VECM. Part 1 of 2. Model Five. EVIEWS

hi hello hello today we shall be developing their broken PCM model right our vector error correction model and here are three here are three steps involved in developing be ECM model so step number one lacks election right that is step number one and step number two Johanssen Johansson test off cointegration and the final step is the vector error correction model okay so that so the first step is the lack selection that we do first right lacks selection how many lags we should choose we should select to run our model again here you can see we have three variables cons IX and GDP three variables cons X and GDP I open all variable as a group dad you can see from here right chorus cons cons means consumption e^x means export and GDP means gross domestic product and we have the data from 1960 until 1990 95 okay so the first step is the lack selection okay I go too quick then I go to estimate far right then here I put my three variables cons e^x and GDP and here I take lakh - for the time being I take lakh - then I press ok so so actually this is the vector error correction model and here you can see cons e^x and GDP these are dead dependent variable and fro from this model we can select the lakh right okay go to view then I go to lag structure then I go to lag length criteria that I go ok here the lakhs to include better I take 4 lakhs to include right 4 lakhs to include then I press ok so here here you can see the outcomes right so here are few cry 2dr here you can see so here the the criteria is lr f be AIC SC + hu and here is given LR means this one since this sequential modified LR test FB final prediction error AIC SC HQ and these are the lakhs right these are the next okay so here the first one is a log and here one thing you can see a star here right star indicates lag order selected by that right here so here suppose in case of LR so the star is here that means twenty seven point four seven meaning that three lakh should be chosen okay and here we have the second criteria FB star is here meaning that three lakh should be chosen and here the AIC so here what is happening lower the value better the model all the time and here you can see 17.3 two is the highest and two point one eight nine is the lowest lower the value better the model so this value is the lowest and that is why there is a star right so meaning that three lakh should be chosen and also here here also the same guideline the lower the value better better the model and here three point four four this value is the lowest and that is why the star is there and here HQ also the same criteria lower the value better the model and here two point six four is the lowest and that is why there is a star okay now actually all criteria are good now it is up to you to decide which criteria you will choose that is up to you all are good okay but but here what is happening if I take L R so three lag is asking for and FP also three lag has been recommended and also AIC three lakhs have been recommended and also HQ three lakhs have been recommended so that means here one two three four criteria are asking to take three lakh so better we take three lakh in our model meaning that optimum lack optimum optimum lakh would be three and and we shall use this three lakh in Johnson test and also in the vector error correction model right okay so the latch selection is done okay now the second step is the Johnson test of cointegration okay we proceed with the Johnson test of point that is number two step but the precondition of Johnson test is what there is a precondition variables must be non stationary at Louisville but when we convert all the variables into first difference right then they will become stationary meaning that all the variables should be integrated integrated of same order only then we can run the Johnson desktop cointegration sir variables must be non stationary at level but when we convert the variables into first difference then they will become stationary and if it is happens then only we can run the johnson desktop cointegration okay we check actually what happens about our three variables right they become I put it become right okay okay now we check our three variables okay so we take one by one first I go to quick series statistics then I go to correlogram so the first variable is coms right that I select first then I press okay okay first I check at the level that means the variable is at the raw data the initial data that I select Louisville then press ok and here you can see the the correlogram of comms at Louisville and you can see the queue statistics and corresponding probability value right and here what is our null hypothesis null hypothesis that variable is stationary right and alternative hypothesis variable is not stationary right that is alternative hypothesis and here the p-value is very small less than 5% so we can reject the null hypothesis and can accept the alternative hypothesis meaning that variable is not stationary at level right so that means variable is not stationary at number okay now I check what happens after the first difference of the coms variable so I go to view I go to Cola Graham so this time I select fast difference this one right then I press okay okay here you can see the queue statistics and a corresponding probability suppose this one probability is 6.6 percent this one 18.3% and this this one 30 percent all are more than 5 percent meaning that we cannot reject null hypothesis rather we accept null hypothesis meaning that variable is stationary after first difference that means D cons you can see D comes this is stationary after first difference right okay so so that is that done now we check our second variable so I go to as before series statistics Corolla graph I take the e^x or second variable right I check this one export okay so I take at level so this is the outcome you see from here right this outcome and correlogram e^x as before Q statistics and corresponding probability here the p-value is less than 5% we can reject the null hypothesis and can accept the alternative hypothesis meaning that this e^x variable is not stationary at level okay now I go to view and I do that coca-cola gram as therefore and I take the first difference of e^x variable right I press ok as before you can see this is the first difference of e^x and queue statistics is here and the p-value you can