Indiana & North Carolina Primaries: Which Polls will be Wrong?

Since taking statistics class some 20 years ago, I've been troubled by the media's repeating of pre-election polling data, as though the numbers are reliable.  Fact: sometimes polling numbers turn out to be right, and sometimes they don't (e.g., the pre-primary polls in Iowa, New Hampshire, and California). 

It's like off-track betting -- you  have to wait until the race is over to know whether you've won money.  All media personalities should grasp the uncertainty of pre-election polling data by now, yet they continue to broadcast the numbers with an authoritative tone.

That said, I've created tables (below) of some recent polling data on tomorrow's Indiana and North Carolina Democratic primaries.

Indiana
...............SUSA.......Zogby-2.....ARG.........Zogby-1.......Suffolk
..............(May 2-4)......(May 3-4).......(May 2-4)...(Apr 30 - May 1)....(May 3-4)

Clinton.......54%..............42%.............53%............42%................49%

Obama.......42%...............44%...........45%..............42%................43%

Und..............1%..................8%...............2%.................9%..................6%
.

North Carolina

..............SUSA..........Zogby-2......ARG........Zogby-1.......PPP
.............(May 2-4).........(May 3-4).......(May 2-4)....(Apr 30-May1)....(May3-4)

Clinton.......45%................40%............42%............34%...............43%

Obama.......50%................48%...........50%.............50%...........53%

Und...............2%.................8%...............4%................8%................7%

Additionally, SurveyUSA did a tracking poll yesterday for a Raleigh TV station: it has Obama at 50% and Hillary at 45%, and it doesn't list undecided's.  About 25% of the survey participants said that they had already voted, and 16% of them said they'd voted for Obama.

In North Carolina, most polls show Obama ahead but that Hillary has shrunk his lead.  Nate at FiveThirtyEight.com did some sort of analysis based on demographics and predicts that Obama will win North Carolina by double digits.

On the other hand, Congressional Quarterly predicts a "slim" delegate victory for Obama in North Carolina. 

Frankly, I don't know why so many people continue making predictions, given how many pre-primary predictions turned out to be wrong this year.

Whatever the varying polling numbers say, no one really knows what's going to happen. 

If you haven't voted yet and you're properly registered in Indiana or North Carolina, please vote today -- knowing that your vote really does count tomorrow's election.

Memeorandum has commentary.

Related BN-Politics Posts:

* Limits of Polling: John Zogby Tells Jon Stewart

* Nationwide Polls Conflict: Just Pick the one You Like? 

* Super Tuesday: Which Polls and Predictions were Wrong?

* Not Just New Hampshire: Some Polls Called Wrong Winners in Iowa

 

Comments

[Adam]

Zogby has been, on average, 4.6% in favor of Obama, while ARG is about the same, except in Clinton's favor. I'm sure you remember Zogby's multiple bad guesses.

Fivethirtyeight's analysis is actually not based on polls at all. It's based on demographics (AA%, age, income, party leanings in past election, et cetera). That's why it can predict a result outside the range of ALL the polls. Should be an interesting test case of his methods, and/or the "likely voter" models the pollsters use.

[D. Cupples]

Hi adam,

You're up late! Yeah, I linked to posts that mention Zogby's screw-ups.

Before Ohio (or maybe PA), Zogby had outlier polls in favor of Obama and quickly released new polls that were more in line with the others. I don't understand why Zogby has been so wrong this season.

Thanks for the tip on 538: I'll correct the post.