ORGANIZATIONAL RESEARCH METHODS Ganzach et al. / LINEAR VERSUS LOGISTIC MODELS
Interaction in Linear Versus Logistic Models:
A Substantive Illustration Using the Relationship
Between Motivation, Ability, and Performance
YOAV GANZACH
ISHAK SAPORTA
Tel Aviv University
YAACOV WEBER
The Hebrew University of Jerusalem
A binary performance measure (high school graduation) is examined as a function
of motivation (educational goal), ability (scores in an intelligence test), and their
interaction. The interaction was positive when a logistic model was used and neg-
ative when a linear probability model was used. The reason for the difference in
the results of the two models is examined, and the conditions under which this dif-
ference occurs are discussed.
Many of the important dependent variables in organizational research are binary. For
example, between 1995 and 1997 about 5% of the articles in the Journal of Applied
Psychology and about 8% of the articles in the Academy of Management Journal used
binary dependent variables. Nevertheless, methodological issues associated with the
modeling of such variables, and, in particular, the modeling of interaction effects
among their determinants, have received little attention in the literature. The purpose
of this article is to examine issues associated with the modeling of interaction when the
dependent variable is binary within the context of an important, yet little researched,
substantive issue: The relationship between motivation and ability in the determina-
tion of performance. A binary performance measure (high school graduation) is exam-
ined. Scores in a cognitive ability test are used as a measure of ability, and educational
goal is used as a measure of motivation.
The article is organized as follows. We first review the relevant literature regarding
(a) the modeling of binary dependent variables using linear and logistic models, and
(b) the relationship between motivation, ability, and performance. We then present
data showing that modeling performance as a function of motivation and ability by a
linear probability model and modeling it by a logistic model lead to diametrically
opposite conclusions about the interaction between motivation and ability. In the last
Organizational Research Methods, Vol. 3 No. 3, July 2000 237-253
© 2000 Sage Publications, Inc.
237