Simultaneity bias OLS

For econometric discussions not necessarily related to EViews.

Moderators: EViews Gareth, EViews Moderator

Aktar
Posts: 74
Joined: Thu May 21, 2009 5:08 am

Simultaneity bias OLS

Postby Aktar » Mon Jan 31, 2011 6:57 am

Hello,

I estimate this model :

dlog(EC_thb/aud) = BETA1 dlog(EC_USD/aud) + BETA2 dlog(EC_ACU/aud)

where EC is the exchange rate. I explain the movement of thain bath (against a numeraire - asutralian dollar-) by movement of USD (against the same numerair) and an asian currency unit (composed by different asian currencies). Normally the beta give me the weight or the importance of usd and amu in the evolution of the thai bath. But thai bath and acu are in some extense determined by the same variable (us dollar)... so the coefficent BETA 2 is very strong.

So One of my explanatory variable is apparently determined simultaneously with the dependent variable of my model. In this case normally, the former is generally correlated to the error term which leads to inconsistent and biased OLS estimates.

How i can check correlated error term with the explanatory variable in eviews (control this simultaneity bias) ? and how i can remove this effect ?

thank you very much.

Best regards

startz
Non-normality and collinearity are NOT problems!
Posts: 3775
Joined: Wed Sep 17, 2008 2:25 pm

Re: Simultaneity bias OLS

Postby startz » Mon Jan 31, 2011 7:57 am

In general, you need to use TSLS. In this case, perhaps you could add dlog(AUD) to both sides?

Aktar
Posts: 74
Joined: Thu May 21, 2009 5:08 am

Re: Simultaneity bias OLS

Postby Aktar » Mon Jan 31, 2011 8:20 am

Thank you very much,

What do you mean by "In this case, perhaps you could add dlog(AUD) to both sides?"

bests

startz
Non-normality and collinearity are NOT problems!
Posts: 3775
Joined: Wed Sep 17, 2008 2:25 pm

Re: Simultaneity bias OLS

Postby startz » Mon Jan 31, 2011 8:32 am

If you add dlog(AUD) to both sides of the equation, you still have a valid equation. Doing this removes AUD from the left hand side. I was thiking that might help with your endogeneity problem, but on reflection I'm not sure it makes any difference.

Aktar
Posts: 74
Joined: Thu May 21, 2009 5:08 am

Re: Simultaneity bias OLS

Postby Aktar » Mon Jan 31, 2011 8:41 am

ok thank you,

what do you mean by dlog(aud) ?

"aud" mean australian dollar, which is express against an other currency (it is the base currency of different exchange rate in the equation)

Thank you very much for your help, i must find good instruments for my estimation.

startz
Non-normality and collinearity are NOT problems!
Posts: 3775
Joined: Wed Sep 17, 2008 2:25 pm

Re: Simultaneity bias OLS

Postby startz » Mon Jan 31, 2011 9:04 am

dlog(EC_thb/aud)+dlog(aud)=dlog(EC_thb).

But yes, finding good instruments is the way to go.

Aktar
Posts: 74
Joined: Thu May 21, 2009 5:08 am

Re: Simultaneity bias OLS

Postby Aktar » Mon Jan 31, 2011 9:05 am

How i can check for correlation between one of my explanatory variable ad error term ? is there a specific test ? or i only keep residual series (ordinary or standarized ?) and have look to the correlation of the group variables ?

when i look for correlation matrix between the series of my ordinary residual and amu/aus, there is no correlation... so what does it mean ?? However when i look for correlation between my dependent variable (thb/aus) and my explanatories variables (usd/aus) and (acu/aus) they appeared to be very correlated. Moreover, two of my explanatories variables (usd/aus and acu/aus) are very correlated too. It is normal because acu is a basket of asian currencies which are correlated to the dollar also.

But when i check residual of my estimation there is no correlation between residual and all the variables. Maybe i should make a test of correlation ?

startz
Non-normality and collinearity are NOT problems!
Posts: 3775
Joined: Wed Sep 17, 2008 2:25 pm

Re: Simultaneity bias OLS

Postby startz » Mon Jan 31, 2011 9:40 am

Least squares residuals are always correlated with the dependent variable. They are never correlated with any of the independent variables. If you want to test for correlation between the errors and the independent variables you need instruments.


Return to “Econometric Discussions”

Who is online

Users browsing this forum: No registered users and 26 guests