The authors explore factors leading to successful implementation and use of macroeconomic and microeconomic models by agencies of the federal government. Case studies of two sets of models, the TRIM/MATH (Transfer Income Model/MicroAnalysis of Transfers to Households) models and the DRI (Data Resource, Incorporated) national econometric model, form the basis of the research.
Economic models constitute the preponderance of those computer models currently in regular use. Examples of both microeconomic and macroeconomic models provide good case study analyses. Two-thirds of the book talks in some detail about the development, modification, maintenance, and use of the TRIM/MATH and DRI models.
The TRIM/MATH models are microeconomic models of a variety of income transfer programs and their effects on individuals and households. DRI is a macroeconomic simulation model of the United States economy. Each of these models was developed during the 1970s and continues to be used today by a variety of agencies of the federal government.
The time span under consideration is particularly significant as it allows for discussion of the impacts associated with the shift from a long period of dependence upon Keynesian theory in the sixties and seventies to Reaganomics and monetarist theory in the eighties. Modeling suffered during the transition due to the lack of an adequate economic theory. However, the traditional bureaucracy reasserted itself after only a few years and the earlier, more highly-developed and familiar models again came to dominate.
The research focuses on four sets of variables affecting model implementation and use--environmental preconditions, organizational attributes, features of the technology, and transfer policies. Subjective analysis of the case studies then identifies those factors having primary significance.
Several environmental preconditions are important. The presence of a body of substantive theory in the area of concern (Keynesian economic theory here) as well as related modeling theory, adequate funding sources for model development, appropriate computer technology, and adequate quantity and quality of data are necessary but not sufficient causes for model implementation and use.
The federal agency’s need for the analysis and its possession of resources, in the form of finances and personnel, are key demand features in successful model implementation and use. Since research has shown that models are no better predictors of the economic future than experts, an agency’s need is often determined by political considerations, that is, numbers lend credibility. The demands of the user organization are the most critical element in model implementation and use.
The disconcerting finding for computer professionals is that computer models are being widely used at a cost of millions of dollars a year not because of their inherent ability to produce new or more accurate data but because of their political utility. The models are known to be only moderately accurate in their predictions, but they are significant for their ability to model policy impacts on politically-relevant subsectors of the population. While the authors indicate interest in this finding at several points in the book, they do not appear to be particularly disturbed by it.
This book, like most research studies, is perhaps more interesting because of the questions it raises rather than those it answers. It is clear that, in the governmental arena, political decision making rather than purely rational decision making is the central issue. Where modeling facilitates that political process it is widely used, where it does not it is ignored or suppressed. Political decision making is a process of legitimizing actions. Currently numbers and computers add significantly to the aura of legitimacy. Therefore models that use computers and generate numbers are worth millions of dollars per year.
The greatest long-term utility to the use of models may be the opportunity it provides to experiment on a massive and extended scale with the new modeling and computer technology. By the time someone raises serious and embarrassing questions, the modeling process, and the undergirding economic theory, may have become sufficiently refined to stand on its own merits. However, the bureaucratic entrenchment of these present models, the enterprises that support them, and the agencies that use them may mean that significant improvements will come from new outside sources.
The findings of this study are highly subjective but reasonable on the basis of the information provided. The case studies are interesting and informative in and of themselves and provide sufficient detail to allow readers to draw their own conclusions. The accounts once again illustrate how significant nonrational elements are in human interaction and how substantial their impact is upon the perhaps all too rational world of the computer scientist.