Commentaries on “The Delphi technique as a forecasting tool: issues and analysis” by Rowe and Wright

https://doi.org/10.1016/S0169-2070(99)00013-8Get rights and content

Section snippets

Comments by Peter Ayton

Perhaps it is a myth – for my analogy it hardly matters – but I am told that what we know as aspirin was originally obtained by boiling willow bark. It was deemed effective and regularly prescribed long before anyone had the faintest idea how it worked. Though its efficacy is not much in doubt, how it works is still largely a mystery to medicine. The research reviewed by Rowe and Wright gives one a similar sense of the nature of Delphi. Generally, it seems to be a better procedure for producing

Comments by William R. Ferrell

Rowe and Wright's criticisms of Delphi research and their recommendation for more attention to the underlying judgment processes are well taken. From a practical design standpoint – determining how best to make judgments with groups – I draw two conclusions. The first is that the often quoted opinion of Lichtenstein et al. (1982) that calibration research was "dustbowl empiricism" applies also to research on Delphi; there has been a dearth of theoretical work. The second observation, however,

Comments by Thomas R. Stewart

In their thorough and insightful review of research on the Delphi technique, Rowe and Wright conclude that the research that has compared the Delphi to other methods is of limited usefulness and that "research should focus on an analysis of the process of judgment change within nominal groups." They conclude that their critique of Delphi could also be applied to a larger body of research on "judgment-enhancing techniques".

Rowe and Wright are correct to generalize their Delphi critique to all

Importance of the task

Judgmental forecasts are the joint product of the forecasters and the forecasting task.

Task properties determine the difficulty of the task and place an upper limit on human performance. For example, Stewart, Roebber and Bosart (1997) compared several tasks and found that the level of uncertainty in the task explained most of the variation in the performance of judges. Furthermore, the experience with a task affects the judgment process. People learn to make judgments in different ways for

Subject population

Rowe and Wright find that most Delphi studies involve homogenous groups of undergraduates. Much research that is based on the responses of college students to abstract tasks is appropriate and useful. However, studies of methods for improving judgmental forecasts should include experts working with task content that is familiar to them. This requirement follows from the same principle that requires the use of representative tasks: Forecasts are the joint product of the task and the forecaster.

Group average analysis

Rowe and Wright view the individual as the appropriate unit of analysis for studies of the effectiveness of the Delphi method, while the research that they review generally ignores individuals and focuses on group averages. This approach assumes that judgment processes are universal and are independent of the individual (and the general environment). This view is accepted, often without question, in much psychological research.

The argument for analyses based on individuals (idiographic

Conclusion

It is unfortunate that the research on the Delphi method conducted during the last three decades has not led to an understanding of the conditions (task and forecaster) that favor its use or of the best techniques for implementing it. It is also unfortunate that the same is true of research on judgmental forecasting. The remedy for this situation has been available for at least 50 years. It lies in research conducted with full recognition of the importance of the task and the individual.


First page preview

First page preview
Click to open first page preview

References (0)

Cited by (0)

View full text