Socio-economic and urbanization profiles of internal migration in Egypt: A canonical correlation analysis

Data on in and out migration in Egypt by governorate and on variables believed to interact with migration are examined. The purpose is to identify the dimensions through which the basic demographic phenomenon can be rationally structured. This objective is approached through a general multivariate analysis technique known as canonical correlation analysis. It is an appropriate tool for investigating relationships between two sets of variables, one of them may be thought of as a set of dependent variables and the other as a set of independent variables. The basic step in a canonical correlation analysis procedure is the generation of orthogonal paris of canonical variates. For any pair, the first variate is a linear combination of the dependent variables and the second is a linear combination of the independent variables. Once these variates are generated, a number of criteria are applied to assess the type and strength of relationship between the dependent and the independent variables. For the purpose of migration analysis, in and out-migration rates by governorate are used as a set of dependent variables. A number of variables believed to act as determinants of migration are used and arbitrarily classified into four groups to represent four distinct migration profiles, namely, socio- demographic, economic, urbanization, and health profiles. The SAS computer program CANCORR is used to generate the information necessary for analyzing these profiles. The analysis shows that the economic profile of migration is the most pronounced one of the four profiles examined. Canonical correlation between the migration and the economic variates is at its highest level (0.926) and so is the amount of variance overlap between them (85.8%). Next to it are the socio- demographic and the urbanization profiles. On the other hand, canonical correlation between migration and health variables does not significantly differ from zero and the percent of variance overlap is very low (17%). Migration canonical variates within the four specified profiles extract between fifty two and sixty percent of the variance in their own variables and between forty and forty nine percent of the variance in the set of opposite variables. An exception to this last result is the low amount of variance in the health variables explained by the migration variate. This is another indication to the weak relationship between migration and health indicators. The amount of variance in the independent variables explained by the canonical variates is much lower, ranging from 19% to 65% for own variates and from 11% to 30% for opposite variates. Finally, there is an indication for several independent variables across ne socio-demographic, the economic, and the urbanization profiles to exert larger interaction with migration variates than do other variables. This is evidenced by the high correlations between these variables and both own and opposite canonical variables. Among these variables are the proportion of illiterate population, the proportion of population engaged in occupations, and type of occupational affiliation. The above findings, however, summarize the specific migration experience in Egypt in the early eighties. Their validity is hinged upon considerations of quantity and quality of available baseline data.  The procedure of canonical correlation analysis proves to be a very useful technique for illuminating relationships of complicated or general nature such as the ones dealt with in this paper.