Content uploaded by J. Scott Armstrong
Author content
All content in this area was uploaded by J. Scott Armstrong on Aug 06, 2018
Content may be subject to copyright.
International Journal of Forecasting 15 (1999) 351–352 www.elsevier.com/locate /ijforecast
Introduction
Introduction to paper and commentaries on the Delphi technique
Rowe and Wright’s paper ‘‘The Delphi Technique forecasting. This framework works well in Rowe and
as a Forecasting Tool’’ was initially reviewed by Wright’s (2000) discussion of principles for judg-
four experts in the area of judgmental forecasting. mental forecasting. The term also offers a convenient
Following three rounds of revisions, the paper was way for people to communicate about judgmental
accepted for publication. It was then sent for com- forecasting procedures. Finally, there is an existing
mentary by Professors Ayton, Ferrell, and Stewart. useful body of research on Delphi.
The lead paper should be of interest to researchers In my opinion, the primary problem is that few
because it identifies important aspects of the Delphi validation studies have been done. Here is an exam-
procedure that have not yet been studied. In par- ple of the further research that might be considered:
ticular, there are few validation studies and these One of the basic principles used in Delphi is that
often omit descriptions of the relevant conditions. experts’ judgmental predictions should be made
This makes it difficult to identify which aspects of independently because of evidence that group
Delphi are related to accuracy and the conditions pressures can harm accuracy. Perhaps the traditional
under which Delphi is most useful. design of Delphi is not implementing this principle
Ferrell suggests that it might be time to abandon in the most effective manner. To reduce group
research on Delphi. This is not as radical as it pressure, it seems sensible to provide information
sounds. His point is that we should take a more after round one, but not the forecasts made by other
general view of group processes. The key is to find experts. For example, if Expert A learns that Experts
what group processes provide the best way to make B and C considered factors that he overlooked, an
forecasts in which situations. Ayton and Stewart also adjustment could be made based on that information.
argue that findings from research on group process But if Expert A also receives information about B
and decision making are relevant to this issue. and C’s forecasts, this might produce a bias towards
Another way of saying this is that one should have conforming with the group without providing useful
a broad definition of Delphi. Here is a broad information. In this case, Expert A has no infor-
definition: ‘‘Delphi involves anonymous forecasts mation on whether B and C also took account of all
made on two or more rounds by a group of in- the things that A had considered. So I would like to
dependent heterogeneous experts who receive feed- see an experimental study that compares Delphi
back between rounds.’’ This allows for incorporation where the feedback consists only of information
of key elements of judgmental forecasting and it also relevant to the prediction versus a traditional version
allows for flexibility for such things as the amount, that transmits both information and forecasts. I
type, and timing of feedback. It would also allow one would expect the former to be more accurate. It
to identify salient characteristics of the experts. would also be useful to see whether it is better to
Does the definition serve any purpose? I think it provide information from all experts, or merely from
does. Delphi offers a convenient framework to those whose predictions are at the extremes. This
implement basic principles for effective judgmental research can be aided by computers that can control
0169-2070/99 /$ – see front matter 1999 Elsevier Science B.V. All rights reserved.
PII: S0169-2070(99)00012-6
352 Introduction
the type of feedback provided to panelists and it can Reference
monitor the use of this information by each particip-
ant. Rowe, G., Wright, G., 2000. Expert opinions in forecasting: Role
of the Delphi technique. In: Armstrong, J.S. (Ed.), Principles of
Additional research might also look at im- Forecasting: A Handbook for Researchers and Practitioners.
plementation. It seems clear that Delphi is much Norwell, Mass: Kluwer Academic Publishers, forthcoming.
more accurate than traditional group meetings.
Nevertheless, people seem happier in traditional J. Scott Armstrong
meetings, so this is how they usually make judg- Department of Marketing
mental forecasts. Would this situation change if The Wharton School
,
University of Pennsylvania
Delphi procedures were more readily available in Philadelphia
,
PA
software?