The Impact of Return Migration and the Determinants of Lifetime Migration in Low Income Countries
To study the impact of return migration on the urban labour market, we extend the basic monitoring model of efficiency wages to include two urban sectors (formal and informal) and a rural sector, and a labour force composed of permanent (or resident) and temporary (or migrant) workers. The proposed model provides a micro-economic foundation for a rural-urban wage gap as well as an intra-urban formal-informal sectoral wage structure. Unlike previous models, our explanation is based on imperfect information on the part of the employer instead of the employee. The proposed model appears to conform closely with stylized facts as discussed in the first chapter.
The difference in expected tenure between resident and migrant workers provides a rationale for discrimination by formal sector employers in favour of resident workers, who then have an incentive to engage in signalling, usually by choosing unemployment over informal employment. In this way, the model suggests a non-human-capital based explanation for the incidence of luxury unemployment and predicts that migrants tend to be less likely to endure periods of unemployment. Under certain specified conditions, informal sector wages may be higher than formal sector wages, and informal sector wages may be negative.
Using the 1988 Malaysian Family Life Survey (MFLS-2), we provide a description of the migration process in Peninsular Malaysia. The data reveal that three quarters of the moves are other than the standard rural-urban move, and that return migration is widespread, particularly for rural workers. Using cross tabulations, we find that geography and ethnicity variables and the level of education are important determinants of migration patterns. We also find the presence of life cycle effects with a common starting age of between eighteen and twenty.
To study the determinants of lifetime migration in a multivariate context, we estimate double hurdle event count models using MFLS-2. Double hurdle models distinguish between factors which affect the probability of migration from factors which affect the number of migrations chosen over a lifetime. Specification testing and in-sample forecasting confirm that double hurdle models perform significantly better than the standard Poisson and Negative Binomial models.
The availability of information about different facets of the individual's background allows us to control for factors overlooked in most studies. Specifically, we are able to disentangle the effects of parent's and own education and we find that both effects induce both higher participation rates and a greater number of moves. Studies which do not control for parent's education levels would then be likely to overestimate the effect of own education. We also find support for the effects of location specific capital, particularly in rural areas.