Web Survey Bibliography

Title Length of Input Field and the Responses Provided in a Self-Administrated Survey: A Comparison of a Paper&Pencil and a Web Survey
Author Fuchs, M., Couper, M. P.
Source Methodology and Statistics, 2001
Year 2001
Access date 22.05.2004
Abstract It belongs to the common wisdom of survey methodology that respondents react to the question wording of a particular item when searching a response. In addition, respondents also react to other more formal aspect of the presentation of an item, e.g. question order, response order, numbers and symbols associated with the response categories etc. In this experiment we tested the length of an input field for numeric information as an independent variable. In a previous experiment in web survey conducted by Couper it was found that longer input fields lead to more instances where the respondent provides ranges or estimates of the correct answer. This lead to the more general hypotheses that longer input fields or lines/boxes for the responses reveal more extensive information. The more space is provided by the designer of the questionnaire the more the respondent assumes that extensive information is expected by the researcher. In a self-administrated survey on right wing attitudes among Germany high school students (n=5,042) two versions of the same questionnaire were used. The versions were randomly assigned to the respondents. In one version each input field or line/box had twice the size compared to the other version. In addition both versions were administrated on paper as well as on the Internet. The resulting 2x2 design allows a detailed assessment of the effect of long vs. short response boxes or input fields. In Addition, data is available for a comparisons of a paper&pencil and a web questionnaire in terms of that particular effect. During the course of the questionnaire 6 items were included in the experiment. The items differ in the degree of salience of the response (very low to very high). The results indicate that respondent react to the length of the input field or the line/box when the response is not readily stored in their minds. When respondents need to guess or estimate the response they use different strategies to indicate the week status of their response: (1) they use ranges instead of a single value (’between 10 and 20’), (2) they qualify their response as an estimate (e.g. ’about 10’, ’˜10’), (3) they provide responses that are easily detected as estimates (bunching/hyping). In addition, respondents provide lots a additional information associated with the response that. E.g., when asked how long it took to answer the questionnaire a respondent might answer ’45 minutes, it was an awful long and boring questionnaire’. From our results we can draw the conclusion that longer input fields induce more ranges, more estimates and more additional information. Furthermore, with longer input fields or lines/boxes we have in decrease in the amount of bunching/hyping. In summary, the results suggest that respondents use different strategies when dealing with uncertaincy: long input fields and lines/boxes lead to more ’ranges’ and ’explicit estimates’, short input fields and lines/boxes show more hyping and bunching. This is especially true for items where the respondent needs to guess the correct answer. The comparison of the paper&pencil and the web questionnaire indicates that this effect shows up differently in a web environment. In a computer assisted situation respondents use far less ’ranges’, ’explicit estimates’, and provides less ’additional information’. On the other hand bunching/hyping seems to be the same or slightly higher. From a researchers point of view both approaches (long an short version) have different trade-offs: confronted with long fields and lines/boxes respondents qualify their response as an estimate. As a result more editing is necessary. Short input fields and lines/boxes on the other hand lead to less explicit estimates and show more hyping/bunching. This requires less data editing but leads to a situation where a researcher takes a response for granted were he or she should be more careful.
Access/Direct link Homepage - conference (abstract)
Year of publication2001
Bibliographic typeConferences, workshops, tutorials, presentations
Print

Web Survey Bibliography - Methodology and Statistics, 2001 (4)