Abstract
Segregation in workplaces and individuals' social networks based on ethnicity, race, and/or religion may have serious social and economic consequences. The relationship between social segregation and workplace segregation has been traditionally studied as a one-way causal relationship mediated by referral hiring. In this thesis, an alternative framework is introduced which describes the dynamic reciprocal relationships between social segregation, workplace segregation, individuals' homophily levels, and referral hiring. An agent-based simulation model was developed based on this framework. The model describes the process of continuous change in composition of workplaces and social networks of agents (individuals), and how this process affects levels of workplace segregation and the segregation of social networks of the agents. The simulation results indicated that a labour market may experience significant levels of workplace segregation and social segregation even when the hiring of workers occurs mainly through formal channels. The results also show that majority groups tend to be more homophilous than minority groups, that referral hiring may be beneficial for minority groups especially when the population is highly segregated, and that the relationship between referral hiring and minority unemployment is curvilinear. Levels of workplace and social segregation were found to be negatively correlated with minority proportion, average size of individuals' social network, and firm size, while they were positively correlated with overall unemployment level and hiring discrimination. The research is based on primary data involving structured interviews with a sample of 39 employers and 122 workers (81 Muslim and 41 Coptic workers) in industrial firms in Egypt. Two secondary data sets were also used: the Social Contract Survey (SCS) and Workers' Status in Industrial Enterprises Survey (WSIES). The data were used to assess the levels of social and workplace segregation in Egypt (which found to be high), and to validate the simulation model.