Having problem with estimating ces production function

[Kanan]

Hi,

I need to estimate CES production function like log(Y)=log(aK^-p+(1-a)*L^-p)/(-1/p)+yt+e. I am trying to estimate but cant get a result. I did Cobb-Douglas but have trouble with CES production function.

Great thanks in advance

Kanan

[startz]

Nobody here can guess what you did. Post as many details as possible about what you did and what went wrong and someone may be able to help.

[Kanan]

Nobody here can guess what you did. Post as many details as possible about what you did and what went wrong and someone may be able to help.

Thanks for reply,

I read from one source that i can estimate CES production function on eviews like this

Y = C(1)*(C(2)*K^(-C(3)) + (1− C(2)) *L^(-C(3))) ^(−1/C(3))

but it didn't work, I get NA in my st errors and probabilties. So i need to know how can I correctly write this equation that would work. I need to estimate the equation that i first posted and there is a trend also considered i guess.

thanks again,

Kanan

[startz]

Try changing the starting values in the c() vector.

[Kanan]

Try changing the starting values in the c() vector.

Sorry, I didn't fully get what you mean. I changed the order of c and started from 2, in addition I changed the starting value of one of the variables but still didn't get any result. Is the above writing ok? If so the coefficients should be similar to Cobb-Douglas function, but i get NA in probabilities and R2 is negative large number greater than 100.

Maybe I didn't understand you correctly. Would be glad for any help.

Thanks

[startz]

In the workfile window you'll see an object named "c". That's where EViews stores coefficients. Go in an change the values to ones you think might be reasonable before you start the estimate. (The order of the coefficients doesn't matter.) Sometimes this helps in nonlinear estimates. If that doesn't work, you might want to post your workfile, including the equation you've estimated.

[Kanan]

Thanks for your help,

I tried but still R2 is not getting a value that should be. I have uploaded the work file. The equation need to estimated is like this

log(GDP)=log(a*Capital^-p+(1-a)*Labour^-p)/(-1/p)+yt+e

It is a log written version of ces production function. Capital is with t-1. I will put the picture of the equation as well.


Thank you again

[Kanan]

This is the picture from a paper.

Thanks

[startz]

I don't think you've done anything wrong. The solution is highly sensitive to the starting values. For example, if you add in the trend term and start at c=1, you get a perfectly nice estimate...it just isn't the same estimate as in the picture.Getting a reliable nonlinear estimate with only 24 observations is tough. Sometimes there's nothing to do but stat estimates at lots of different values and see which one is best. Alternatively, one can plot the likelihood..but EViews isn't the right software for that.