Non-Response Error within Internet Surveys: A Cautionary Note

Abstract

The use of the internet as a method to conduct survey research has expanded rapidly over the past decade. High speeds of response and lower expenses have driven this rapid growth. Relatively low response rates, however, suggest online surveys may be compromised by high levels of non-sampling error. This paper examines a major component of non-sampling error and the consequences that may be associated with internet survey non-response. Known population parameters are compared to point estimates from a census as well as a random sample of non-respondents in order to provide insight on the magnitude and direction of non-response error. Issue salience and response latency are found to exhibit a significant relationship to self-selection and response valance biases. Specifically, lower rates of non-response were obtained from respondents who perceived the topic of the survey as more important and patterns of response were more favorable among initial study participants.