Web Survey

Web surveys allow respondents to complete questionnaires that are delivered to them and administered over the World Wide Web. This offers advantages over other survey methods but also poses problems for quality data collection.

A main advantage of Web surveys, compared with other survey methods, is lower costs. Lower costs enable researchers to collect larger samples than affordable with other survey methods and have made Web surveys an option for many who cannot afford other survey methods. Technology allowing for complex questionnaire designs using features available in computer-assisted interviewing programs (such as CAPI, CASI, and CATI) is inexpensive. There are other advantages to Web surveys. The ability to quickly contact respondents over the Internet and process questionnaires through Web sites can shorten field periods compared to other survey methods. Web surveys can be used in conjunction with other methods, giving respondents another option for survey participation. The Web's role as an electronic meeting place has increased the ability to survey some rare, isolated, or marginal populations. In principle it is possible to conduct [Page 956]completely anonymous Web surveys, which may make surveying some populations easier.

Web surveys, however, have limitations that prevent them from being a simple substitute for other survey methods. A major limitation of Web surveys is their inability to represent the general population, or even the population of Internet users, without special efforts or techniques. Internet access, a requirement for participation in Web surveys, is more common among those with higher, as opposed to lower, socio-economic status. This bias means that Web surveys are likely to have substantial coverage error for many survey topics (e.g., those related to health, finances, and education) if they are intended to represent the general population. Further, sampling those who have Internet access is problematic. There is no single sampling frame of the population of Internet users, and it is not possible to construct one because there is no standard system for assigning email addresses, the Internet counterpart to street addresses or telephone numbers.

Researchers have different approaches to dealing with these coverage and sampling issues. Mick Couper has categorized Web surveys according to their methods of respondent selection. Some approaches do not use probability sampling techniques for initial selection of respondents. A common approach is to rely on high traffic through Web sites to produce convenience samples. Another approach is to use Web sites to recruit pools of potential participants who are later sampled for Web surveys. Demographic information about those in a pool can be used for sampling quotas or for post-survey data weighting. These approaches rely on potential participants to opt into a survey or pool, which can produce a self-selection bias. The ability of such samples to represent a full population of Internet users is controversial. Other Web survey approaches use probability sampling methods for respondent selection. When Internet use is high among a population, email lists can serve as sampling frames (e.g., all students, staff, and faculty at a university). Web surveys based on these list samples are potentially representative of those populations. Another approach is to systematically sample users of Web sites using intercept or “pop-up” surveys; those data are potentially representative of users of those Web sites. A probability sampling approach for the population of Internet users could use random-digit dialing telephone surveys or in-person surveys of the general population to recruit Internet users who will then be sampled for Web surveys. Those without Internet access are considered out of the sample. Web surveys based on such samples are potentially representative of the population of Internet users. Finally, the last approach can be modified for conducting Web surveys of the general population. For example, random-digit dialing telephone surveys of a population are used to recruit potential participants for Web surveys. Those without Internet access are not considered out of the sample but are instead provided with Internet access and technology in order to participate in Web surveys. Web survey data collected from such samples are potentially representative of a population.

...