Several Historical Considerations

Abstract

Early in Chap. 4, in order to introduce the set of values later presented as “Core” or “Basic” statistics, several examples of early regression displays were shown, starting with one from 1962 for the program RGR. The purpose was, in part, to demonstrate the historical antecedents for considering such a set of statistics as a group, for as that discussion illustrates, starting from the earliest days of econometric software, what later occurred historically was for evaluative statistics to be added progressively to the standard regression display from time to time. Actually, to say “from time to time” may convey too much of an impression of steady change. Although there was some variation from program to program during the period from 1962 to 1985, for most of this almost quarter of a century time period there was both a high degree of uniformity in the statistics displayed and relatively little change, notwithstanding that each econometric software developer was in principle able to chose independently which statistics to display and in what format. In contrast, from 1985 to the present day, not quite 25 years, has been a period of more change and less uniformity, especially the past 10 years, reflecting both the propagation of contesting econometric ideas and the “market dynamics” of the “econometric software industry.” Such differences possibly mirror the change in econometric practice and this thought stands behind the content of this chapter.

However, these most obvious aspects do not tell the whole story. The provisions made for data management and access, including both software facilities and usable data sets, also affect both the quantity and quality of empirical research. It is possible to argue that these facilities are today the best they have ever been, reflecting the development of the Internet and the other capabilities of the modern computer. But, as demonstrated earlier, they are far from perfect as an enabling factor. In addition, there are more general questions that might be asked concerning the way in which research results are customarily presented. For instance, considering the modern economics literature, it is troubling that it is both possible and seemingly generally acceptable to publish books subtitled Theory and Empirical Evidence (Gong & Semmler, 2006) that neither precisely identify the empirical data used nor the details of the computational methodologies applied, often using the word “algorithm” as a synonym for an ideal mathematical formula, rather than the implemented code that computationally expresses that formula. It is not clear whether this type of approach reflects an increasing degree of unfamiliarity with economic measurement concepts on the part of economists generally or instead an unwillingness to take the trouble to work with actual data and then describe that work in a way that would facilitate its replication by others. But whatever the reason, it has become all too common for analysts to focus nominally upon techniques, rather than to consider the correspondence between theory and the real world. The consequence is an increasing lack of communication: once data have been gathered, fed into a computer program that is only generically identified, and the findings then described in too cursory a fashion, the report on this research effectively becomes imaginary to the reader – irrespective of its actual underlying content.