Mixed-modes survey media and data comparability issues: a French case study

The propensity for non-responses is tending to reduce confidence that survey results are representative of the studied population. A way to increase response rate is to combine different survey media. An Internet survey of non-respondents to the 2012-2015 Rhone-Alps phone travel survey has been conducted. If web respondent declared fewer trips than phone respondents (3.00 vs. 3.82 trips per day), covered distance was on average longer (24.38 vs. 20.47 km per day) and time budget was similar. These gaps could be explained by a double effect: a higher immobility of the web respondents (18.6%, vs. 10.1%) and a lower statement of their daily trips (3.69 vs. 4.24 daily trips for mobiles). Individuals who respond by phone might be different than those who respond on web. But when sociodemographic differences are corrected by weighting procedure (margin calibration), some gaps have increased. Responses might be not comparable, because the presence of respondents in web or phone sample might be determined by external factors that may also affect their mobility. It is therefore necessary to implement an endogenous selection model, allowing taking into account selection bias. The aim is to estimate the number of trips and the distance and time budgets thanks to specific models, which eliminate the selection bias.