In this dissertation, we develop four essays to analyse several aspects of migration for Pakistan on the basis of household level panel and cross-sectional data from 1986 to 2007. The aspects analysed are decision to migrate domestically and internationally, financing of migration, decision to send remittances and spill-over effects on remittances. These essays are presented in Chapters four through seven. Other chapters include a general introduction, literature review and concluding remarks and policy discussions.
Chapter four (essay one) studies the determinants of the two locational migrations (internal-and-international migration) using discrete choice models. The study begins by investigating migration as a whole and then looks at the more disaggregate choice. On the snapshot of migration as a whole we detect an intriguing size-composition effect on the household’s probability of migration: the larger the household, the greater the probability of migration. On the other hand, the more dependants a household has (in terms of the number of children and young females), the less likely it is to have migrant member. This relation is by and large true for both types of migration – internal or international. We also find that the probability of migration is inversely related to the pre-migration initial (as observed at the beginning of the study period) landholding of the household. For a one-acre decrease in landholding will results in 11% increase in international migration relative to never migration, and comparatively it shows a 1.3% increase in internal migration. So it is most likely that households have depleted their landholding to raise finances for their migration.
Using the same dataset (of chapter four) in chapter five (essay two), we study the determinants of receiving remittances by constructing a Box-Cox double hurdle model for households. The first
iii
hurdle deals with the decision to receive remittance, and conditional on participation, the hurdle deals with the amount or level of remittances. The first stage regression does reveal household characteristics such as household head age, number of children aged less than 11 years, and the district level fixed effects to be important for remittances, be it for internal or international migration. From these factors, it may be concluded that household level characteristics and regional factors are the important determinants for the probability of remittances. The second hurdle deals with the determinants of remittance amounts, conditional on being a remitter. The household head education and age appear to affect the level of remittances, unlike the participation results. Comparing the impact across the participation and the level of remittances received by the household, we find that both members travelling within and outside of the country are the important determinant for both hurdles. We find that there exists opposing effect of probability to receive remittances and the level of remittances at the district level.
Chapter six (essay three) studies the determinants of internal- and –international remittances by introducing the random- and –fixed effects by employing a multilevel econometric methodology to study the spread of remittances at different levels of spatial aggregation. Few studies use multilevel analyses on remittance data, but to the best of our knowledge, there is none for in the context of a developing country such as Pakistan. We use data from Pakistan Household Integrated Economic Survey of 2001-02, employing household-,village-, district-, province- and regional-level variables to understand how heterogeneities at these different levels impact on the probability of receiving (or sending) remittances, relative to other households in the same village or district. Our findings suggest that the determinants of internal remittances are different from those of international remittances. We find greater variation in the odds of receiving remittances (both internal and international) among households from same villages within the same district than those located in different villages within same district. Also with regional fixed effects, the correlation is refined and thus it becomes smaller.
iv
In chapter seven (essay four), we empirically study the possible multiplier effect (spill over) triggered by remittances by using the Pakistan Household Integrated Economic Survey of 2007-08. While the existing studies provide a compelling empirical evidence of remittance income is more likely to be saved and invested in land, housing, and human capital, but it is unclear to what extent it contribute to the origin community. The existing literature is unable to answer the question whether remittances causes multiplier or spill-over effect. If they do exists, then to what extent? Motivated by the concern that the remittance can generate multiplier effects in origin communities. We investigate the role of household interdependencies of the remittances by exploiting a novel method of identification based on the comparison of the variance of household behaviour at the different level of aggregation within and between a different districts of Pakistan (Graham, 2008). This method allows for identification of two problems that arise due to self-selection and unobservable heterogeneity. We quantify the social multiplier of remittances to be 1.12, suggesting sizable spill-over between eighty one districts in Pakistan. Our result suggests that social multiplier in terms of remittances has contributed to the development of rural household of Pakistan.