Abstract

When the non-response rate in a survey increases, the quality of the survey’s statistics becomes degraded due to the occurrence of a non-response bias. We can pre-suppose that the survey interviewer’s career length – as a measure of one’s experience in the field – is one of the determining factors that can lower the non-response rate. There are relatively few studies, however, on the relationship between an interviewer’s experience and the non-response rate.
In this study, I divided a group of interviewers in charge of a particular survey in the Seoul metropolitan area into three groups based on how long they served as interviewers. I then investigated the non-response rate for an existing household survey group and a new household survey group. This took into consideration the sample replacement characteristics of the economically-active population survey in Seoul, South Korea. First of all, I confirmed whether there was a relationship between the interviewer's career length and the respondents' response status in the economically-active population survey. Subsequently, I analyzed whether there was a significant difference in the average non-response rate for each interviewer’s career length group among the existing survey group and the new survey group.
The statistical assumption review of the variance analysis and the results of confirming the p-value are as follows. First, when analyzing the non-response rate of the existing survey group by the interviewers’ career length, the average non-response rate was lower for the group with the most amount of experience. At a 95% significance level, however, it was not statistically significant that an interviewer’s longer length of service produced a lower non-response rate. Based on this information, we did not discover a meaningful relationship between the interviewer’s career length and non-response rates while allocating survey tasks to interviewers for an existing household survey group in progress.
When persuading the new survey group, however, we confirmed that the difference in the non-response rate according to the interviewer's career length was statistically significant at the 95% significance level. Through the post-hoc test, there was a significant difference in non-response rates between the group with less than three years of experience and the group with more than three years, but less than five years of experience. Through this study, we deduced that when trying to persuade new respondents to participate in a survey, it is possible to expect a positive effect on reducing the non-response rate by allocating work based on an interviewer’s length of service.