NONRESPONSE BIAS IN MIXED-MODE SURVEYS FOR DISASTER RECOVERY: THE CASE OF MASHIKI FOLLOWING THE 2016 KUMAMOTO EARTHQUAKE

Abstract

 Mashiki town conducted questionnaire surveys following the 2016 Kumamoto earthquake. The survey aimed to capture the disaster victims’ residential preference and predict the demand for disaster public housing. To achieve a high response rate, reminders were sent in a mail-based survey, and additional interview surveys were administered to non-respondents from the mail-based survey. This study aims to examine (1) the effects of these additional efforts in mixed surveys by revealing the household sample differences in various examples of response timing, and (2) the nonresponse bias in estimated public disaster housing demand. We found that households with younger heads-of-households, households with children living in temporary housing, and households in post-disaster public-funded rental housing considering housing relocation tend to delay their response or to be non-respondent for mail-based surveys. Our findings indicate that demand might be overestimated with simple prediction methods. Specifically, the demand by temporary housing tenants was overestimated more than that by post-disaster public-funded rental housing tenants.