Web surveys have been included in many national longitudinal studies and panels as an additional option offered to respondents together with, or instead of, traditional modes (e.g. CATI or CAPI). Web surveys cost less and this mode generally requires a shorter data collection time compared to the traditional modes of data collection. Evidence is mixed

regarding the presence and size of coverage error and nonresponse error in Web surveys. Typically, there is little information about non-respondents to allow for the measurement of nonresponse bias. This paper takes advantage of the 2011 Health and Retirement Study (HRS) Internet Survey to examine factors affecting people’s decisions to respond to a Web survey. Respondents to the 2011 HRS Internet survey are randomly selected from respondents to the 2009 HRS internet survey and the 2010 HRS respondents who had Internet access. As a result, a lot of information is available on both respondents and non-respondents to the 2011 HRS Internet survey. Making use of the rich auxiliary data, we first characterize non-respondents to the 2011 HRS Internet survey. Specifically, we examine whether health status is a factor that might influence respondents’ willingness to participate in the Web survey after controlling for age, gender, and socioeconomic status (SES). We hypothesize that respondents with lower self-rated health in previous waves were less likely to participate in the 2011 Internet survey. We then explore the role of health status in nonresponse bias in several key survey variables from the 2011 Internet survey for older adults.