos’ uses ‘theory’ in order to refer to a certain interpretation of probability theory that purports to explain a certain phenomenon of interest. In the jargon of SVT, a theory is far more than that. It is the general description of a class of set- theoretical structures, plus a family of intended applications including qualitative descriptions of their structure. In Spanos’ use, it is the application of the concept of a probability space to a particular target system, together with several estimation methods intended to determine T-theoretic parameters. Hence, it seems only natural to propose a definition of econometrics as probability theory together with a (rather large) class of random target systems of an economic nature, with corresponding methods to estimate the parameters that characterize them.
The former view implies that the basic theory-element of econometrics is given by Kolmogorov’s axioms for probability, but there is far more to econometrics than such axioms. In the final section I will present a detailed, albeit compact characterization of econometric theory.
12.3