Multilingual survey research: Do poor translations cause bias?
Survey results may be biased in multilingual research if consumers are unfamiliar with translated terms, according to a new study in the Journal of Consumer Research.
"Consumers are influenced by the specific labels used to mark the endpoints of a survey response scale. This is particularly important in multilingual research. If the response category labels used in different languages are not equivalent, this could bias survey results," write authors Bert Weijters, Maggie Geuens (both Ghent University and Vlerick Business School), and Hans Baumgartner (Pennsylvania State University).
Many surveys use response scales asking consumers to indicate their agreement or disagreement with certain statements. These agreement rating scales are typically anchored by category labels such as "strongly (dis)agree" or "completely (dis)agree."
In a study with English and French speaking consumers in the US, UK, Canada, and France, differences in familiarity with labels led to different endorsement rates for the endpoints of agreement scales. In both English and French, response categories received more responses when the associated labels were more commonly used in day-to-day language (completely agree or tout à fait d'accord versus extremely agree or extrêmement d'accord).
In another study, self-reported awareness of the cholesterol level of various foods was significantly stronger when using the Dutch equivalent of a scale ranging from "completely disagree" to "completely agree" than when using a scale ranging from "strongly disagree" to "strongly agree."
"Survey researchers should pay more attention to the labels assigned to response categories on rating scales and make sure that the response category labels used in different languages are equivalent in terms of familiarity. Differences in the category labels used in different languages may lead to differences in responses resulting from poor translation," the authors conclude.