| Abstract: |
Existing studies on the impact of migration on income inequality at sending
communities suffer from severe methodology defects and data limitations. This
paper analyzes the impact of rural-to-urban migration on inequality using a
newly constructed panel dataset for around 100 villages over a ten-year period
from 1997 to 2006 in China. To our best knowledge, this is the first paper
that examines the dynamic aspects of migration and income inequality employing
a dynamic panel data analysis. Unlike earlier studies focusing exclusively on
remittances, our data include the total labor earnings of migrants in
destination areas. Furthermore, we look at the gender dimension of the impact
of migration on wage inequality within the sending communities. Since income
inequality is time-persisting, we use a system GMM framework to control for
the lagged income inequality in estimating the effect of emigration on income
inequality in the sending villages. At the same time, contemporary emigration
is validly instrumented in the GMM framework because of the unobserved
time-varying community shock that correlates with emigration and income
inequality, as well as with the potential reverse causality from income
inequality to emigration. We found a Kuznets (inverse U-shaped) pattern
between migration and income inequality in the sending communities.
Specifically, contemporary emigration increases income inequality, while
lagged emigration has strong income inequality-reducing effect in the sending
villages. A 50-percent increase in the lagged emigration rate translates into
one-sixth to one-seventh standard deviation reduction in inequality.
Contemporary emigration has slightly smaller effects in raising the income
inequality within villages. These effects are robust to the different
specifications and different measures of inequality. More interestingly, the
estimated relationship between emigration and the gender wage gap also has an
inverse U-shaped pattern. Emigration tends to increase the gender wage gap
initially, and then tends to decrease it in the sending villages. |