Evaluating the equivalence of an employee attitude survey across languages, cultures, and administration formats

Stephen G. Sireci
James K. Harter
Yongwei Yang
Dennison S. Bhola
International Journal of Testing, 3(2003), pp. 129-150

Abstract

Modern multinational organizations have found that testing or surveying their employees is difficult when the employees come from a variety of language backgrounds. In such situations, surveys and tests are often adapted for use across multiple languages. However, different language versions of an instrument are not necessarily equivalent, which may lead to misleading interpretations. In this study, we used weighted multidimensional scaling (MDS), analysis of covariance (ANCOVA), and ordinal logistic regression (LR) to evaluate the structural equivalence and differential item functioning (DIF) of an employee attitude survey from a large international corporation. Specifically, we evaluated the functioning of the survey items across 3 different languages, 8 different cultures, and 2 mediums of administration (paper-based and Web-based). MDS was used to evaluate structural equivalence, and ANCOVA and LR were used to evaluate DIF across selected employee groups and the 2 administration formats. The results indicated that the structure of the survey data was consistent and that the item functioned similarly across all groups. The results also illustrate the utility of MDS, ANCOVA, and LR for evaluating translated instruments. The implication of the results for future research in this area are discussed.

Research Areas