Selected Published Papers - J. Scott Armstrong

J. Scott Armstrong
· Faculty Page
· Selected Papers
· Educational Materials

Major Findings
· Forecasting
· Marketing
· Scientific Methods
· Social Responsibility
· Strategic Planning
· Education
· Applied Statistics

People
· Faculty
· Former Faculty
· PhD Students
· Phd Alumni
· MBA Students
· Undergraduates
· Staff

Home

Major Findings

This page provides a summary of findings in which I was involved (the multiple asterisks denote the more important findings).

Forecasting

****	Developed "Rule-based Forecasting" (in collaboration with F. Collopy), an expert systems approach to improve the selection and combination of extrapolation forecasts (also see Principles of Forecasting)
****	Proposed and tested the casual forces for the selection and weighting of extrapolation methods.
***	Developed and tested simulated interaction as a way to more accurately predict outcomes in conflict situations and negotiations, with K. Green
***	Developed and tested structured analogies (with K. Green) as a way to more accurately predict outcomes in conflict situations and negotiations
***	Combining forecasts reduces error by about 12%
***	Concluded that game theory has no value for forecasting (with K. Green)
***	Found that judgmental bootstrapping substantially improves accuracy relative to judgmental forecasts
**	Proposed and evaluated the Relative Absolute Error (RAE) for comparing the accuracy of extrapolation methods across series.
**	Proposed that trend factors in extrapolation models be damped (based on a literature review).
**	Modified (damped) seasonal factors that use domain knowledge reduce forecast error by 6%
*	Developed the Seer-Sucker Theory: high expertise in forecasting change has little relationship to forecast accuracy.
*	Developed a procedure to conduct the "forecasting audit."
*	Obtained evidence that econometric methods provide more accurate long-range forecasts (see also Principles of Forecasting)
*	Showed simple extrapolation methods to be as accurate as more complex methods in forecasting under a wide variety of conditions (see also Principles of Forecasting)
*	Concluded that company earnings forecasts are more accurate than those by analysts which, in turn, are more accurate than extrapolations (based on meta-analysis).
*	Concluded that the mean square error is inappropriate for comparing forecasting methods (see also Armstrong and Fildes).
*	Prediction intervals for series where trends conflict with expectations are asymmetric in the logs

****	Market-share and competitor-oriented objectives are detrimental to profit (the myth of market share).
***	Extrapolation across waves can be used to correct for non-response bias in mail surveys.
***	Experts' predictions of consumer behavior lack validity.
***	Use of the BCG portfolio matrix leads to less profitable decisions
***	Found strong relationship between monetary incentives and mail survey response.
**	Identified the conditions under which multiplicative decomposition is useful for estimation
**	Brief descriptions are sufficient for intentions-to-purchase studies in some conditions
**	Found that texts on marketing principles contain few principles
*	Practitioners do not use well-established principles on how to select employees
*	Showed that business-reply postage is not cost-effective for mail surveys
*	Marketing journals publish few replications and these are published after much delay.
*	Extensions of marketing studies typically fail to reproduce the original findings.
*	Conducted experiments showing that escalation bias does not generalize to marketing decisions.

Scientific Methods and Peer Review

***	Concluded that method of multiple hypotheses is underutilized in management science (also see on advocacy)
***	Obtained evidence of a bias against the publication of papers with controversial findings
**	Found relationship between less intelligible writing and academic prestige
*	Identified conflicts between scientific advancement and the advancement of scientists
*	Proposed and used the "survey of researchers" to assess reliability and validity of coding in meta-analyses
*	Proposed and implemented a "Results-blind reviewing" procedure for reviewing journal articles with controversial findings (this reviews design without knowledge of results)

Social Responsibility in Management

****	The stakeholder role, in combination with social accounting, reduces the likelihood of socially irresponsible decisions in marketing

Strategic Planning

****	Formal planning improves organizational effectiveness (used in meta-analysis). (see also planning 1 and planning 2)

***	Provided support for the hypothesis that business school prestige rests upon research, not teaching
***	Teaching ratings of faculty are detrimental to learning skills
**	Developed and assessed the use of time contracts as a way to increase learner responsibility
*	Developed and assessed a method of "learning by objectives"

Applied Statistics

*	Illustrated dangers with the failure to assess reliability in factor analysis.
*	Illustrated dangers with the failure to assess reliability in regression analysis.
	Return to top of this page Return to Professor Armstrong's top page

Comments or questions
Copyright © 1994-2002 Trustees of the University of Pennsylvania
Certifying Authority: The Wharton School Marketing Department