Regional differences in survey response rates and their individual and geographic determinants: A multilevel analysis

The response rate is an important indicator of the reliability and validity of a sampling survey. Recently in Japan, especially since 2005, major social surveys have experienced a substantial decline in response rates, and researchers have become concerned about the worsened survey environment In this paper, we sought to examine the regional differences in survey response rates and their individual and geographic determinants. The data we used were individual data on survey responses to the national representative sample survey with face-to-face interviews in Japan (JGSS: Japanese General Social Surveys) in 2005 and 2006. We distinguished the two-stage process of survey response: (a) whether the investigator was able to meet with those surveyed; and (b) whether they participated in the survey. The differences were then analyzed based on individual and regional characteristics. Large regional differences, measured by the degree of urbanization and neighborhood typology, were observed in both processes. This means that the regional difference in the response rate was also large, because the rate can be obtained by multiplying (a) by (b). We also performed multivariate analysis adjusted for individual attributes such as age, gender, and house type and considering the clustering nature of the samples within the same surveyed area (i.e., multilevel analysis). The results showed that the regional differences in the survey response remained statistically significant although they were partially explained by the respondents' individual attributes. The current study elucidated that the survey response rate to a face-to-face interview is determined by geographic factors as well as individual attributes of the respondents. Additionally, we confirmed regional variations in the survey response rate between surveyed areas which cannot be explained by the degree of urbanization and neighborhood type, indicating the existence of "local survey environments" that have geographically unique and contextual factors that affect the response rate.