A meta-analysis of the Fishbein and Ajzen theory of reasoned action

Access restricted to the OSU CommunityMeta-analyses provide a useful, systematic, and logical method to summarize the effect size o f a particular attribute. Determining whether there are differences in the effect sizes among the four major subgroups that report effect sizes- published journal articles, "fugitive" unpublished studies, book articles, and unpublished articles, is necessary to establish whether inclusion of each is essential in meta-analyses. The problem o f this study was to conduct a metaanalysis for the Fishbein and Ajzen theory of reasoned action, using published journal articles, "fugitive" unpublished studies, book articles, and unpublished dissertations, to determine the effect sizes of (a) attitudes towards the behavior and subjective norms on behavioral intentions, and (b) behavioral intentions on actual behavior, in the aggregate, and for each o f the four subgroups - published journal articles, "fugitive" unpublished studies, book articles, and unpublished dissertations. Several sources were used to procure studies from each o f the four subgroups. Previous meta-analyses on the topic were reviewed for published journal articles. The Social Science Citation Index was searched under the subject title "attitudes/behaviors." Two clearing houses were also used to locate studies - ERIC and PSYCHLIT. The work of Cohen and Cohen (1975) was used to establish needed sample sizes to procure power of .99 with an alpha of .05. The meta-analytical methods prescribed by Rosenthal (1984, 1991) were used to conduct the metaanalysis. Weighted average r's and z's were used to establish effect size and significance per subgroup and in the aggregate. After converting r's to a 's, skewness and kurtosis checks indicated the subgroups were not all normal. A nonparametric Kruskal-Wallis Test indicated significant differences among mean average effect sizes o f the subgroups for the A-BI multiple correlation. The Kruskal-Wallis Test also indicated significant differences existed among the mean average effect sizes of subgroups for the BI-B correlation. Mann-Whitney H tests were used as "L.S.D" like multiple comparison follow-ups. For the A-BI linkage the only significant difference among subgroups was the published journal articles had mean average effect sizes that were greater than "fugitive" unpublished studies. For the BI-B correlation, two significant differences emerged: Both published journal articles and book articles had larger average effect sizes than unpublished dissertations. Both linkages were found to have effect sizes that were statistically significant, both in the aggregate, and for each of the four subgroups. The aggregate multiple correlation (A-BI) of 0.61, and the aggregate correlation (BI-B) of 0.47, are not likely large enough, however, to be useful in actually predicting behavior in groups, based on criteria suggested by Borg & Gall (1989)