How to Apply Panel ARDL Using EVIEWS

How to apply panel ARDL using EVIEWS. you can perform this test easily with the help of few steps following are eviews steps while in following section section you will n! theory a"out panel ar!l an! how to run panel ARDL using stata #$%&urther panel ARDL have two estimators %means we can run ARDL with two metho!s "ut EVIEWS only one estimator 'nown as ()* while stata haive "oth ()* an! )* *o to +uic' an! select estimate e+uation Select from !rop !own "utton,com"o "o- or from metho! ta" "elow e+uation specication ()*,ARDL ow in e+uation specication ta" write your e+uation li'e co/ f!i g!p here you can can see see rst rst i writ write e !epe !epen! n!en entt varia aria"l "le e whic which h is co/ co/ an! aft after this his all in!epen!ent varia"le an! !on0t inclu!e 121 means constant ow select -e! -e! or automatic automatic lags its all up to "ut if you choose automatic automatic lags you can assign !i3erent lag to !epen!ent an! in!epen!ent varia"les %li'e its up4 to you% mean i can assign assign ma-imum ma-imum 5 lags to !epen!ent !epen!ent varia"le% varia"le% while $ to in!epen!ent varia"le% eviews automatically choose optimal lags even you assign 5 %eviews can select / as optimal%"ut if choose -e! lags option then assign same ma-imum lags "oth to !epen!ent an! in!epen!ent varia"le. When you will 67 results will "e pro!uce ow if you want to see in!ivi!ual e3ect for !i3erent companies or country% go to view of you resulte! win!ow an! clic' cli c' on view an! select cross option an! 67.  8heory of panel ARDL (recon!itions regar!ing sationarity of panel ar!l We can run ar!l mo!el in three cases When all varia"les are stationary at level When all varia"les are stationary at rst !i3erence 6r when varia"les are stationary at level an! rst !i3erence means in mi-ture nature of sationarity 9 Suppose I have : varia"les suppose ; stationary at level an! ; at rst !i3erence so I can go now for ARDL mo!el Remem"er we can mo!el  8he main characteristic of ()* is that it allows short4run coe?cients% inclu!ing the intercepts% the spee! of a!@ustment to the long4run e+uili"rium values% an! error variances to "e heterogeneous country "y country% while the long4run slope coe? coe?ci cien ents ts are are restr estric icte te! ! to "e homo homoge gene neou ous s acr across oss coun countr trie ies. s. 8his 8his is particularly useful when there are reasons to e-pect that the long4run e+uili"rium relationship "etween the varia"les is similar across countries or% at least% a su"4

set of them. 8he shortrun a!@ustment is allowe! to "e country4specic% !ue to the wi!ely !i3erent impact of the vulnera"ility to nancial crises an! e-ternal shoc's% sta"iliation policies% monetary policy an! so on. However% there are several re+uirements for the vali!ity% consistency an! e?ciency of this metho!ology. &irst% the e-istence of a long4run relationship among the varia"les of interest re+uires the coe?cient on the errorBcorrection term to "e negative an! not lower than 4/. Secon!% an important assumption for the consistency of  the ARDL mo!el is that the resulting resi!ual of the error4correction mo!el "e serially uncorrelate! an! the e-planatory varia"les can "e treate! as e-ogenous. Such con!itions can "e fullle! "y inclu!ing the ARDL =p%+> lags for the !epen!ent =p> an! in!epen!ent varia"les =+> in error correction form. 8hir!% the relative sie of 8 an!  is crucial% since when "oth of them are large this allows us to use the !ynamic panel techni+ue% which helps to avoi! the "ias in the average estimators an! resolves the issue of heterogeneity. E"erhar!t an! 8eal =/C#C> argue that the treatment of heterogeneity is central to un!erstan!ing the growth process. 8herefore% failing to full these con!itions will pro!uce inconsistent estimation in ()*.  8he ()* estimator constrains the long term coe?cients to "e the same across countries an! allows only the short4term coe?cients to vary. )ean *roup =)*> estimator  8he secon! techni+ue =)*> intro!uce! "y (esaran an! Smith% =#> calls for estimating separate regressions for each country an! calculating the coe?cients as unweight means of the estimate! coe?cients for the in!ivi!ual countries. 8his !oes not impose any restrictions. It allows for all coe?cients to vary an! "e heterogeneous in the long4run an! short4run. However% the necessary con!ition for the consistency an! vali!ity of this approach is to have a su?ciently large time4series !imension of the !ata. 8he cross4country !imension shoul! also "e large =to inclu!e a"out /C to ;C countries>. A!!itionally% for small  the average estimators =)*> in this approach are +uite sensitive to outliers an! small mo!el permutations =see &avara% /CC;>.  Dynamic &i-e! E3ects =D&E> mo!el &inally% the !ynamic -e! e3ects estimator =D&E> is very similar to the ()* estimator an! imposes restrictions on the slope coe?cient an! error variances to "e e+ual across all countries in the long run. 8he D&E mo!el further restricts the spee! of a!@ustment coe?cient an! the short4run coe?cient to "e e+ual too. However% the mo!el features country4specic intercepts. D&E has cluster option to estimate intra4group correlation with the stan!ar! error =Flac'"urne an! &ran'% /CC5>. evertheless% Faltagi% *ri% an! Giong =/CCC> point out that this mo!el is su"@ect to a simultaneous e+uation "ias !ue to the en!ogeneity "etween the error term an! the lagge! !epen!ent varia"le in case of small sample sie. How to perform (AEL ARDL using stata #> import your !ata le into stata

