Policy making in the area of immigration and ethnic affairs relies increasingly on the findings of social science research. A new "conventional wisdom" is emerging in sociological and economic studies. Its claim to objectivity is based on the use of a methodology which emphasises the use of sophisticated statistical methods, especially multiple regression models. The present paper suggests that this apparently objective approach is actually based on a set of of subjectivist and voluntaristic assumptions on: the nature of society, the character of the groups being examined, and on the practice and logic of scientific research.

This type of analysis leads to the reconstitution of the "ethnic group" through the use of dummy variables, which mask many of the specific structural features of labour migrants. Moreover, the emphasis on multiple regression models makes it necessary to quantitify the variables to be examined. This involves a subjective decision by the researcher: either to give a numerical value to something for which there is no single correct form of quantification, or to exclude it altogether from the analysis.

The approach is based on a human capital model, in which migrants are seen only as economic subjects in a rational market system. This blocks historical understanding of the way labour relations function in capitalist societies, in particular of the significance of labour migration, and of the role of segmentation of the labour market according to ethnciity and gender in class relations.