This was, in fact, the methodology we were given at the state level in 2004 by Karl Rove to identify "the base." But it's much different than just saying, "Are you a Republican or a Democrat?"
There are other independent surveys taken all over the place through a variety of means to get a better idea of what "the base" really is.
You can use that to validate/verify the probability of a recent survey to more correctly reflect the popular sentiment.
BTW, there also some expensive ways to do this that are used frequently in marketing. First you get your peeps to sample. You know everything about them. Then, you hit them with questions down the line ~ on all sorts of things. You correlate your KNOWN UNIVERSE OF RESPONDENTS to product market penetration, or politics, or whatever.
The biggest problem I can see with using that method for politics is that it's more expensive than you need AND people do shift political preferences over time so you'd need to keep updating your universe.