The use of online panels (probability-based or volunteer opt-in) as a mode of data collection has become increasingly popular in market, social, psychological, and medical research (Callegaro and DiSogra 2009). As part of the sign up process, potential panel members are required to complete an online recruitment survey that asks for their basic demographic information. In this study, we analyze data from an opt-in panel in the United States that is composed of respondents who voluntarily sign up (opt-in) to become members of the panel. Like any other online survey, responses to some questions are optional (i.e. skip is permitted), while for others a response is required to advance to the next page. In this paper, we draw upon data obtained from the welcome survey to examine the issue of missing data (i.e., item nonresponse) in the optional set of demographic questions. Specifically, we examine the item nonresponse rate for these questions and dropout rates across all pages in the web survey questionnaire. Then, we attempt to answer one of the fundamental questions related to the practice of survey science "how can we believe what respondents say?" To answer this, we use panel members‘ behavioral data, which is captured from their online and offline activity on their computers, to validate some of their responses to the demographic questions. We construct an overall measure of data quality, defined in terms of item response rate, accuracy of responses, and completeness of responses, and identify demographic predictors of data quality in a multivariate context. Lastly, we discuss the findings from the study and its implications for survey science and conclude with recommendations for future research.