With the worsening of the environment for traditional social surveys, the use of internet research has rapidly increased and has been expanding in academic research including the discipline of geography. However, the availability of geographic information through internet research and the possibility of regional analysis (i.e., investigating geographical distribution of a certain event and its association with another event) has not been sufficiently discussed. In this study, we collected micro data with detailed residential address information through internet research. We then tried to evaluate a sampling bias by comparing the results of statistical analysis for both individual- and regional-level data. The results showed that, although more than half of the respondents answered a question inquiring about their residential address, the response rates to the question varied by their attributes such as educational attainment. However, individual-level analysis of self-rated health exhibited similar results between the acceptance group and the refusal group for providing residential address. These results were also similar to the results based on the nationally representative samples, which were collected by traditional survey methods, indicating that internet research may be useful as an alternative survey method. For the regional-level analysis, we calculated three indices by the internet research by using prefectures as the unit of analysis, and compared them to the results obtained by the population census of Japan. Although only one out of three indices highly correlated between data sources, we suggest that internet research may be useful in cases when geographical distribution of some events or phenomena have large regional differences in nature. Further studies are needed to understand the sample characteristics of internet research.
