HOW'D THE POLLS DO?
TEXAS
Kudos to the pollsters overall: every single one of the pollsters that kept polling into the very last days of the campaign -- and there were seven of them -- correctly pegged the last-minute trend towards Hillary. All of them had Hillary gaining on Obama in their last poll out.
For over a week the three-day average had shown Obama in the lead; only on March 2, two days before the election, did the balance swing to Hillary again. On average, the polls whose 'midpoint'* was in the last three days of the campaign (March 1-3) showed Hillary leading by 1.4%. If you look only at the polls whose midpoint was in the last two days, the average even pegged Hillary's lead at 2.3%. In reality, it was 4%. Not bad.
Not all pollsters did equally well though, of course. This update of my graph shows that: the big red ball on the right is the actual result.
InsiderAdvantage, Zogby and ARG were closest; their last poll out had Hillary leading by either 3% or 5%. Zogby's kudos here rest on a last-minute correction; the second last edition of its daily tracking poll still had Obama leading by 3%.
But overall none of them did badly. PPP and ABC/WaPo also had the lead within 3 points, well within the margin of error (remember that the MoE doubles if you look at the difference between two candidates rather than at one candidate's number). Survey USA and Rasmussen were five points off, which should still be within the MoE.
OHIO
Again, the pollsters did pretty well all in all. In Ohio, too, all the six pollsters that tracked public opinion into the last three days of the campaign picked up on the movement towards Hillary in those last days.
In all, a three-day average of polls had Hillary's lead down to 3-4% a few days before the primaries. But come the day before the elections, Hillary's lead was at over 7% again. Narrow it to a two-day average and her lead was 8.5%.
In reality, it was 10%. So again, Hillary outdid the polls still, but they got close.
There were bigger differences between the polls here though. Or one, anyhow. See this update of my graph, with the big red ball on the right designating the actual result. Most all of the polls trended right towards the actual result, though ARG predictably overshot the target somewhat. But one pollster was off in leftfield, and ended up looking a little silly: Zogby.
The last edition of Zogby's daily tracking poll was off a full 10%. Survey USA on the other hand was right on spot, polling a 10-point Hillary lead. Which is an exact echo of what happened in California, when Zogby predicted an Obama win and ended up way off, while Survey USA pegged the size of the Hillary win spot-on.
SUSA wasnt the only pollster who did well though. PPP and the University of Cincinnatti came within 1 point. The Suffolk poll was very close too (sorry Cyclo), at 2 points from the actual result. Rasmussen and ARG were both off 4 points. Rasmussen did a good job in its last poll out, which corrected a mere 2-point Hillary lead to a 6-point one; ARG would have been better off not doing its very last poll, as its previous one had been closer (still). But all of them well within the margin of error. Only Zogby failed.
LAST TWO WEEKS OF CAMPAIGN A WASH
Interesting point to note: the end results in both states basically mirrored where the race had been two weeks out from the elections. On 16-17 February, the development of the polls basically showed a 10-12 point Hillary lead in Ohio and a 4-5 point Hillary lead in Texas, and that's pretty much where the race ended up at. So in the last two weeks out campaigning, the two candidaes basically scored a draw (whereas in the weeks before Obama had been steadily winning).
* By "midpoint" I mean that, if a poll was conducted over several days, the middle of that period fell in these three days. (And I've included polls that were conducted on Feb 29 - March 1.)
