opinion on automatic general to specific modelling selection

[FilipinoEconomist]

Hi everyone,
I just want to know if you have any knowledge/opinion about automatic selections like general-to-specific (e.g. PCGETS). Is this valid practice or bad econometrics? Any Econometrica-type publication- "evidence" (other than papers from the Oxford group selling the software) to support the position would be gratefully appreciated. Thanks!