CES function estimation using NLS

[londonphd]

Hi,

I am trying to estimate the a set of two parameters of a CES function:

Code: Select all

Y = (((AL*L)^C(1)+(AK*K)^C(1))^(C(2)/C(1))+(AE*E)^C(2))^(1/C(2))

The iteration stops abruptly due to division by 0. Can I limit the range of parameters to non-zero values?
I used the

Code: Select all

param

function to set starting values, but it has not helped.

thank you.

[EViews Gareth]

take logs?

[startz]

You can also replace the coefficient c(i) with exp(c(i)). That guarantees that exp(c(i)) will be positive.

[tvonbrasch]

this article may also be of interest
t

[londonphd]

take logs?

Many thanks.
How will I type in the log of the CES function (((al*l)^c(1)+(ak*k)^c(1))^(c(2)/c(1))+(ae*e)^c(2)) ?
Thank you.

[Elderfield.A]

You can restrict the parameters as you like, see for example this post:

http://forums.eviews.com/viewtopic.php?t=12122


The authors of the paper provided by @tvonbrasch also have an R package called micEconCES which offers some insights into how one estimates the production function directly using NLS.