Test-taking behaviors have been hypothesized as culturally learned, providing an inherent disadvantage to those from ethnic backgrounds different from the test source (Geisinger, 1992). Lack of technical equivalence results in method bias. Variations in data collection or test administration procedures can prove to be critical in cross-cultural comparisons of the performance of translated measures. Adapted test administration procedures can produce variability in performance on translated tests that are technical, rather than substantive, but nonetheless, have a significant impact on results of cross-cultural comparisons (Gilmer et al., 1995). If variations in test administration occur on a translated test, these variations typically impact most items on the test, rather than being selective in influence. Discovering method bias can be accomplished by repeated administrations of a translated test. Variations in performance across administrations for individuals who initially hold similar scores can suggest method bias. Additionally, nonstandard administration of a translated test can be deliberately conducted with a variety of instructions, item variations, response alternatives, and prompted motivations to elicit the impact of method bias. Method bias, as well as scale eqivalence, can also be studied utilizing different methods to collect information on the construct of interest, referred to as triangulation. Triangulation is a procedure that emphasizes the use of multiple measures to identify a particular target construct (Van de Vijver & Leung, 1997).
It may also be advantageous to include measures of social desirability to assist evaluation of method bias (Van de Vijver & Hambleton, 1996). It is therefore important that translated versions of tests be administered in the same standardized format, as that described for the original instrument, in order to provide for reliable comparisons across cultures.
At times, the test protocol itself will pose a problem, due to its inherent structure. Aday, Chiu, and Andersen (1980) found that Hispanic respondents were more likely to respond to "yes" answer options than non-Hispanics on a survey. Social desirability has also been shown to affect response patterns of Spanish-speaking consumers (Marin, Gamba, & Marin, 1992). Social desirability test response patterns have been reported for low socioeconomic status, older aged individuals, and individuals of Mexican origin. This response pattern dissipated as acculturation level increased (Guffey, 1997). Additionally, Hui and Triandis (1989) found that Hispanics made more extreme responses than non-Hispanics on a 5-point scale questionnaire, with these differences fading on a 10-point scale questionnaire. Other studies have shown test items with double negatives may prove difficult to translate, or may be confusing (Butcher et al., 1998).
Was this article helpful?