Background

Lecture notes for EC484 Econometrics Analysis Lecture 10


Under the two assumptions, we could derive the new expression of .

We want and in the model. recall that is the time-invariant variable.

衡量的是不可观测的个体异质性 与可观测的时间不变特征 之间的相关程度

用刚才的工资例子: 是性别, 是不可观测的”能力”。如果男性和女性的平均未观测能力不同(比如因为历史性的教育资源分配不均),那么 就是相关的,

同理 的角色: 是个体 的时间均值(比如平均工时),如果工作更多的人平均能力也更高,那 相关,

本质上, 这个方程就是在说: 我不知道 是什么,但我可以用可观测变量对它做一个线性投影(linear projection), 就是投影系数。投影之后的残差 按构造与 不相关,RE 方法就可以安全使用了。

The good IV for is

4.1

Recall that , thus the FE1 method has Endogeneity Issue. The validity of the Instrumental Variable depends on which (period) you are talking about

7 Can PD LDV models with high dimensional integrals be rendered feasible with classical methods.

Reference