Professor Sam Wang of Princeton provides the gory details,
The negative GOP bounce. As I stated before, the GOP convention was of no help to them in the Electoral College. Indeed, it appears that the race shifted towards President Obama by 6-15 EV, or about 1.0% of Popular Vote Meta-Margin. From an analytical perspective, a negative bounce is quite remarkable because all the talk in recent weeks has been of bounces being smaller or zero, but always in the hosting party’s favor. It is all the more remarkable because of the relatively small number of state polls over the last week, so that the Meta-analysis’s inputs have not fully turned over (for discussion see comments). So the negative bounce may be larger than what is shown in the graph. Such an event would have been missed in past years (and even this year) because national polls don’t have the best resolution.
The natural question arises: why would the Republicans be hurt by their own convention? Two answers come to mind.
(1) The Ryan-VP bounce effectively used up whatever room there was for a bounce. This year, opinion seems to be fluctuating in a very narrow range: Obama up by 1.0-5.0%. Maybe there was no room for improvement.
(2) The GOP convention was not particularly inspiring. Indeed, the most notable event was Clint Eastwood’s empty-chair routine, which overshadowed Romney’s acceptance speech.
The GOP convention negative bounce: a final look