Principles for constructing web surveys

There can be little doubt that the number of surveys being conducted over the World Wide Web is increasing dramatically. The ability to collect large amounts of data without interviewers, stationery or postage, and to process answers without separate data entry, makes the cost of doing web surveys very attractive (Witt, 1998). The prospects of moving into an era of web surveying are in other respects, quite sobering. Although the number of individuals with web access is increasing rapidly, in October, 1997, only 37 % of U.S. households reported owning a computer, up from 24 % in 1994. A total of 17 % had e-mail connections, compared to only 3 % in 1994. (National Telecommunications and Informa-tion Administration, 1998)2. Therefore, substantial coverage problems exist for most household web surveys. Access is higher when considered without reference to households. For example, a series of five CBS News/New York Times Polls conducted by telephone from March to November 1998 reported that 67 % of respondents had “access to a computer ” and 31 % “had an e-mail address through which you can send or receive electronic mail messages by computer.” (Frankovich, 1998). Even these higher numbers are much too small to facilitate general public surveys. Although certain survey populations have nearly 100 % access to e-mail, e.g. employee