Sources of nonresponse error in the translation process of survey instruments: the impact of language mismatch and on-the-spot translation on the quality of birth date data

ABSTRACT

In data collection process of Turkey Demographic and Health Surveys (TDHSs), only printed questionnaire in Turkish is used. Therefore, on-the-spot translation was performed during the data collection process, rather than using printed questionnaires in Kurdish or Arabic. From the item-level nonresponse perspective, this study examines the quality of birth date data with two different but interrelated angles (1) impact of the language mismatch between language of interview and mother tongue of women; (2) impact of on-the-spot translation from Turkish into Kurdish and Arabic. The findings suggest that interviews with language mismatch and interviews conducted with on-the-spot translations, especially translations performed by local interpreters tended to produce high item-level nonresponse when compared with matched interviews. For this reason, an organized interviewer training session for increasing the effectiveness of on-the-spot translation with explanatory documentations and/or printed questionnaires in Kurdish and Arabic may be considered as options to decrease the item-level nonresponse further.

KEYWORDS:

• Item-level nonresponse
• on-the-spot translation
• language mismatch
• data quality
• Turkey DHS

Acknowledgments

We are grateful to the DHS Program and Hacettepe University Institute of Population Studies (HUIPS) for allowing us to use TDHS data sets. Additionally, we are indebted to the Hacettepe Teknokent Technology Transfer Center (HT-TTM) for the advanced editing service.

Disclosure statement

There is no conflict of interest.

Supplementary material

Supplemental data for this article can be accessed here.

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.

Notes

1. Although the printed questionnaires of the TDHSs were prepared in Turkish only, there does not exist any impediments on the language of survey materials. Since the official language is Turkish in Turkey, researchers generally assume that all the individuals can speak Turkish and understand the questions in a certain extent. However, as we shown in the paper this is not the case.

2. The Arab population in the TDHS-2013 was the native Arabs who are living in Turkey from the past. Although migration of Syrian refugees to Turkey has started in 2011, the main influx started mainly after 2013. Therefore, the migration flows of Syrian refugees to Turkey had no impact on the ethnic distribution that calculated from TDHS-2013.

3. All TDHSs cover a question on the ethnicity of the respondent. Additionally, if the respondent is ethnically Kurdish or Arabic, TDHSs ask a question on whether respondent speaks Turkish or not. However, this question gives very limited information about the individual ability of speaking Turkish.

Additional information

Funding

We have no specific funding for this manuscript.

Notes on contributors

Melike Sarac

Melike Sarac is a research assistant at the Department of Social Research Methodology, Hacettepe University Institute of Population Studies. Her research interests include survey methodology, survey quality and total survey error, nonresponse, interviewer characteristics, interviewer and respondent relations, data quality, respondent selection techniques, social desirability bias, infertility, and ARTs use and mismeasurement of survey responses.

Ismet Koc

Ismet Koc is a professor at the Department of Demography, Hacettepe University Institute of Population Studies. His research interests include demographic techniques, population policy, population and development, and family demography as well as survey methodology, social desirability bias, and methods of survey response mismeasurement.