Internet mobility survey sampling biases in measuring frequency of use of transport modes

We develop a quantitative analysis of the biases that arise when measuring trip frequencies for a general population through an online survey instrument. Data from a national official survey in Italy, concerning both mobility behaviors and skills in using computers and internet, have been deployed to assess differences in mobility levels between those that can answer a computer/internet survey and those that cannot. Positive correlations were found between ability in using ICT tools and trip frequencies. These latter are about 15% to 150% higher for the “ICT literate”, according to the travel means under consideration. A Heckman sample selection model showed us that these biases have different explanations. People knowing how to use internet are different from the others in they car driving behavior due to a range of self-related factors. Conversely, public transport patterns of use are more similar between the two groups: the observed bias is mainly due to the fact of using internet in itself, which could for example lead to a more active lifestyle. Such distinction is of practical interest because it can help defining a method to correct these biases. According to our results, the overestimation of public transport frequency of use of an internet survey could be corrected by looking at the internet diffusion in the population. On the contrary, corrections for car driving frequencies are more complex and should be based on differences in attitudinal and personal characteristics between internet survey respondents and the remainder of the population.