Result interpretation - two stage least squares in Eviews

[hanidv]

Hi all,
I am using two stage least square to test the endogenous relationship between 2 variables Y and M
Y = a1 + b1*X + b2*Z + b3*M
M = a2 + b4*Y + b5*Z
Did OLS regression, I got b3 is negative and b4 is negative. However, when I run 2SLS, I got b3 is positive and b4 is still negative.
Anyone know how to interpret this result.

[startz]

Hi all,
I am using two stage least square to test the endogenous relationship between 2 variables Y and M
Y = a1 + b1*X + b2*Z + b3*M
M = a2 + b4*Y + b5*Z
Did OLS regression, I got b3 is negative and b4 is negative. However, when I run 2SLS, I got b3 is positive and b4 is still negative.
Anyone know how to interpret this result.

How did you do 2SLS? You don't appear to have enough instruments.

[hanidv]

Hi, thank for reply.
here is my full models:
(1) LEV = a1 + b1*GR + b2*SIZ + b3*PROF + b4*TANG + b5*LIQ + b6*NDTS
(2) LIQ = a2 + c1*GR + c2*SIZ + c3*CFR + c4*LEV
I do OLS with b5 and c4 are negative
With 2SLS, The instruments are
For (1): C GR SIZ PROF TANG NDTS CFR
For (2): C CFR GR SIZ PROF TANG NDTS
And I got b5 negative but c4 is positive.

[startz]

Oh, you're obviously fine on the instrument count then.

The standard interpretation would be that running 2sls corrects for bias in OLS, and the 2sls results would be preferred.

[hanidv]

Oh, you're obviously fine on the instrument count then.

The standard interpretation would be that running 2sls corrects for bias in OLS, and the 2sls results would be preferred.

LEV and LIQ are negatively related in both OLS and also in the first 2SLS, however, for the second 2SLS model, it show a positive relationship. It is contrary with the first one.
Do u think is there any problem inside the model or which model should I believe in?