Global Journal of Human Social Science, E: Economics, Volume 22 Issue 3

authors have long been inclined to erroneously conclude in favor of the persistence of the inertia of inequality of opportunity on the basis of the standard log-linear constant association model. However, the latter very often turns out to be incapable of modeling in a parsimonious way the way in which the statistical association between two variables varies according to the modalities of a third variable. Taking this situation into account, in addition to the previous models which allow us to test situations of inequality of opportunity in the classroom and on the labor market separately, we also use the “log-multiplying layer effect model”. Still called the Unidiff model, it was proposed simultaneously by Xie (1992) and Erikson and Golthorpe (1992). It (the Unidiff model) assumes a stable structure of the association between O and D and detects the dominant tendencies of this association. These trends evolve in terms of the variation between the modalities of the education variable (E) through the beta parameter of expression 4 below: Log where OED = (OE)(ED)(OD) (4) In expression (4), is a parameter reflecting the form of the intergenerational statistical association. (or according to the presentation model adopted) represents a parameter which translates for a given level of education k the relative strength of this association. If for a given level of education (such as "no level"), the coefficient β is set to 0 (or 1 when applying the exponential), a negative parameter (or less than 1) for another level of education (for example “primary”) implies that the intergenerational link on the labor market is weaker among people with “primary” level compared to those with “no level”. If on the other hand > 0 (respectively greater than 1) this reflects the fact that the link is more important. In other words, social mobility is lower among less educated people. Under this assumption, the relative odds ratios become Log (). That is to say that the logarithm of each odds ratio is decomposed as the product of a common structure) and a parameter specific to each level of education. In fact, the models mentioned above are nested models. They are comparable with each other based on specification tests of log-linear models. The choice of a model is based on the criterion of the quality of adjustment of the model to the observed data and on the criterion of parsimony. It is the ability of the model to reproduce the observed data with fewer parameters. These criteria are a priori contradictory. In other words, the reference model that assumes independence between variables not only contains fewer parameters, but is the least able to translate reality. To this end, the specification tests therefore consist in examining whether the best predictive quality of the models with the most interaction justifies their large number of parameters. Although log linear models allow us to analyze intergenerational mobility and to determine the place of education in the process of social reproduction, they do not, however, allow us to determine causal relationships. For this, we propose to accompany these models with structural models such as analyzes of causal paths. They also make it possible to test the robustness of the results obtained from the log-linear models. c) Results and robustness As said above, we analyze in this part of our work the effect of family capital on the professional integration of young graduates in Cameroon. It is a question of examining the association that exists between the socioeconomic status of the father and that of his child, the latter having reached a certain level of education. It is also a question of examining the evolution of this association over time. Referring to the segmented structure of the Cameroonian labor market, the labor market status of an individual and that of his father are analyzed in terms of institutional sectors (public, private formal and informal). The structure of the labor market as presented in Cameroon is marked by significant income differences between these different sectors. It is therefore a question for us of testing beforehand the hypotheses of the presence or absence of inequalities of opportunity both at school and in the labor market. In table 1 below are grouped the results. Table 1: Patterns of association (OD) by level of education (E) in 2005 and 2010 EESI 1 (2005) EESI 2 (2010) Modèles df rG 2 p BIC DI df rG 2 p BIC DI (1) OE ED (Indep Conditionnelle) 14 0,00 0,00 -26,9 4,6 7 0,00 0,00 32,0 3,9 (2) OE ED OD (Association constante) 10 82,5 0,10 -68,1 1,6 7 95,5 0,77 -4,2 0,5 (3) OE ED β E OD (UNIDIFF en OD) 7 93,9 0,60 -53,3 0,9 5 96,9 0,73 -8,8 0,4 Source: Author from STATA15 Table notes: G² represents the likelihood ratio statistic, BIC the Bayes information criterion and DI the dissimilarity index which indicates the percentage of people misclassified by the model. Volume XXII Issue III Version I 8 ( ) Global Journal of Human Social Science - Year 2022 © 2022 Global Journals E Family Capital and Professional Integration of Young Graduates in Cameroon