about the 2sls estimates

[tony]

I estimate a model using 2sls in EViews 7,
but the outputs are different from my manual two stages OLS, namely
first, regress y2 on c x1 z1 z2, where y2 are endogenous, x1 are exogenous, z1,z2 are IVs, and predict y2, say y2_p
second, regress y1 on c x1 y2_p.
I konw the standard error is wrong, but the estimates should be the same as those of 2sls in EViews.
So, could you please help me to interpret this?
Thank you!

[startz]

Perhaps you should post each of your regressions and the 2sls output.

[EViews Gareth]

Here's some code I wrote a long time ago that replicates various 2SLS models.

Code: Select all

create u 100 series x=nrnd series y=nrnd series z=nrnd series w=nrnd 'simple 2SLS equation eq2sls.tsls y x @ z equation eq1.ls x z c eq1.fit xf equation eq2.ls y xf 'simple 2SLS with an AR(1) term equation areq2sls.tsls y x ar(1) @ z equation areq1.ls x c x(-1) y(-1) z areq1.fit xf equation areq2.ls y=c(1)*(xf-c(2)*x(-1)) + c(2)*y(-1) 'Simple 2SLS with a difference term equation difeq2sls.tsls d(y) x @ d(z) equation difeq1.ls x c d(z) difeq1.fit xf equation difeq2.ls d(y) xf 'Simple 2SLS with even more difference terms equation compeq2sls.tsls d(y) d(x) w @ d(z) c equation compeq1a.ls d(x) d(z) c compeq1a.fit(d) xf equation compeq1b.ls w d(z) c compeq1b.fit wf equation compeq2.ls d(y) xf wf