On-Line Versus Phone Surveys: Comparison of Results For a Bicycling Survey

For decades, travel behavior researchers and transportation planning agencies have relied on phone surveys as the primary means of collecting data on household travel patterns. But these surveys are increasingly problematic. In particular, the shift from land-lines to cell phones makes it more difficult than ever to achieve a representative same. On-line surveys offer an intriguing alternative, particularly given their relatively low cost. But for these surveys, too, sampling is difficult. For a general household survey, no complete sampling frame of email addresses yet exists. Instead, researchers have used letters sent via regular mail to recruit household to participate in the on-line survey. Because not all households have access to the Internet, they may be given the option of requesting a paper survey instead, but this puts an extra burden on the respondent and discourages participation. Non-response bias is thus a serious concern for both phone surveys and on-line surveys. But how do the biases compare between the two surveys? In this paper we examine this question by looking at results from on-line and phone surveys with identical questions conducted in Davis, CA. The primary purpose of the surveys was to measure bicycling and various potential explanatory factors. Although the project did not directly examine the non-response bias in each survey, the project can compare the characteristics of the samples for each survey as an indicator of biases. In addition, we can compare the coefficients of models of travel behavior estimated for each sample separately to assess the degree to which the models yield consistent conclusions. Even though the bicycling levels measured in the two surveys are similar, the characteristics of the sample and the relationships inherent in each could differ. Analyses could yield different conclusions and thus point to different policies.