Problem backcasting with fit/forecast

[aarnon]

EViews 9 Enterprise, 64-bit, April 3, 2015 build.

I am encountering a problem using the fit or forecast procedures to backcast from an equation (putting this here since this issue did not exist in Eviews 8 ). In my estimated equation, I have a RHS variable that starts earlier than my LHS variable, and I would like to use those RHS values to create fitted values for the periods before my LHS variable begins. If I set the sample to the max range of my RHS variable, fitted values are generated only for the periods in which LHS values already existed. If I set the sample to exactly the periods for which RHS exists but LHS does not, it creates fitted values for that sample. However, if I add the first period in which LHS values do exist, fit will only create a value in that one period of overlap.

See the attached program for an example. In it, the series LHS exists from 1980 to 1990 and the series RHS exists from 1970 to 2000. I estimate "ls lhs c rhs" over 1980-1990 and want to use the equation to create fitted values for LHS for 1970-2000. If I set the sample to 1970-2000, fit (or forecast) creates values for 1980-2000. If I set the sample to 1970-1979, fit creates values for 1970-1979. If I set the sample to 1970-1980, fit creates a value only for 1980...

Any help much appreciated.

[EViews Gareth]

We'll have a fix soon.