Tuesday, November 20, 2012

How Exit Polls changed as election night wore on


One interesting feature of the 2012 general election is the fact that historically unreliable EARLY exit poll data actually proved to be more favorable to Romney than Obama, going against the popular notion that early numbers tend to favor Democrats. Several metrics, ranging from Party ID, to views of government, to race/ethnicity showed  more promising numbers for Romney earlier in the night, only to magically change as more and more returns were actually counted.

For example, just two hours before the first polls closed on the East Coast, politico reporter Jamie Dupree tweeted the following:  
"41% say the government should do more - 53% say the government is doing too much; it was flipped 51-43 in 2008"

But by the time the night was over, that 12 point lead for those thinking the government is doing TOO much had dwindled to just an 8 point, 51-43% lead.

Jon Karl of ABC News also provided some flawed preliminary exit poll data. See these tweets from just before 6 pm E.S.T., election night 2012:

"In two ABC prelim exit polls, Romney slightly edges Obama on handling of the economy and the deficit (51-47 and 50-46 percent)"

"In our ABC preliminary exit poll, voters give Obama the edge over Romney on who is more in touch with them (52-44)"
 What were the actual numbers? In the end, Romney's edge over Obama on the economy and deficit was NOT the 4 points early exit data indicated. In fact, it wound up being just 1 point, 49-48%, and just 2 points on the budget deficit (49-47%). Meanwhile, the number who felt the President was more in touch with them than Romney grew from an 8 point deficit to a ten point deficit (53-43% for Obama).

The difference between early, preliminary exit poll data, and the final result was evident in several other places, but perhaps none more consequential than partisan identification. At about 6:05pm, E.S.T., one hour before the first polls closed on the East Coast, CNN provided viewers with (and I tweeted) the preliminary Party ID data:


 "Preliminary PARTY ID is 37% Democrat, 34% Republican, 29% Independent (D+3) (all per EARLY exit polls. Could change)"
Had those numbers held through the night, they would have represented a 2% drop in the number of Democrats voting from 2008, a 2% increase in the number of Republicans voting, with Independent voters staying the same. But alas, the D+3 party ID first reported by CNN did not last, and it eventually became what we know it to be today, (38D/32R/29I, D+6). But what if the CNN preliminary reports had been correct? Would Romney's 5 point victory among Independent voters have been enough to put him over the top in the event the Democrats turnout advantage was D+3 instead of D+6%?

2012 election results reweighted to EARLY exit poll party I.D. data:


As you can see, had early exit polling data held through election night, all other things remaining the same, the race would have essentially been tied at 49-49% in the popular vote, as opposed to Obama's actual 51-47% victory.

As someone who was duped by reports that early exit polling data always favors Democrats, let  election 2012 be your reminder: don't fall for it!           

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.