Faced with the difficulty of collecting reliable and representative behavioural data for an acceptable cost, complex survey protocols associating several modes have been developed. Combining different media improve data quality, as it increases the overall response rate, but the question of data comparability remains unresolved. By comparing the results of endogenous selection models estimated using three French mixed-mode travel surveys (two in Lyon in 2006 and 2012-2015 and one in Paris in 2018), we show that it is possible to isolate the impact due to socio-economic differences from that induced by the survey mode.