Satisficing and language proficiency

Especially in multicultural diverse populations, unit nonresponse can be attributed in part to language barriers, if no proxies are available or willing to answer on behalf of people, and if there is no translated version of the survey into the native language of the potential respondent. In such cases, sampled individuals may simply opt out of the survey. However, little is known about the data collected from sampled individuals who are not native speakers of the survey language and who decide to participate in surveys. Specifically, there is little in the literature addressing the data quality of survey respondents who may not be fully proficient in the language of survey administration. Our research examines the question of whether the data of nonnative speakers may  be compromised to some extent due to problems of comprehension of survey questions, especially questions that are more complicated both in terms of content and form. Operationalising data quality within the terms of Krosnick’s satisficing theory, we compared native and nonnative speaking groups in two large-scale Swiss national surveys with respect to a set of dependent measures, including item nonresponse, extreme responding on scales, recency effects, and straightlining. The characteristics of the selected surveys and the questions within them allowed us to distinguish possible effects of language comprehension from other effects on data quality, such as age and level of education, and effects of motivation. Our results indicate that foreigners from neighbouring countries that speak one of the three Swiss national languages (i.e., from Germany, Austria, France, and Italy) tended to produce data on the same level of quality as the Swiss nationals. On the other hand, there was significantly poorer data quality for the different dependent measures for foreign populations in Switzerland that are not native speakers of the available survey languages (German, French, Italian). However, in addition to language proficiency, the poorer quality may be attributable in part to respondent motivation.