Instrumental Variables

[ripperpool]

Anybody has any idea about them? Can I refer to any notes or online material regarding these as I am still a newbie at this

[startz]

The wikipedia entry is not bad.

[suki1121]

it is used in the case when the initial assumptions about x and error term do not hold.

e.g. if x and error are correlated, we need an IV that is highly correlated with x but uncorrelated with error, say z.

then two stage least square applies, 1st, regress x on z to get x hat; 2nd regree y on x hat.

i hope my ans is clear.

[startz]

we need an IV that is highly correlated with x but uncorrelated with error, say z.

then two stage least square applies, 1st, regress x on z to get x hat; 2nd regree y on x hat.

although you should use the EViews command TSLS. Don't run two separate regressions.