The link below is a time machine going back to October 2004 when the maggot infected mediots and their knee pad pollsters were declaring Kerry would be president:
They were wrong then and now!
There's nothing terribly statistically valid when 80% of those you sample in an opinion or marketing survey never give you a response.
There's an entire science behind replacement of one sample with another ~ which is particularly important in manufacturering quality control. You don't just grab something at the end of the line to make up for the designated samples you missed in the process. Unfortunately, when pollsters simply keep polling until they get enough samples to give them a reasonably high confidence level from as few as possible number of samples, they end up violating all those principles ~ the result is G.A.R.B.A.G.E.
Building a body of semi-pro respondents probably makes sense ~ provided they are always available ~ and I don't mean a response group ~ those things are just too small, and putting all the folks in the same sitting leads them to flirt with each other and minds wander (e.g.'Wow she's stacked') ~ not good at all.
Alas, now that we're down to 9% actually having input to your typical poll these days it's time for a new approach.