Does increasing parents' schooling raise - Ivie
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Now assume the errors distributions depend on the X 0 s (e.g. heteroskedasticity).<br />
The coe¢ cients in (3A) become:<br />
j = cov(vj ujX)<br />
var(v j jX) = Aj (X)<br />
which implies that the impact of v j i<br />
on u i depends on the value of X i : Under the<br />
conditional correlation assumption KV show:<br />
p<br />
j = cov(vi ujX) V<br />
var(v j jX) = ar(ui jX i )<br />
j q<br />
V ar(v j i jX i)<br />
which given our assumptions in the text gives:<br />
u i = M H ui<br />
v M<br />
Hvi<br />
M i<br />
+ F H ui<br />
v F<br />
Hvi<br />
F i<br />
+ e 1i<br />
Estimation is now feasible as the matrix M 1 = [X; S; S M ; S F ; Hu<br />
H M v<br />
bv M ; Hu bv F ] is of full<br />
Hv<br />
F<br />
rank due to the non linearity induced by the multiplicative role of the X 0 s: KV show<br />
that the parameters of the model are identi…ed even without parametric assumptions<br />
regarding the form of heteroskedasticity.<br />
While KV (2006) provide a semiparametric estimation procedure for (4A), retaining<br />
the semiparametric aspect in practice is associated with demanding computational<br />
requirements. Thus we employ the following parametric version:<br />
i) Regress S M and S F on X to get bv M and bv F :<br />
ii) Use assumption (8) and estimate j 2 and 2j through non linear least squares using<br />
ln(bv j2 ) as the dependent variable. With these estimates compute b H j vi = qexp(b j 2(Z j i b 2 ):<br />
iii) To estimate the primary equation parameters we can proceed in two ways.<br />
24