Confound It!

Abstract. Many investigators have noted “reverse-coding” method factors when exploring response pattern structure with psychological inventory data. The current article probes for the existence of a confound in these investigations, whereby an item’s level of saturation with socially desirable content tends to covary with the item’s substantive scale keying. We first investigate its existence, demonstrating that 15 of 16 measures that have been previously implicated as exhibiting a reverse-scoring method effect can also be reasonably characterized as exhibiting a scoring key/social desirability confound. A second set of analyses targets the extent to which the confounding variable may confuse interpretation of factor analytic results and documents strong social desirability associations. The results suggest that assessment developers perhaps consider the social desirability scale value of indicators when constructing scale aggregates (and possibly scales when investigating inter-construct associations). Future investigations would ideally disentangle the confound via experimental manipulation.