We analyze Internet and telephone Stated Choice (SC) survey methods in the context of the Swedish value of time study 2008. In this study, extensive piloting and follow-up surveys were undertaken to assure high quality data. We use these data and data from the main survey to analyse properties of the different data collection methods.

We find that Internet-based data collection gives less random error in the SC data than telephone-based data collection. Moreover, the times to respond to each SC question are consistently shorter for telephone respondents compared to Internet respondents. This suggests that the time pressure caused by the interview situation prevents respondents from considering the alternatives carefully enough, perhaps by not taking the time to fill in sheets in order to be able to properly compare both alternatives.  

On the other hand, the response rate drops considerably, to less than 20 percent, in the pilot where the Internet was used as the only response and recruiting mode. Lack of Internet access could not be the only reason for the drop in response rate in the present surveys, given that Internet is available to 82 percent of Swedes in the relevant age group. Hence, the large drop in response rate would to some extent be due to lack of personal contact in recruitment and response.

The main survey was designed as a mixed mode survey where the Internet was the primary method but where respondents knowingly would be subject to a telephone follow up survey or telephone reminder. This survey design is found to give substantially higher Internet response rates. In the mixed mode survey all respondents were initially recruited by Internet and non-respondents were contacted by telephone repeatedly until reached, but at most three times. Respondents were requested to respond to the questionnaire over the Internet, but if they stated that they could not (for some reason) respond by Internet, they were given the chance to respond to the survey over the telephone. Most respondents subject to telephone reminders chose to respond to the survey by Internet, showing that the telephone call in most cases served as a reminder and not a SC data collection mode.

Respondents declining to respond to survey questions by Internet but accepting to participate in a telephone-based survey, differ from Internet respondents in terms of VTT. However, a large part of this bias seems to be explained by socio-economic characteristics, such as income and age, which are cheaper to collect. The fact that we can predict the VTT of respondents declining to respond to survey questions by Internet but accepting a telephone-based survey, suggest that we can do reasonably well without full telephone follow-up interviews as long as information of the socio-economic characteristics of non-respondents is available. Telephone reminders do, however, appear to be crucial.