A Comparison of Four Approaches to Modeling Information Insufficiency

Abstract

Information insufficiency, or the disparity between the level of knowledge needed to confidently judge an issue and the perceived level of current knowledge, is a key motivator of risk information seeking and processing. This study compared 4 approaches to modeling information insufficiency within the planned risk information seeking model. These approaches included the raw difference score, regression approach, partial variance score, and direct measure. Statistical modeling used data from large samples in Singapore (n = 2,124) and the United States (n = 2,125). The results of ordinary least squares regression analysis and structural equation modeling pointed to several issues. First, while the raw difference score is conceptually straightforward, it is susceptible to omitted variable bias when constructing explanatory models. The regression method is effective for data sets with low multicollinearity, while high multicollinearity warrants the analysis of partial variance. The direct measure, though simple, is prone to common method bias. Researchers should use the regression approach or partial variance score after assessing the degree of multicollinearity in their data sets.