Tackling With Item-Level Missingness Among Older Respondents: Evidence From World Values Survey/European Values Study Joint 2017

Abstract

The thesis examines item nonresponse, one of the data quality indicators, by focusing on individuals aged 50 and over. It shows how item nonresponse varies by age, socio demographic characteristics, country affiliation, and survey design variables, specifically for religiosity/religious values, institutional confidence, and retrospective questions, using binary and multinomial logistic regression analyses separately for individuals under 50 and those aged 50 and over. In this study, the EVS/WVS Joint 2017 data set, which is obtained by combining the World Values Survey and the European Values Study data sets, including data from Turkey, Zimbabwe, Italy and Switzerland, is used. The findings indicate that the occurence of item nonresponse varies by country, age group, and question group. Additionally, independent variables such as sex, education level, subjective health status, marital status, region, immigration status and remaining variables are also statistically significant. Women, those with lower education levels, and those reporting poor health are more likely to leave these question items unanswered. Among individuals over 50 and those under 50, women are more likely than men to leave confidence questions unanswered. The same pattern is observed among individuals with lower education levels. In Switzerland, immigrants aged 50 and over are more likely to leave both confidence and retrospective questions, while immigrants under 50 are more likely to leave retrospective question unanswered. Furthermore, item nonresponse is less likely in self-administered surveys in Switzerland. The thesis results indicate that item nonresponse varies by country, age group, and question type. For older respondents, additional survey modules, supplementary qualitative interviews, and conversational interviewing are presented as survey research methods, and detailed studies on these methods are recommended. Rather than offering general methodological recommendations covering all countries, age groups and question groups, it is emphasized that these methods should be tested in different contexts and survey settings.