Extensions and robustness tests
Migration into one labor market can have spillover effects on adjacent labor markets
through, for instance, trade or migration of native workers (Borjas 2003). Moreover, even
in the absence of spillovers, a mismatch between the boundaries of districts and the
boundaries of labor markets can induce spatial correlations between districts (Anselin and
Bera 1998: 239). We investigate spatial dependency by estimating a model with spatial
lags,
(5) d d d d d d w = a + g m + d b + bX + rW¢w + e
where w is a vector containing all values of wd, Wd is a vector of weights, and 㰐 is the
spatial correlation coefficient. Wd is constructed by setting the i-th element equal to 1 if
district i shares a border with district d and 0 otherwise, and then normalizing so that Wd
sums to 1. The spatial correlation coefficient 㰐 governs the rate at which correlations die
off with distance.
The spatial lags model is not designed for situations where explanatory variables are
endogenous (Kelejian and Prucha 1998: 101). We therefore replace migrant intensity
with distance to the Myanmar border. A positive coefficient on distance would imply a
negative relationship between migration and wages, though without providing
information about the magnitude of the relationship. The presence of w on the right hand
side means that Equation 5 cannot be estimated using ordinary least squares. Instead we
use instrument on spatial lags of Xd, which allows heteroskedasticity-robust standard
errors to be calculated (Anselin and Bera 1998: 258-60)10. We use as dependent variables
the wages of all private employees and the wages of low-skilled private employees.
An alternative way to incorporate spillovers is to use larger geographical units (Borjas
Extensions and robustness testsMigration into one labor market can have spillover effects on adjacent labor marketsthrough, for instance, trade or migration of native workers (Borjas 2003). Moreover, evenin the absence of spillovers, a mismatch between the boundaries of districts and theboundaries of labor markets can induce spatial correlations between districts (Anselin andBera 1998: 239). We investigate spatial dependency by estimating a model with spatiallags,(5) d d d d d d w = a + g m + d b + bX + rW¢w + ewhere w is a vector containing all values of wd, Wd is a vector of weights, and 㰐 is thespatial correlation coefficient. Wd is constructed by setting the i-th element equal to 1 ifdistrict i shares a border with district d and 0 otherwise, and then normalizing so that Wdsums to 1. The spatial correlation coefficient 㰐 governs the rate at which correlations dieoff with distance.The spatial lags model is not designed for situations where explanatory variables areendogenous (Kelejian and Prucha 1998: 101). We therefore replace migrant intensitywith distance to the Myanmar border. A positive coefficient on distance would imply anegative relationship between migration and wages, though without providinginformation about the magnitude of the relationship. The presence of w on the right handside means that Equation 5 cannot be estimated using ordinary least squares. Instead weuse instrument on spatial lags of Xd, which allows heteroskedasticity-robust standarderrors to be calculated (Anselin and Bera 1998: 258-60)10. We use as dependent variablesthe wages of all private employees and the wages of low-skilled private employees.An alternative way to incorporate spillovers is to use larger geographical units (Borjas
การแปล กรุณารอสักครู่..
