The Use of Within- and Between-Subjects Experimental Designs in Behavioral
Accounting Research: A Methodological Note
Phillip D. Harsha and Michael C. Knapp
The growing popularity of ANOVA designs in behavioral accounting research
has focused attention on several methodological issues related to their
use. Among the most debated of these issues are the comparability of key
statistical measures for within- and between-subjects ANOVA designs and
the potential impact that experimental artifacts may have on the results
of within-subject designs. Regarding the first of these issues, this paper
suggests that when comparing the results of the two types of ANOVA designs
researchers must consider both the levels of statistical significance achieved
and the strength of association between independent and dependent variables.
If only levels of statistical significance are considered when making such
comparisons, researchers may incorrectly conclude that the results of the
two different types of studies are substantively different. This paper
also examines the structure of recent within-subject designs and concludes
that demand effects can be induced in such designs if a sufficiently "transparent"
instrument is used. However, this ability to artificially induce demand
effects does not establish that the results of previous, structurally sound
within-subject studies have been influenced by demand effects, nor should
it deter researchers from using this design when appropriate.
Return
to BRIA 1990, Volume 2 Contents