In the context of Strict Exogeneity on the classical regression model $$y_i = \beta x_i+ \varepsilon_j$$ we have: $$E(\varepsilon_i | x_i , x_j )=0,\quad \forall i,j$$ Under the linearity assumption, we have that $\varepsilon_i$ is a function of $(y_i, x_i)$

But, if $(y_i,x_i)_{i=1}^n$ is an i.i.d. I want to show that:

  1. $(y_i, x_i)$ independent of $(y_j, x_j)$ for $i \neq j$ implies $(\varepsilon_i, x_i)$ is independent of $x_j$ for $j \neq i$;
  2. $E(\varepsilon_i | x_i,x_j)=E(\varepsilon_i | x_i)$

I don't know how to prove the first item. As for the second one, I'm trying to use the following property, $$E(\varepsilon_i | x_i) = E( E( \varepsilon_i |x_i,x_j) |x_i) $$ but I don't know how to complete it.

Any help for my two items above?


1 Answer 1


First if $X$ and $Y$ are independent random vectors, then $f(X)$, any function of $X$, is independent of $g(Y)$, any other function of $Y$. So if $(y_i,x_i)$ is independent of $(y_j,x_j)$, then taking $f(x,y)= (y- \beta x,x)$ and $g(x,y)=x$ we have that $$(\varepsilon_i, x_i) \hbox{ is independent of } x_j.$$ This shows the first item.

For the second item, if $(x,y)$ is independent of $z$, then $$E(x|y,z)=E(x|y).$$ Thus, using the first item, we conclude that $$E(\varepsilon_i|x_i,x_j)=E(\varepsilon_i|x_i)$$

Quod Erat Demonstrandum!


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.