Bias in Internet-Based Surveys
- A Comparative Study Using the Census and an Interview-Based Survey-

Given the fact that they are generally less expensive to conduct, Internet-based surveys have today become more widely used than interview-based- and mail-based surveys, which use random sampling methods to select survey participants. However, questions have been raised concerning the representativeness of samples in Internet-based surveys.

In order to clarify the characteristics of Internet-based survey respondents, the project discussed here conducted an interview-based survey using random sampling and an Internet-based survey with target collection numbers for gender and age group, using the same survey items and the same survey timing. The study then compared the results of both surveys with census results.

The analysis revealed that, compared to the census results, the Internet-based survey showed bias in items such as area of residence and educational background, while the interview-based survey showed bias in home ownership rate. Correction based on propensity scores using census questionnaire data indicated that while there were limitations in correction for basic attributes, there was a certain correction effect for variables such as Internet usage time.

President, Nippon Institute for Research Advancement/ Professor, Graduate Schools for Law and Politics, The University of Tokyo
Shoko Omori
Research Coordinator & Research Fellow, Nippon Institute for Research Advancement