/> now create a pool or simple stata give co!es to each cross section or entity li'e if you have !i3erent countries !ata or companies the u have to give specic co!e all countries or companies% further if you have assign co!e "y yourself  suppose u !i! not write company name li'e nestles "ut you in!icate! nestle with ### now u see you have alrea!y given the co!e "ut if you have simple right the name of company then u nee! to give also co!e  egen country#Jgroup= country> =noteK if you have countries !ata>  egen 2ompany#Jgroup= 2ompany> =noteK if you have companies !ata> ;> ;.now set time which is most important $> -tset 2ompany# year% yearly =noteK hear I have yearly !ata an! company# is new varia"le which I genrate in step />  here we shall Run )* =average>K > &irst of all install this pac'age to run (AEL ARDL ssc install -tpmg% replace :> Suppose you thin' you have installe! this pac'age "ut still you are not sure then type in comman! "ar type -tpmg 5> If u see message of no foun! then install otherwise you have alrea!y install it. > here we shall Run ()* =average>K -tpmg !.26/ !.energy !.g!p % lr=l.26/ energy g!p > ec=E28> replace pmg  #C> here we shall Run ()* =in!ivi!ual>K =8he main characteristic of ()* is that it allows short4run coe?cients% inclu!ing the intercepts% the spee! of a!@ustment to the long4run e+uili"rium values% an! error variances to "e heterogeneous country "y country% while the long4run slope coe?cients are restricte! to "e homogeneous across countries.> -tpmg !.26/ !.energy !.g!p % lr=l.26/ energy g!p > ec=E28> replace full pmg here we shall Run )* =average>K -tpmg !.26/ !.energy !.g!p % lr=l.26/ energy g!p > ec=E28> replace mg =It allows for all coe?cients to vary an! "e heterogeneous in the long4run an! short4run. However% the necessary con!ition for the consistency an! vali!ity of  this approach is to have a su?ciently large time4series !imension of the !ata.>  here we shall Run )* =In!ivi!ual>K -tpmg !.26/ !.energy !.g!p % lr=l.26/ energy g!p > ec=E28> replace full mg to select "etween ()* AD )* run housman test following is comman!  hausman mg pmg% sigmamore ow if our pro"a"ility value comes more than M we run ()* If our pro"a"ility value comes less than M we run )*  Running D&EK -tpmg !.26/ !.energy !.g!p % lr=l.26/ energy g!p > ec=E28> replace !fe N Running Hausman test to choose "etween )* an! D&EK hausman mg D&E% sigmamore oteK Suppose you want to run all these tests on your !ata% so simple @ust import your !ata into stata an! copy comman! from here into stata comman! "ar an! replace my varia"les name with yours. *oo! luc'. O

Which oneK )* vs D&EP Hausman4type test is applie! to the !i3erence "etween the )* an! the Di3erence! &i-e! E3ects =D&E>. If p4value Q C.C% we conclu!e that the D&E mo!el is preferre! over the )* mo!el.

If p4value QM% then use D&E If p4value M% then use )*