That raises the obvious question: why rely on 2008? They're pollsters after all. Can't they do a poll to estimate the turnout percentages for 2012 and go from there?
Because they actually came out to vote. Adults/registered voters/likely voters are an unstable number, but actual voters who do go to the polls - for sure - are a "sure thing." The odds of them voting again are pretty solid.
2008 tells them what % voted for who, so they use that as their guide. Those same people will vote again (although, personally, I think a lot of democrats will just stay home. They won't want 4 more years of this disaster, but they won't vote Republican either).