Conferences

The debates on response modalities measurement are virtually endless between those who prefer questions which force respondents to take side (the agree / disagree format) and those who advocate for both a median answer and a wide scale of response. Clearly the European Social Survey teams have taking side since the questionnaires contain a high number of response modalities ranging individuals on 0 to 10 scales. This preference in itself is justified but creates several substantial and statistical problems when using methods of Geometric data analysis, particularly specific Multiple Correspondence Analysis (Le Roux, Rouanet, 2004).

Most of the problems are caused by the median answer (5) chosen regularly by a high proportion of respondents. This kind of distribution creates quite systematically a Guttman effect, a well-known problem in GDA which ends up with a factor opposing extreme and median answers. Substantially, this effect challenge the aims of MCA: basically 1) to identify the main cleavages upon which the electorates oppose and 2) to sort out groups of respondents on value dimensions.

We propose a solution of this Guttman effect based on the Benzecri’s design (Benzecri, 1980) and apply it to the ESS round 2 and 3 French data. This method helps to solve the Guttman effect’s problem and give substantial results regarding the political and normative space of the French respondents on dimensions such as politics, Europe and globalization, ethnocentrism, institutions, economy.