Danaher &
Crandall,
in press
This paper offers a
re-analysis of data collected and presented by
Stricker & Ward (2004),
in which the claim was made that there were no "effects
of inquiring about ethnicity and gender on performance
on [standardized] tests" (p. 685). The current authors
critique the use of conservative decision criteria used
in the original analyses. Specifically, Stricker and
Ward (2004) only considered results to be significant if
p<.05 in the overall ANOVA, p<.05 for
planned comparisons using a very conservative post hoc
Bonferroni correction, and effect sizes were η>.10
and
d>.20. While these criteria do minimize the chances
of making a Type I error (erroneously claiming a
significant effect), they also elevate the likelihood of
making a Type II error (failing to detect a real, even
if small, difference). Given the importance of
standardized test performance in determining educational
opportunities, career paths, and life choices, Danaher
and Crandall argue that use of standard statistical
decision criteria is misplaced in this context.
Accordingly, these authors re-examined the data
presented by Stricker and Ward (2004), but used criteria
of p<.05 from the overall ANOVA and η≥.05 for
determining significance. These small changes produced
several conclusions differing with those offered by
Stricker and Ward (2004). Timing of soliciting identity
information interacted with gender in affecting AP
Calculus AB Grades and Formula Scores. Women’s
performance on these measures improved and men's
performance declined when when they were not asked about
gender before taking the test. Put in different terms,
soliciting identity information at the end rather than
at the beginning of the test-taking session shrunk sex
differences in performance by 33%.
Analysis of
performance on the Computerized Placement Test
also yielded an identity timing x gender interaction,
with women's performance improving substantially, and
men's scores declining slightly, when they did not
report their identities before taking the test. This
re-analysis suggests that soliciting social identity
information prior to test taking does produce small
differences in performance consistent with previous
findings in the stereotype threat literature that, when
generalized to the population of test-takers, can
produce profound differences in outcomes for members of
different groups.Back to top | Previous
Page
|
