Contenu du sommaire

Joachim Marcus-Steiff : Sales effects of advertising. Some results of the analysis of « natural » data. « Proofs » of the effectiveness of advertising on sales are scarce and, most often, questionable. Particularly in the case of « natural » data, the correlations frequently reported between sales figures and advertising expenditures seem to reflect the influence of sales on advertising rather than the contrary. However, when practically identical products are offered under different brand names, advertising might increase the share of the market of a given brand, the total sales of the product (all brands cumulated) being indépendant of advertising expenditures. These conclusions are tentative, as the number of data analyzed in this paper is small.

Pierre Kende : The advertising myth. On the macro-economic scale (growth, consumption, economic cycle) advertising does not seem to have any of the influences of which it is credited. It is true that the economist considers advertising expenses as synonymous with advertising action; however, two screens separate the first from the second: the media and the message. Advertising is primarily communication and therefore, it is not absurd to think that its efficiency is essentially related to the content of the « message ». Compared study of European markets shows that advertising expenditure is a relatively stable structural factor; it depends on the form of competition, rhythm of innovation, tyoe of distribution etc. and perhaps also on certain cultural habits.

Jacqueline Feldmann-Högaasen : Dependence analysis : regression and correlation. This article presents the dependence analysis, and different formulas relating dependence, regression and correlation coefficients. It gives a brief discussion of the cases where structural zeros or cycles are present. This type of analysis supposes the (causal) ordering of the variables given. Conversely, it is possible, starting with the observed correlation coefficients, to deduce an order on the variables, using appropriate criteria. Two examples are given.

Jean-Michel Chapoulie : A type of explanation in sociology : the systems of variables in causal relations. The purpose of this study is the examination of theoretical and statistical problems posed by the analytical methods of relations between variables proposed especially by Simon, Blalock, and Boudon. The theories of causality justifying the statistical models constructed by the methodologists seems insufficient : or rather ; their absence of logical precision authorizes illegitimate use of statistical methods, or rather, by defining in logical mathematical terms, the concept of cause, these methods cannot guarantee the construction of any model. To avoid these difficulties, which limit the application of statistical technique to sociology, such as the analysis of regression and the recessive models, a definition of the notion of causal system is proposed which tries to make coherent the relationship between the conceptual definition of causality and its translation into statistical language.

Anselme Flavigny : Notes on dependence analysis. In his book, L'Analyse mathématique des faits sociaux, R. Boudon proposes a method which he calls « dependence analysis ». This method permits the analysis of complex structures with many variables. By using a more simple mathematical apparatus, based on the composition of conditional frequencies, the same conclusions are reached. This method permits great flexibility in interpreting complex structures because it stays close to the reality of the research in question.