Les déterminants économiques du vote 1976-2006 : trente ans de fonctions de vote en France

The aim of this article is to present a very detailed survey on politico-economic models of voting built in the French case. Since 1976 and the first study ever published on this topic more than seventy papers have been written. The 30th anniversary is the occasion to explore this growing literature. After an introduction that highlights the main features of the scientific production and recalls the theoretical grounds of vote-functions, the first section studies the type of elections and the aggregation level in French voting models. Issues linked to the construction of the dependent variable like left-right cleavage, extreme right status, and relevant round are tackled in the second section. The third one shows the explanatory variables used. They are gathered in three groups: economic variables, institutional variables, and political variables. The conclusion regarding each of them is indicated. The section ends on the role of the campaign and on the French voter?s degree of memory. A paragraph is also devoted to seats-functions that translate vote into seats. This kind of functions is very useful in majority systems like the French one because to be majority in vote does not necessary mean to be majority in seats. Section four deals with econometric problems. After having examined the way that authors treat usual diagnoses (autocorrelation, heteroskedasticity...), two remarks are drawn concerning the use of the intercept term in pooled-data models and the estimation of systems formed by a vote-equation and a seat-equation. Section five presents the forecasts produced by the models. Their accuracy is assessed by comparing the forecasted value to the actual value for some elections and by examining the forecasts stemming from vote intentions polls. This evaluation shows several brilliant successes on the recent period. The conclusion sketches some trails for future researches. Among them, we can cite the building of a finer model for the legislative elections (by constituencies for example) and the development of a model for cantonal elections. Moreover, seats-functions have to be improved and more attention has to be paid to econometrics. Taking spatial interactions into account may also constitute an appealing way to increase our knowledge of voting behaviour.