Page 1 of 1

Can eviews perform backcast in ARDL/ECM models?

Posted: Tue Jun 08, 2021 1:18 am
by stpan
I have estimated an ARDL model for the period 2015-2020. However, I would like to backcast the period 2010-2015 for which no data of the dependent variable is available. I have though all the values of all the other variables.

Is it possible with eviews? Is there a trick or loop that I can use?

Note 1: I type in the forecast sample the respective period (2010-2015). But I received an error message stating “Unable to compute due to missing data”. In my view, the missing data are exactly the ones that I would like to forecast, i.e. the dependent variable.

Note 2: I tried to use a previous forum contribution on reversing the time series but it did not work esp. as it was not clear whether I could use the same model (co-efficients), applied though in reverse order (e.g. http://forums.eviews.com/viewtopic.php? ... ing#p25848)

Re: Can eviews perform backcast in ARDL/ECM models?

Posted: Tue Jun 08, 2021 5:38 am
by EViews Gareth
An ARDL model specifies that today’s values are dependent on the past. You will need an initial value to start the forecast.

Re: Can eviews perform backcast in ARDL/ECM models?

Posted: Tue Jun 08, 2021 11:31 am
by stpan
Many thanks, I will try it. Will it be legitimate to use the long-run relationship of the ARDL model to estimate the initial value(s)? Are there any other alternatives if actual data are not available?