ARDL with Cointegrating Bounds using EVIEWS 9

Well we can now have ARDL module in EViews 9 which can replicate same results as compared to what Microfit can do with the advantage that we can have more than two lags and more than 6 variables which currently available demo version of Microfit does not allow. You can download your trial version of EViews 9 at following link

In this post  I will provide a brief tutorial to how to do ARDL in EViews rest of the details can be seen from my previous ARDL manual post.

First of all we have to import the data into the EViews 9


After that we select the variables by pressing control button and selecting the dependent variable first and independent variables after it and right click it and open it as equation. Here in the drop down menu we can see option of ARDL at bottom


Select it. It will show the options of ARDL model


Here we can fix some particular lag or use automatic selection within the maximum lags of dependent variable and independent variable. The automatic lag selection criteria can be changed from default in the option window. press ok to see the ARDL model results in the following

4 these are the basic results see here that there are 4 lags used for the dependent and 2 for the first independent and 3 for the second independent variable using AIC criteria. Now we need the Bounds F test to see if there is cointegration or not, it can be done by pressing view button on the top and going in the coefficient diagnostics


This F test will tell if we can proceed further or not
6Here we can see that our F test value of 3.5 is not bigger than any of the I1 bound value hence there is no cointegration among these variables. Since it is a tutorial I will show you further steps. If the F test value is small then we have to change the variables (add or remove)  or try adding trend variable. and If we find F test value larger then we can go for the Long run results which can be seen by pressing view button and coefficient diagnostics


it will then show the short run and long run results both


Here we can see that there was no cointegration because all the long run coefficients are insignificant and the coefficient of cointEq(-1) is also non negative and insignificant which is with the short run coefficients. These should be significant as they are important. Further diagnostics like hetroskedasticity, Auto-correlation etc can be done by selecting view and residuals diagnostics.

9 This was the brief overview of how ARDL can be operated further details can be seen in the blog link above.

Note: You can generate the CUSUM or CUSUM sq charts from the following link or update the eviews version.



300 thoughts on “ARDL with Cointegrating Bounds using EVIEWS 9

  1. Iriogbe Pamela says:

    I intend to use a dummy variable in my model, I have logged my other variables but I cannot log the dummy variable. Can I work with it like that? If not how can I go about it.

    1. Noman Arshed says:

      you cannot apply log on the dummy variable as there are 0s and 1s in it if you take log all the 0s will disappear your sample will become small. use the dummy as it is, because dummy variable is not a quantity which you can log it is a quality.

  2. Azam says:

    Eviews 9 not showing bounds test after getting ARDL Results. View—>Coefficient Diagnostics—> Shows only Four options and Bounds Test is not one of them.Please suggest.

  3. iqra rani says:

    i have check stationary but my dependent variable is at first difference stationary after taking log but all other variables becomes stationary without taking log.what should i do?guide me plz

  4. Nasir islam says:

    Dear sir
    I m using ardl model having 36 observation annual time series data d statistic value is greater then upper bound value and data have no problem of serial correlation and hetroscadasity my ecm coefficient is positive and significant kindly give me suggestion

  5. shimaa ahmad says:

    I have 11 variables in my analysis for time series, one of them is dependant & the others are independent includes one dummy variable … in fact I dont know which model can use it. By the way iapplied unitroot test and I take 2nd difference . I realy need help

    1. Dhnaya says:

      Dear sir,
      Thank you very much for your assistance with ARDL model, I have used the model to publish a paper with the instruction from this page.
      Now am struggling with GMM model, I Am using panel data, can you please give some guidance regarding the model, How do we choose instrumental variable in the Model.
      Expecting for your valuable information.

  6. Nasir islam says:

    I have 36 observation in my data I can not add any other variable in my model and ecm coefficient is positive and significant and there is no problem of auto correlation and hetroskadacity what I do kindly guide me

  7. Iyiola Adeshola says:

    Good day sir,

    I really appreciate your quick response to various questions.
    Sir, how do I interpret both the long run and short run coefficients?
    Is it the normal way of interpreting the OLS output?

    Thanks in anticipation.

  8. Z says:

    Thanks for the explanation
    I’m running an ARDL model to evaluate the determinants of economic growth using eviews 9, some variables hav long run influence others have short run influence. the estimation in eviews 9 include all variables in the long run relation. I want exclude those which have just short run from the cointegration relationship, does ardl support that ? is it possible to make it automatically on eviews 9 or i have to do it manually ?

  9. bahaa says:

    We employ a hedging model between gold and inflation with the addition of a dummy variable as an intercept and trend at the same time, how can we write the equation.

  10. Josephine says:

    Can I know that when we specify the dependent and independent variables under the dynamic specification, if some of them are I(1), do we need to put a “d” in front?

      1. sam sam says:

        Good Evening
        Sorry professor, I found in the test of stationarity of the serie at the level it is of type DS (stationary defference), and with the first difference that it is of type TS (Trend Stationary). in this case, should we do the detrending or the differentiation? And I can do the ARDL model?

        And if we apply the detrend it implies that this variable is integrated of order 1 ie. (I (1)) or 2 i.e. (I (2))?

  11. student says:

    may i ask a favour, if F-statistic less than the upper and lower bound, I need to proceed with ecm or other method. thanks in advance for responding.

    1. sam sam says:

      Essalamo Alaykom
      Please Sir, If a series has a trend in level, then it is not stationary but it is trend stationary, This means are I(0) or I(1)?

  12. sam sam says:

    Essalamo Alaykom
    Please Sir, If a series has a trend, then it is not stationary but it is trend stationary, This means are I(1)?

    1. sam sam says:

      Essalamo Alaykom
      Please Sir, If a series has a trend in level, then it is not stationary but it is trend stationary, This means are I(0) or I(1)?

    2. sam sam says:

      Essalamo Alaykom
      Please Sir, If a series has a trend, then it is not stationary but it is trend stationary, This means are I(1)?

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s