see all are more than 5% that means we cannot reject null hypothesis meaning that our first difference of e^x is stationary okay then we check our third variable that is GDP if I think I think the GDP is the last term I put it GDP right I do it then I press ok as therefore so I choose they choose that level right so this is the outcome of GDP and say so it is the probability value you can see we can reject the null hypothesis and can accept the alternative hypothesis meaning that GDP this variable is non-stationary at level but but but I go to view correlogram but when I choose first difference right then what happens so I press ok so we call it color gram of D GDP first difference of GDP what happens here you see Q statistics and corresponding probability value so probability value all are more than 5% meaning that we cannot reject null hypothesis rather we accept null hypothesis meaning that the first difference of GDP is stationary so what is our what is our summary summary is our three variables are non stationary at level but when we convert all the variables into first difference then they become stationary that means all our three variables are integrated of same order so now we can easily run Johnson test of go integration because precondition has been okay so we run the Johnson test of cointegration I go to quick then I go to group statistics then I go to Johnson test of cointegration okay so here I put our three variables Khan's IX than GDP and here these three variables are non stationary make sure here we must put non stationary variable right and we know that after first difference they will become stationary that we have just seen okay I press ok and here a lakh selection here I take lakh 3 because lagg-3 has been has been said by the lakh selection criteria and here I select 3 right then I press ok to run the Johnson test of cointegration and it is here right I run it yes the result is coming up you can see the result from here this is the Johnson test of cointegration that you can see from here right just going down going up a John sent and we have three variables okay any Johnson test of cointegration right it is here okay okay first we take tray statistics okay so and we have three variables okay and it is the null hypothesis number of cointegrating equation or number of cointegrating model so the first all hypothesis is none meaning that there is no cointegration there is no co-integration among the three variable such as coms e^x and GDP that is the null hypothesis that means none ok here that retraced statistics how much 35.6 s and critical value at 5% is thirty nine point seven nine so what is the guideline the guideline is that if that trace statistics is more than critical value we can reject null hypothesis and here thirty five point six six is more than thirty nine point seven nine meaning that we can reject this none and also we can check it from the probability value here probability value is very small 0.94% which is less than five percent which is less than five percent meaning that we can reject this null hypothesis which is none okay now our second null hypothesis is at most one meaning that there is one girl cointegration model right at most one okay here is this that the statistics is how much 8.5 one critical value is fifteen point four nine so here statistics is less than critical value so we cannot reject null hypothesis rather we accept null hypothesis meaning that there is at most one cointegration model and also from the probability value the p-value is forty one point two five percent which is more than five percent so we cannot reject null hypothesis rather we accept null hypothesis meaning that there is at most one cointegration model and the result is also given here test test indicates one cointegration model at 5 percent level because here is 5 percent right so meaning that our three variables are cointegrated that means cons e^x and jdb are cointegrated or in in other words our three variables cons e^x and GDP had have long-run Association XI or in the long run they move together okay then our second test is maximum eigenvalue test and and and and the thing is same almost same number of cointegrating a equation or model the null hypothesis is none alright there is no cointegration among the three variables and here is the statistics twenty seven point one five is more than critical value twenty one point one three meaning that we can reject this non hypothesis and also probability is very small 0.63% we can also reject null hypothesis meaning that this one which is none okay then we come here at most one here seven point eight four is less than fourteen point two six so we cannot reject null hypothesis rather we accept null hypothesis meaning that there is at least one way integration model and also the probability value is telling the same thing and the probability value is thirty nine point five one percent meaning that we cannot reject null hypothesis rather we accept null hypothesis meaning that there is one cointegration at least one cointegration model and the decision also given here max eigenvalue test indicates one cointegration model at five person level meaning that our three variables such as khan's e^x and gdp are cointegrated or they have long-run association shape or in the long run they move together okay now what we have seen tray statistics and maximum eigenvalue test they are telling the same thing and that is our three variables cons x and gdp are cointegrated or or they have long run association shape okay so and now so now now what is the guideline if the variables are cointegrated or have long run association shape have long run association shape then we can run restricted bar and that is and that is VE see a model that means vector error correction model but if the variables are not cointegrated we cannot run b ECM model rather we shall run unrestricted bar but here our three variables are cointegrated so we can easily run vector error correction model and in the next video in the next video we shall run it we shall run the B cm model in the next video thank you very much for being with me

Channel: Sayed Hossain

Views: 69,380

Rating: 4.8266668 out of 5

Keywords: regression, vecm, eviews, econometrics, VECM, VAR, Lag Selection, JOHANSEN COINTEGRATION, Sayed Hossain

Id: MSLNkgyygP0

Length: 28min 26sec (1706 seconds)

Published: Thu Sep 05 2013