Dear All,
My name is Herbert. I would like to raise a question (or perhaps a confirmation) about my CGE modeling using GAMS (PATH Solver).
I am recently finished to build a CGE model code in GAMS. This model generates optimal solution (at 0 iteration / iterlim) at the initial state (preshock). I have checked everything, including the homogeneity of the model. So, no problem with the calibrations and equations.
However, I do would like to ask a confirmation about the shocked model results I had. In other words, after I ensure that the preshock reached optimal solution I then continue to do some shock experiments to the model. I saw a strange result in the statistics screen.
For examples, given a certain scaling I simulate some shocks. First, I simulate a 10% increase of household income tax rate (HAtaxrate) to my model with this command:
 A CHANGE IN TAX RATE FOR HOUSEHOLDS
HAtaxrate_SHOCK(‘BASE’) = HAtaxrate0;
HAtaxrate_SHOCK(‘SHOCKED’) = 1.1 * HAtaxrate0;
HAtaxrate_SHOCK(‘PCHANGE’) = 10;
When I execute this, the statistics screen showed as follows:
EXIT  OPTIMAL SOLUTION
However, I checked that the infesibility exists (by typing INFES in the search box of GAMS) of which this INFES is given to those eq that related with household tax as shown below:
eq22b(HH_AGR_L)…  0.0236240381204785*IH(HH_AGR_L) + TR(GOV,HH_AGR_L) =E= 0 ;
(LHS = 0.00443672856287591, INFES = 0.00443672856287591 ****)
Can somebody help me to clarify why is this INFES still comes even the optimal solution is found?
Also, I tried another shock, I simulate a 10% increase of government spending on goods (CGADJ) to the model with below command:
 E.G. 10% SHOCK TO GOV EXP CGADJ
CGADJ_SHOCK(‘BASE’) = CGADJ0;
CGADJ_SHOCK(‘SHOCKED’) = 1.1 * CGADJ0;
CGADJ_SHOCK(‘PCHANGE’) = 10;
With the same scaling, the statistics result is given as follows:
EXIT  OPTIMAL SOLUTION
Without any infesibility (no INFES)!
Finally, I again tried a different shock, a simultaneous 10% increase of government spending on goods (CGADJ) and 10% increase in world import price (P_MW) as follows:

E.G. 10% SHOCK TO GOV EXP CGADJ
CGADJ_SHOCK(‘BASE’) = CGADJ0;
CGADJ_SHOCK(‘SHOCKED’) = 1.1 * CGADJ0;
CGADJ_SHOCK(‘PCHANGE’) = 10; 
OR:

A CHANGE IN IMPORT PRICES, P_MW(i)
P_MW_SHOCK(i, ‘BASE’) = P_MW0(i);
P_MW_SHOCK(i, ‘SHOCKED’) = 1.1 * P_MW0(i);
P_MW_SHOCK(i, ‘PCHANGE’) = 1;
And now the statistic reports show:
EXIT  OTHER ERROR
With some infesibility in the equations (by typing INFES in the search box) of which this INFES is given to those equation that linked with the shock: government spending and import price. The infesibility is shown below:

(1.38644762049494)*EXR + IG  (0.104533516883162)*M(AGRI_C)

(0.0964900725404228)*M(CATLE_C)  (0.0741589758645914)*M(CHFCC_C)

(0.00459183110906082)*M(COMOIL_C)  (0.0790671886550816)*M(FISH_C) and so on…
(LHS = 0.126040692772266, INFES = 0.126040692772266 ****)
EG  (0.481406600172975)*P_Q(CHFCC_C)  (0.010373019131026)*P_QA(AGRI_A) 
(0.0116876059899586)*P_QA(AISCOM_A)  (1.3216473300235)*P_QA(CHFCC_A)

(0.999535973506986)*P_QA(ELEGD_A)  (0.0011421712953687)*P_QA(FISH_A) and so on
(LHS = 0.344277540843711, INFES = 0.344277540843711 ****)
It seems that any shocked simulations I had are always generating a similar pattern; and it looks like a random luck!! So, I noticed there are 3 general statistic reports to any shocks I made:
 EXIT  OPTIMAL SOLUTION without INFES (which is the best result in my opinion);
 EXIT  OPTIMAL SOLUTION with INFES
 EXIT  OTHER ERROR with INFES
Can someone give me a confirmation that (1) must always be reached for any shocked model?? Is it ok if I use the results of the shocks when (2 or 3) appears in the report? I have tried to use a scaling approach in order to help GAMS easier in finding the solution. However, it seems only tried to minimize the INFES of equation. I noticed that scaling is only generating a perfect statistic report (optimal solution and no INFES) in the case of a 10% increase of government spending. Even when I reduced it to 5%, the INFES appeared!
I am sure that my model and calibration is correct since the INFES in the shocked model always link to the chosen shock. I confirm that increasing (or reducing) the scaling value only enables to minimize the INFES but cannot eliminate them. Luckily, I only found a report of (1) in numerous kinds of scaling range!
Last year, I also have constructed a CGE model with GAMS solver. At that time, I was convinced that the model correct since I did not aware about the above cases. But when I rechecked this model for a comparison to the new model I built, the pattern of the reports are similar like above (1  3).
Can someone confirm that the above reports (any 1  3) are NO PROBLEM?? Perhaps, if someone has been doing (had) a CGE model that is (was) already well designed (with optimal solution) could simulate some shocks and try to check the INFES in the search box and could clarify that there are no INFES for all kinds of shocks you simulated it would be great. I suppose it takes only less than a minute!?
Thanks guys. I really appreciate you help.
Regards,
Herbert
\
To unsubscribe from this group and stop receiving emails from it, send an email to gamsworld+unsubscribe@googlegroups.com.
To post to this group, send email to gamsworld@googlegroups.com.
Visit this group at https://groups.google.com/group/gamsworld.
For more options, visit https://groups.google.com/d/optout.