By Nichols. E and B Sedivi.

In this paper, a study of collecting data via the Web by the Census Bureau is introduced and conclusion on the results of this study is made.

First the background of this study is explained. Since the rapidly development of web techniques, using a CSAQ (Computerized Self-Administered Questionnaires) over the internet has several advantages over diskette CSAQ. So several question arise. What is the appropriate software to use for a Web CSAQ? How should one design the questionnaire? And, what is the respondent's perception of the willingness to use the internet for this purpose? Therefore the Census Bureau investigated whether we could collect data confidentially over the Web with a CSAQ that met respondents' needs.

Then the details of the design of the 1996 R&D Web CSAQ are described. Unlike the diskette CSAQs, which are saved to the respondent's hard drive and execute on his/her own PC, an interactive CSAQ is stored on the Web server. Data are stored on the server as well. The respondent's machine only acts as a display and entering device. Username and password are needed to access the Web CSAQ. Netscape Web server software is used.

Next the methodology for the pilot test is described. First a paper screener questionnaire was mailed to all respondents to determine the eligibility in reporting on the Web CSAQ for the 1996 R&D survey. Then fifty companies are chosen to receive the Web CSAQ package for this pilot study. Then the 50 selected Web CSAQ test companies were notified of their selection by mail at the same time the paper questionnaires and diskettes were mailed. Follow-up letters were mailed on schedule. Telephone follow-up of delinquent Web CSAQ was conducted.

Next the results of screener questionnaire, response rate, help desk, evaluation questions and user metrics are reported and analyzed. This is in very detail, I will not repeat the results here.

Later in this paper, the three questions asked at the background section are discussed based on the results of the test. For the first question, the conclusion is there is no ideal software for developing CSAQs so far due to browser configuration issue. For the second question, it is said that the questionnaire design used in this test is a reasonable baseline model. The Windows based design is more favorable to an earlier Dos CSAQ design. Adding username and password in the follow-up mail is important. As for the third question, it is said that most people are willing to use it to report their Data. However, a small percentage have security concerns about transmitting sensitive data over the Web.