Web surveys are becoming more and more popular in survey research, mainly because their costs are usually lower than for other modes of data collection. With the increase of the internet coverage of the population in most of the European countries, the response rates obtained are also becoming high enough to collect huge amount of data and in a short period of time. However, the problems of comparability of the results with studies using more traditional modes of data collection (across time or countries) and of the potential di erences in quality lead to several studies comparing data collected using web to data collected with mail or telephone or face-to-face surveys (Kaplowitz, Hadlock, Levine, 2004; Fricker et al., 2005; Heerwegh, Loosveldt, 2008; Heerwegh, 2009; Kreuter, Presser, Tourangeau, 2009). Our analyses are in this line and compare one study completed by the LISS panel (Web) with the same questions asked in the frame of the European Social Survey (face-to-face) in the Netherlands (Rounds 1, 2 and 3). First the general design and response rates for the surveys are compared, as well as the main characteristics of the samples with respect to some classic background variables. Then, we focus on an example of a composite score (using variables about immigration), and test for invariance across the di erent surveys before to compare the means and the qualities of this composite score. We conclude that so far, it seems possible to compare composite scores between the Liss panel and the ESS rounds. The expected mode e ects are not really found. This could be linked to the speci c design of these surveys: more than the mode itself, the way of implementing the survey might create di erences.