Hovhannisyan, Shoghik. “Emigration by Educational Attainment and Growth”, Boston College, 2014. http://hdl.handle.net/2345/bc-ir:103742.
This thesis includes two essays that analyze growth implications of emigration and immigration. The first chapter studies the impact of emigrants with different education levels on their home countries' GDP per worker and its factors obtained by a production function decomposition. It uses migration data from 195 countries of origin to 30 major destination OECD countries in 1990 and 2000 and applies an instrumental variable approach to correct for endogeneity bias in estimating this impact. Pull factors of migration such as demand for emigrants' labor in destination countries and migrants' networks serve as a basis for instrument construction. Estimation results indicate that growth in emigration rates increases growth in GDP per worker in low and lower-middle income countries for all education groups of emigrants, primarily driven by improvements in total factor productivity (TFP). In contrast, there is no robust significant impact of emigration on other components of GDP. The second chapter studies the impact of immigrant labor on GDP per worker in the U.S. and its components obtained by a production function decomposition, including total factor productivity (TFP), the capital-output ratio, average hours worked, and skill intensity, defined as a productivity-weighted Constant Elasticity of Substitution (CES) function of high-skill and low-skill workers. It uses industry-level data over the period of 1960-2005 and applies two-step Difference Generalized Method of Moments (GMM) with instruments constructed using past distributions of immigrants across industries. The estimation results show that GDP per worker in an industry increases by about 2.24-2.63 percent in response to a one percent increase in the share of immigrants in total employment of the industry. These results are primarily driven by TFP growth with a magnitude of 2.08-2.21 and average hours worked: 0.23-0.29. However, these results are not robust to inclusion of the lagged dependent variables.