Compared to random digit dial telephone surveys, web surveys have major advantages in terms of cost and speed. The major disadvantage of web surveys is that it is usually not possible to obtain a probability sample from the target population. We com-pare cost, time-to-complete, weighted responses, demographic composition and cost from a large-scale survey about health care that was conducted simultaneously and in parallel over the phone and on the web. The web version used propensity methodol-ogy, a statistical technique that with certain assumptions does not require a probability sample for drawing valid inferences - which makes it well suited for use with web surveys. The two samples are different with respect to key demographic variables. Population estimates for response items turn out to be similar in some cases, and very different in others. Differences are re-duced when only comparing users who use the Internet daily or sub-populations that are well represented on the Internet.