Handbook of Web Surveys. Jelke Bethlehem
Two About Web Surveys
2.1 Introduction
The Internet is one of the data collection tools available for conducting surveys. It is a relatively new method. At first sight, it is an attractive mean of data collection because it offers a possibility to collect a large amount of data in a short time period at low cost. Therefore, web surveys have quickly become very popular. Due to the large diffusion of mobile devices connected to Internet, when sending a web survey, it is possible that the invitation and even the completion takes place on a mobile device like a tablet or, especially, a smartphone. Thus, in fact, a web survey becomes a mobile web survey. In this handbook we use the term web surveys, and web surveys only when we want to stress the importance of mobile devices we use the term mobile web survey.
The methodology of web surveys is not yet fully developed. Nevertheless, a lot of experiments and scientific discussion about the theory of web surveys are now in the literature. They are useful, because one can determine the advantages and the problems in doing a web survey. Therefore, the statistical studies described in this handbook are important for the future of web surveys and for those who are interested in learning about web surveys and how to run them.
Traditionally, surveys can use various modes of data collection:
By mail using paper questionnaire forms.
By telephone. The interviewer can use a paper form or a computer program for computer‐assisted interviewing (computer‐assisted telephone interviewing [CATI]).
Face‐to‐face. The interviewer can use a paper form or a computer program for computer‐assisted interviewing (computer‐assisted personal interviewing [CAPI]).
Web surveys resemble mail surveys. Both modes of data collection rely on visual information transmission. Note that telephone surveys and face‐to‐face surveys use oral information transmission. Furthermore, there are no interviewers involved in data collection. Data collection is based on self‐administered interviews.
Of course, a web survey is a computer‐assisted form of data collection (like CAPI and CATI). Therefore, sometimes it is given name CAWI (computer‐assisted web interviewing). Web survey questionnaires can include features like automatic routing through the questionnaire and automatic checking for inconsistencies. These features are not possible for mail survey questionnaires.
Like for any other survey, also web survey respondents have to be contacted first and invited to participate in the survey. In general, the following approaches are possible:
Send an e‐mail or an SMS (cell text message) with a link to the website containing the survey questionnaire. The link may include a unique identification code. The unique code ensures that a respondent will complete the questionnaire only once. It also ensures that only selected individuals complete the questionnaire. Example 2.1 describes a web survey that applied this approach. At the time being, e‐mail and SMS can be read on mobile devices. Consequently, the questionnaire could be completed on mobile devices as well, or alternatively the contacted individual can postpone the completion when he will access to a laptop. This is a challenging situation because factors like the environment and contest where the invitation to the survey (i.e., the e‐mail or the SMS) is received and the time lag between the invitation and the possibility to use a laptop for completion and other factors are requiring special attention.
Send a letter by ordinary mail inviting the potential respondents to go to the survey website. The letter contains the address (URL) of the website and a unique code. Again this guarantees that only the proper individuals participate in the survey.
Catch potential survey participants on the Internet when they are visiting a website. They are invited to click on a link or button to start the survey. They may be directed to a different website containing the survey, or the survey starts as a newly opened window (pop‐up window) on the screen. The web survey may also be embedded in websites visited by the individual.
The last approach is a very simple way to conduct web surveys. No e‐mails or letters have to be send. However, it has the disadvantage that no proper sampling procedure is used. This may lead to a response that lacks representativity. Sometimes such surveys may allow a respondent to complete the questionnaire more than once. Moreover, there is no guarantee that each respondent is a member of the intended survey population. Finally, technical aspects, like pop‐up blockers, may prevent starting the survey questionnaire.
EXAMPLE 2.1 A web survey on technological communication and links between enterprises
The survey is carried out within the survey research activities now provided from the CASI (Center for Statistical Analyses and Surveys/Interviewing, Centro Analisi Statistiche e Indagini) seated at the ex‐Department of Mathematics, Statistics, and Informatics of the Faculty of Economics of the University of Bergamo. The survey topics were the use of e‐commerce, the collaboration with other enterprises and/or the belonging to groups, the markets, and the employment. The questionnaire was kept simple (6 pages of 40 substantial items, 1 welcome page, and 1 final page) and asked mainly for qualitative answers or for percentage data.
The survey addressed to about 2,000 firms of the provinces of Bergamo (used as pilot province), Brescia, Lecco, Varese, and Mantova (each province is in the Lombardy region in Italy) in the manufacturing and building sector. E‐mail addresses and stratification variables collection is from the administrative databases of the Chamber of Commerce for the Bergamo province and from a Unioncamere (Union of Chambers of Commerce) database for the other provinces.
The overall response rate was 21.9%, which is quite high considering that due to the quality of the list, 12% of the follow‐up contacts were explicitly wrong e‐mail addresses. Response rates by size of the firm, legal form, and economic activity did not differ very much. It is interesting to note that for small firms (less than 20 employees), the response rate was quite high compared with the other firm sizes.
Data collection took place, for the first wave, in spring 2000 and has been carried out according to the following steps:
Invitation to participate in the survey sent by e‐mail (survey presentation letter, a survey report as incentive, and other related advantages have been prospected).
A link of the individual firm address for completion of the questionnaire was included in the presentation letter. Therefore no identification code (id) and no password was required.
Three e‐mail reminders were sent (first reminder 14 days after the survey follow‐up and the two reminders with weekly periodicity). Mainly for research purposes a fourth e‐mail remainder was sent after the end of the survey period. As described in literature, the three reminders were effective in improving the response rate; the forth reminder did not have an effect.
A web survey questionnaire consists of one or more web pages. Respondents have to visit the website in order to answer the survey questions. Note that, at early times of web surveys, the Internet was only deployed as a medium to transport the empty questionnaire to the respondent and to transport the completed questionnaire back from the respondent to the survey agency. For example, a simple questionnaire form is implemented in an Excel spreadsheet. The respondents receive the spreadsheet as an attachment of an e‐mail. After they have downloaded this electronic form, they fill it in on their own computer. Completed questionnaire return is also by e‐mail. This type of survey is called an Internet survey, because it uses the Internet in a much broader sense, than just the HTML pages of the World Wide Web.
This chapter describes the various forms of online data collection, from simple e‐mail surveys to advanced web surveys. It shows how to use web surveys for different target populations, for cross‐sectional data collection, and for longitudinal data collection