Reasons for refusals and their collection: Lessons learned from a content analysis of interviewers´ notices in contact protocols
CSDM GESIS Leibniz Institute for the Social Sciences Germany, http://www.gesis.org/das-institut/mitarbeiter-adressen/mitarbeiterverzeichnis/?valpha=&selcat=M%3E%3E&order=sortname&L=&id=145&&selres=16&pagecount=1#16
CSDM GESIS Leibniz Institute for the Social Sciences Mannheim, Germany
Refusals are an important source of non-response in surveys. During field work some surveys collect reasons for refusal in call record data. The European Social Survey even provides such data for secondary research. In addition to providing interesting information for research these data can also reveal some information about non-respondents which could be used to reduce refusals. But, some questions should be posed on the collection of this information: Do people who refuse state valid reasons or only excuses? Is there some usable information? Is there a valid collection of the data? Here two aspects are included: validity of categories used by interviewers and accuracy of data collection done by the interviewer. These questions are handled in this paper. Notices by interviewers regarding reasons for refusals were analysed with the help of content analysis. The data used were provided in call record protocols for the Germany General National Survey (ALLBUS 2008). Prior to this we analysed reasons for refusals in ESS and found that the numbers of undifferentiated outcomes ranged broadly between 0% and 50% across the rounds and countries. Additionally, there were only marginal numbers of several reasons of refusals in numerous countries. ESS applied a category list which included 10 to 13 categories for the collection of reasons for refusals in different rounds. The categories of ESS (blue print and categories applied in Germany) were used as a starting point for the content analysis of interviewers´ notices in ALLBUS. We found numerous new categories, while the percentage of "other" was 1.4%. Additionally, less occupied ESS-categories were marginally occupied in ALLBUS data also. On the basis of these results a schema of categories was developed and applied for the further analysis of ALLBUS 2008 data. Here, correlations with socio-demographics of interviewers as well as interviewer variance were taken into account to find any evidence regarding the question of interviewer impact on data collection. Finally, the results were discussed in light of the posed research questions, the usability of developed schema in surveys, as well as in relation to further research on the topic.