Very few studies have attempted to relate the properties of some ordination techniques to classical tools of population genetics as F-statistics. A multivariate model to analyse population genetics data based on the properties of ‘joint scaling’ of populations and loci is developed. The design of population genetics data means that this model deals with a modified version of the classical Multiple Correspondence Analysis which is called Constant Row Total-Multiple Correspondence Analysis (CRT-MCA) and is an original tool in population genetics. Such a model allows estimates of the degree of population differentiation by studying the variability of the distribution of allele frequencies in different samples. Some clear relationships exist between some model parameters and the classical Fst statistics. The CRT-MCA also allows all the studied loci to be considered simultaneously and the role of each locus in patterns of population differentiation to be expressed. Such a multivariate approach prevents the use of any pooling strategy as is classically used in studies of hierarchical F-statistics. The relevance of the CRT-MCA model is illustrated by the analysis of population structure of 15 dogwhelk (Nucella lapillus) populations in south-west England. The advantages and limitations of CRT-MCA are presented.