The estimation of the effect of variations in wording as been a very complicated task from many years. This is true for answer scales as well as variation in the main question. This is even more complicated in a multilingual context as Switzerland where at least three languages are fielded in the same time (French, German and Italian) and are supposed to be equivalent.
We have generalised the idea of test/retest proposed by Saris et al in the context of the ESS and propose to test variations by using a reference space where the variations can be projected and their effects estimated. This could be done through Multiple Correspondence Analysis or non linear principal component analysis (CATPCA).
Such a tool will also be discussed in the perspective of alternative techniques