Page 1 of 1

Forecasts as differenced data vs actual data

Posted: Mon Jan 27, 2014 1:34 pm
by hlearner
I created a model in second differences. My equation is d(y,2) = c + mx + AR(1).
If I am working in static space, under "forecast equation," I am given the option to forecast d(y,2) OR y.
How is it that when I forecast "y", the second differences of this forecast do NOT equal the forecast values of d(y,2)?

More importantly, given a forecast of d(y,2) how do I convert this to equal the forecast eviews is giving me for y?

Thank you.

Re: Forecasts as differenced data vs actual data

Posted: Mon Jan 27, 2014 2:50 pm
by EViews Gareth
When performing a static forecast, actual values of lags of the dependent variable are used, rather than forecasted values.

When you forecast D(Y) you're actually forecasting Y - Y(-1). In a static forecast you always use actuals for lagged values, so that the forecast is:

Code: Select all

YF - Y(-1)
Where as D(YF) is:

Code: Select all

YF - YF(-1)
Hence the difference.

Re: Forecasts as differenced data vs actual data

Posted: Tue Jan 28, 2014 8:59 am
by hlearner
Thank you for your prompt answer. This makes sense now and I could derive the same output as eViews.