A recent survey by Public Policy Polling asked "Do you think that Barack Obama legitimately won the Presidential election this year, or do you think that ACORN stole it for him?" Only 35% of the respondents who voted for Romney said that he won it legitimately--59% said that ACORN stole it and 16% didn't know. In December 2004, a Gallup/CNN/USA Today Poll asked "Which comes closer to your view of this year's (2004) presidential election--(George W.) Bush won fair and square, or Bush only won because of vote fraud or other illegal means?" 60% of Democrats said he won "fair and square," while 34% said that he won because of fraud.
So it seems like Democrats were more willing to accept defeat in 2004 than Republicans were this year (although even 34% is a lot). The questions aren't quite the same, but they seem pretty comparable. The bigger problem is that PPP is an automated survey, where the questions are recorded and people answer by pushing buttons on the phone. I don't know of any research about differences in answers between automated surveys and conventional surveys with a human interviewer, but I suspect that people may take the automated survey less seriously and be more likely to pick "outrageous" options. Response rates are also lower, although it's not clear what effect that would have. Hopefully one of the conventional survey firms will ask something on this topic.
PS: I once looked for similar questions after the 1960 election, when there were many claims of fraud. But there weren't any, which says something in itself.