Web Survey Bibliography
Title A Comparison between Using the Web and Using the Telephone to Survey Political Opinions
Author Forsman, G., Isaksson, A.
Year 2003
Access date 07.05.2004
Abstract In a political opinion poll before the Swedish September 2002 election, the commercial survey institute TEMO (one of the largest commercial survey institutes in Sweden) experimented with two different modes of data collection: telephone and Web. The methods were applied to independent samples. The telephone sample was the ordinary TEMO Omnibus sample - an approximate probability sample of phone numbers combined with a probability based selection of individuals within households. The Web sample was drawn from a panel of Web users recruited via earlier TEMO Omnibus samples.
In our paper, weighted answers to selected questionnaire items, in particular voting intentions, are compared. The purpose of the weighting is to lift the sample data to the population level. Demographic variables are used for post-stratification, as well as voting behavior in the previous Swedish election. In addition we calculate for both samples propensity scores from specially designed “lifestyle” questions, aimed at identifying subgroups of the population within which Web users and non-users have similar voting intentions. The propensity score distribution of the Web sample is then “calibrated” against the corresponding distribution of the telephone sample. This technique may reduce the potential bias of the Web sample estimates due to a high drop-out rate in the panel recruitment process. The demographic composition of the two samples is also analyzed, and the outcome of the election is used for additional comparisons.
In our paper, weighted answers to selected questionnaire items, in particular voting intentions, are compared. The purpose of the weighting is to lift the sample data to the population level. Demographic variables are used for post-stratification, as well as voting behavior in the previous Swedish election. In addition we calculate for both samples propensity scores from specially designed “lifestyle” questions, aimed at identifying subgroups of the population within which Web users and non-users have similar voting intentions. The propensity score distribution of the Web sample is then “calibrated” against the corresponding distribution of the telephone sample. This technique may reduce the potential bias of the Web sample estimates due to a high drop-out rate in the panel recruitment process. The demographic composition of the two samples is also analyzed, and the outcome of the election is used for additional comparisons.
Access/Direct link Homepage - conference (abstract)
Year of publication2003
Bibliographic typeConferences, workshops, tutorials, presentations
Web Survey Bibliography - Forsman, G. (9)
- On the Cost-Efficiency of Mixed-Mode Data Collection with a Web Response Option: Results of a Survey...; 2005; Werner, P., Forsman, G.
- Mixed Mode Data Collection Using Paper and Web Questionnaries. A Cost and Response Rate Comparison in...; 2005; Werner, P., Forsman, G.
- On the Variability of Estimates Based on Propensity-score-weighted Data from Web Panels; 2004; Forsman, G., Danielsson, S. , Isaksson, A.
- Comparing Propensity Score Weighting with Other Weighting Methods: A Case Study; 2003; Forsman, G., Varedian, M.
- A Comparison between Using the Web and Using the Telephone to Survey Political Opinions; 2003; Forsman, G., Isaksson, A.
- Doec propensity score weighting work for Web Surveys; 2002; Forsman, G., Varedian, M.
- Mail and Web Surveys: A Cost and Response Rate Comparison in a Study on Students Housing Conditions; 2002; Forsman, G., Varedian, M.
- Does Propensity Score Weighting Work for Web Surveys?; 2002; Forsman, G., Varedian, M.
- The design and analysis of reinterview: An overview; 1991; Forsman, G., Schreiner, I.