2sls, probably simple question

For technical questions regarding estimation of single equations, systems, VARs, Factor analysis and State Space Models in EViews. General econometric questions and advice should go in the Econometric Discussions forum.

Moderators: EViews Gareth, EViews Moderator

Jaap.smiths
Posts: 3
Joined: Thu Jun 24, 2010 3:20 am

2sls, probably simple question

Postby Jaap.smiths » Thu Jun 24, 2010 4:46 am

Hello all,

I cannot figure out how I can make the 2SLS function in Eviews 6 working...
I have already gone through the manual, but I get lost when it is talking about instruments...

For instance a small example what I want to do:
Y1 = c + X1 + X2 + residuals
X2 = c + Y1 + X3 + residuals

In other words, the original regression Y1 contains variable X2. But variable X2 is also explainable by Y1 and X3.

This is what I understand I have to do.

1) Open eviews. Click Quick/Estimate equation
2) As method I choose TSLS.
3) Now it gets confusing to me.
I believe I have to put my original regression under 'equation specification'.
So here I put: Y1 c X1 X2
4) I have no idea what I have to write into 'instrument list'
X2 Y1 X3 is not working.

I read that 2sls is done very often. So I am hoping that someone is willing to help me with this problem!

Many thanks in advance,

Jaap Smiths

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

Re: 2sls, probably simple question

Postby startz » Thu Jun 24, 2010 6:33 am

Your exogenous variables go in the instrument list. Given you description, probably just C X1 X3.

Jaap.smiths
Posts: 3
Joined: Thu Jun 24, 2010 3:20 am

Re: 2sls, probably simple question

Postby Jaap.smiths » Thu Jun 24, 2010 7:30 am

I cannot test at this moment, but many thanks for your answer already!
How does Eviews understand that c Y1 X3 is a function of X2?

And what if X1 is also a function of Y1?
So we have:

Y1 = c + X1 + X2
X2 = c + Y1 + X3
X1 = c + Y1 + X4

How is this possible to enter into the instrument field?

X2 x Y1 X3 X1 c Y1 X4 will probably not work.

Help is appreciated :)

lsym
Posts: 11
Joined: Thu Mar 04, 2010 3:11 am

Re: 2sls, probably simple question

Postby lsym » Fri Jun 25, 2010 2:53 am

In general the two steps of the 2SLS are: 1) The estimation of the reduced form equations to obtain the fitted values of the left hand variables and 2) the estimation of the structural equations replacing the right hand endogenous variables by their fitted values from step 1. In EViews this is done as follows: So first you need to obtain the reduced form of your system, for instance for the structural form:
Y = a1 + a2*X + a3*Z + e
Y = b1 + b2*X + b3*S + u

the corresponding reduced form is:
X = p11 + p12*Z + p13*S + e1
Y = p21 + p22*Z + p23*S + e2

To estimate this system in eviews, for the first equation you fill in the "Equation specification" dialog box: x c Z S
and in the "instruments list" all exogenous variables: c Z S

This methodology holds for any econometric system, provided it is just identified or overidentified.

lsym
Posts: 11
Joined: Thu Mar 04, 2010 3:11 am

Re: 2sls, probably simple question

Postby lsym » Fri Jun 25, 2010 3:22 am

Just a supplementary note to my previous answer: In order to determine whether the equation is just identified or over-identified and thus it can be estimated with 2SLS we have:
If G the number of structural equations then:
a) if the "excluded variables" are G-1 (excluded variables are all variables endogenous and exogenous not included in the equation) then the equation is just-identified
b) if the excluded variables are more than G-1 the equation is over-identified

In your 2nd system:
Y1 = c1 + a1*X1 +a2* X2 + e
X2 = c2 + b1*Y1 + b2*X3 + u
X1 = c3 + g1*Y1 + g2*X4 + a

the reduced for equations will be:
Y1 = p10 + p11*X3 +p12* X4 + e1
X2 = p20 + p21*X3 + p22*X4 + e2
X1 = p30 + p31*X3 + p32*X4 + e3

and for the 2SLS estimation for the 2nd equation for example, you fill: X2 c Y1 X3 in the "equation specification" box, and: c X3 X4 in the "instruments list"

Jaap.smiths
Posts: 3
Joined: Thu Jun 24, 2010 3:20 am

Re: 2sls, probably simple question

Postby Jaap.smiths » Fri Jun 25, 2010 5:20 am

I am really thankful for your input! It is much appreciated!

But I still do not get it. I do not think it is your fault; it is just my lack of knowledge...
Maybe if you can help me with this particular problem I have a better picture and perhaps get the idea better.

Suppose, I have the returns of a company and I want to explain those returns by size of the company, age of the company, amount traded and market volatility.

So, I have:
-Returns
-Size
-Age
-Volume
-Risk

I want to explain the following regression:
Returns = c + a1*Size + a2*Age + a3*Volume + a4*Risk + residuals

So far so good, right?
Now assume that I also know that risk and volume can also be explained by return and another variable.

So I have:
Risk = c + a6*Return + a7*Volume + residuals
Volume = c + a8*Return + a9*Risk + residuals

Now, can you explain me how I can program this in 2SLS in eviews with the same variable symbols?
What do I need to fill in 'equation' and 'instrument' box?

I understand that you are thinking 'n00b' or whatever, but I am really trying to get the hang of it.

Thanks!

Jaap Smiths


Return to “Estimation”

Who is online

Users browsing this forum: No registered users and 25 guests