EViews.com

Hello, I´m using Eviews 6 and I need to replicate an estimation for a growth equation (from Barro 1999, "Inequality and Growth in a panel of countries")
I have data for 138 countries with info for 4 periods 1960,1970,1980 and 1990 and I ned to estimate a growth regression in 3SLS, meaning 4 equations in the system that have the same variables, only differing in the period (year). At the end I need to obtain one coefficient for each independant variable and four intercepts as well as four R2. This is the system I have untill now (including instruments)

tcrec60_55_64 = C(1) + C(2)*lnpib_55 + C(3)*lifee60_1+ C(4)*grenrrh60 + C(5)*lnfert60 + C(6)*invsh51+ C(7)*gini_60 @ lnpib_55(-5) lifee60_1 GrEnrRH60 lnfert60(-5) invsh51(-5) gini_60

tcrec70_65_74 = C(8) + C(2)*lnpib_65 + C(3)*lifee70_1+ C(4)*grenrrh65 + C(5)*lnfert70 + C(6)*avinvr70+ C(7)*gini_70 @ lnpib_65(-5) lifee65_1 GrEnrRH65 lnfert70(-5) avinvr70(-5) gini_70

tcrec80_75_84 = C(9) + C(2)*lnpib_75 + C(3)*lifee80_1+ C(4)*grenrrh75 + C(5)*lnfert80 + C(6)* avinvr80+ C(7)*gini_80 @ lnpib_75(-5) lifee75_1 GrEnrRH75 lnfert80(-5) avinvr80(-5) gini_80

tcrec90_85_94 = C(10) + C(2)*lnpib_85 + C(3)*lifee85_1+ C(4)*grenrrh85 + C(5)*lnfert90 + C(6)*invsh56+ C(7)*gini_90 @ lnpib_85(-5) lifee85_1 GrEnrRH85 lnfert90(-5) invsh56(-5) gini_90

I figured out that by assigning the same c( )to each ind. var. (i.e. C(2)*lnpib_55; C(2)*lnpib_65; C(2)*lnpib_75; C(2)*lnpib_85) on each of the equations is the correct way for obtaining only one coefficient for the whole system. Also, assigning different C( ) for each equation intercept will give me 4 different intercepts (as in the model I´m trying to replicate).
The problem is that after estimating I get a very bad result,,, none of the coeff. are significant, not even the intercepts and the R2 have negative values,,,

What am I doing wrong???
Please Help!!!

This is what I get after estimating,,, as you can see there is something wrong. I am shure the D.B. is correct so I think is the estimation... please help!


System: SYS04
Estimation Method: Iterative Three-Stage Least Squares
Date: 04/14/10 Time: 14:37
Sample: 9 137
Included observations: 80
Total system (unbalanced) observations 183
Simultaneous weighting matrix & coefficient iteration
Convergence achieved after: 6 weight matrices, 7 total coef iterations


Coefficient Std. Error t-Statistic Prob.


C(20) 1.112998 16.50432 0.067437 0.9463
C(2) 0.327021 1.741322 0.187800 0.8513
C(3) 338.7375 223.6979 1.514263 0.1318
C(4) -1.369356 1.870139 -0.732221 0.4650
C(5) -3.368659 3.680732 -0.915214 0.3614
C(6) -5.254112 7.898300 -0.665221 0.5068
C(10) 0.010313 0.047015 0.219357 0.8266
C(21) 1.454950 16.47005 0.088339 0.9297
C(22) -1.268664 16.26656 -0.077992 0.9379
C(23) -1.388824 16.04069 -0.086581 0.9311


Determinant residual covariance 1041.461



Equation: TCREC60_55_64 = C(20) + C(2)*LNPIB_55 + C(3)*LIFEE60_1+
C(4)*GRENRRH60 + C(5)*LNFERT60 + C(6)*INVSH51+ C(10)
*GINI_60
Instruments: LNPIB_55(-5) LIFEE60_1 GRENRRH60 LNFERT60(-5)
INVSH51(-5) GINI_60 C
Observations: 19
R-squared -0.361114 Mean dependent var 3.282843
Adjusted R-squared -1.041671 S.D. dependent var 2.268528
S.E. of regression 3.241433 Sum squared resid 126.0827
Durbin-Watson stat 0.309324

Equation: TCREC70_65_74 = C(21) + C(2)*LNPIB_65 + C(3)*LIFEE70_1+
C(4)*GRENRRH65 + C(5)*LNFERT70 + C(6)*AVINVR70+ C(10)
*GINI_70
Instruments: LNPIB_65(-5) LIFEE65_1 GRENRRH65 LNFERT70(-5)
AVINVR70(-5) GINI_70 C
Observations: 42
R-squared 0.101253 Mean dependent var 3.683439
Adjusted R-squared -0.052818 S.D. dependent var 2.972238
S.E. of regression 3.049721 Sum squared resid 325.5280
Durbin-Watson stat 0.404566

Equation: TCREC80_75_84 = C(22) + C(2)*LNPIB_75 + C(3)*LIFEE80_1+
C(4)*GRENRRH75 + C(5)*LNFERT80 + C(6)* AVINVR80+ C(10)
*GINI_80
Instruments: LNPIB_75(-5) LIFEE75_1 GRENRRH75 LNFERT80(-5)
AVINVR80(-5) GINI_80 C
Observations: 50
R-squared -0.110948 Mean dependent var 1.181002
Adjusted R-squared -0.265964 S.D. dependent var 2.241414
S.E. of regression 2.521929 Sum squared resid 273.4854
Durbin-Watson stat 0.271499

Equation: TCREC90_85_94 = C(23) + C(2)*LNPIB_85 + C(3)*LIFEE85_1+
C(4)*GRENRRH85 + C(5)*LNFERT90 + C(6)*INVSH56+ C(10)
*GINI_90
Instruments: LNPIB_85(-5) LIFEE85_1 GRENRRH85 LNFERT90(-5)
INVSH56(-5) GINI_90 C
Observations: 72
R-squared 0.107205 Mean dependent var 1.496797
Adjusted R-squared 0.024793 S.D. dependent var 2.215215
S.E. of regression 2.187581 Sum squared resid 311.0582
Durbin-Watson stat 0.267351

It isn't clear you are doing anything wrong in EViews. Perhaps your model is just bad.

But I am replicating exactly the same model as the author I mentioned (barro 2008). I would expect at least similar results. Something I don´t quite understand is that he explains this as a panel estimation, but I think the database cannot be arranged as panel i.e. because this would not allow me to specify a system of 4 equations.
Do you think the problem may come from how I am structuring the d.b.? or in other words, which is the right way to structure it,, as (A) or (B)

(A)
NumID DateID Country Year Var 1 Var 2 Var 3 ......
1 1 Afghanistan 1960 x x x
1 2 Afghanistan 1970 x x x
1 3 Afghanistan 1980
1 4 Afghanistan 1990
1 5 Afghanistan 2000
2 1 Algeria 1960
2 2 Algeria 1970
2 3 Algeria 1980
2 4 Algeria 1990
2 5 Algeria 2000
3 1 Angola 1960
3 2 Angola 1970
3 3 Angola 1980


(B)

No. Pais Var 1 Var 2 Var 3
1 Algeria x x x
2 Angola x x x
67 Argentina
131 Australia


Thanks!