The structure of the Levenson locus of control scale in young adults: Comparing item and parcel indicator models

Research on item parcels suggests that previous reliance on item-level indicators may have resulted in the poor model fit often obtained for the original three-factor structure of the Levenson locus of control scale. The present research tested two- and three-factor models with items and two forms of parcels serving as indicators. In addition, the models were tested in the prediction of attitudes toward computers and depressive experiences. Results found that a two-factor structure with item-to-construct balanced parcels provided the best model fit. However, regression analyses indicated that a three-factor conceptualization resulted in the most informative prediction of computer attitudes and depression.