Search

This article presents Paul Kay's reaction to several comments on his paper which appeared in a previous issue of The American Sociologist, which addressed the nonacademic employment among doctor of philosophy degree holders in the U.S. On the whole, according to Kay, there seems to be agreement among the commenters (a) that social science generally is facing an oversupply of doctor of philosophy degree holders and (b) that there is substantial adherence in some quarters to the claim that the glut can be alleviated by expanding nonacademic employment. Similarly, although some commenters express doubt regarding the degree to which what Kay describes for anthropology applies to sociology, most state or assume that the current employment situations of the two disciplines are essentially the same. That is, the consensus of the comments is that his argument is either right for both disciplines or wrong for both disciplines, but probably not right for one and wrong for the other. Several commenters are worried by the lack of statistical report. Kay admits that his paper presented no statistics. According to him, reliable data in this area are hard to come by especially on the demand side.

This article focuses on the utility of population forecasts. Given the many uses of population forecasts and the need to provide a warning about the size of their errors, it is not surprising that population forecasts have been evaluated frequently. These evaluations have created an atmosphere in which the "stand-alone" accuracy of a given forecast, by itself or relative to competing forecasts, is the dominant criterion and consequently the normative expectation for judging adequacy. Researchers made ex post facto comparisons of 1990 population forecasts and 1990 census counts using a "purposive and convenience" sample of counties from seven states and census tracts from three metropolitan areas. All of the county forecasts were based on the cohort component method and used a top-down approach, whereby these counties are "controlled" to an independent state forecast. The research agenda for population forecasting should include projects aimed at improving the ability to incorporate constancy and decline into the forecasting process. The evidence presented in this paper strongly suggests that utility, in the form of "value-added" accuracy, would improve markedly if researchers could identify, in a timely and cost-efficient manner, areas that are likely to remain stable or decline and could understand what differentiates them from areas that are likely to grow. One way of doing this may be to examine the history of areas and develop generalizations in the form of typologies, protocols or models.

In this paper I argue that the selection of appropriate statistical procedures is at best haphazard without a careful consideration of the consequences of one's research, and that significant distortions often will result when applied concerns are ignored. I advance the hypothesis that typical conceptions of basic research in sociology preclude the proper use of statistics. From this it follows that mainstream, quantitative sociology can begin to save itself only by taking seriously the question of applications. [ABSTRACT FROM AUTHOR]