I haven't seen a diary of NC/IN polls recently, so I decided to do it myself. So, I just gathered up all the latest polls out of North Carolina and Indiana.
FiveThirtyEight.com, which makes predictions based on demographic data from past polls, has predicted the margins for every Congressional district in both states. Their models predict a net gain of 17 delegates for Obama (102 to 85). However, in both states, FiveThirtyEight's predictions are more optimistic for Obama than the polls seem to be.
All the polls you could ever want, below the fold.
There is good reason to be wary of polls showing a close race in North Carolina. Up till now, the polls have consistently underestimated two things in southern states: Obama's margin of victory among African-Americans, and African-American turnout. Check out the final poll-based predictions for the states closest to North Carolina, along with the actual results:
South Carolina
final RCP average: Obama +11.6%
final Pollster.com average: Obama +14.6%
actual result: Obama +28.9%
Virginia
final RCP average: Obama +17.7%
final Pollster.com average: Obama +22.0%
actual result: Obama +28.2%
Georgia
final RCP average: Obama +18.0%
final Pollster.com average: Obama +13.9%
actual result: Obama +35.3%
By now, maybe the pollsters have corrected their old models. But maybe not
NORTH CAROLINA
FiveThirtyEight.com's prediction:
Clinton 41.4%, 49 delegates
Obama 58.6%, 66 delegates
diff +17.2%, +17 delegates
Real Clear Politics average:
Clinton 42.7%
Obama 49.7%
diff +7.0%
Pollster.com average:
Clinton 42.2%
Obama 49.9%
diff +7.7%
____________________________________________________
Insider Advantage/Majority Opinion. 5/4 (5/1) (4/29)
Clinton 45% (44%) (44%)
Obama 48% (49%) (42%)
diff +3% (+5%) (–2%)
Public Policy Polling. 5/3–5/4 (4/26–4/27)
Clinton 43% (39%)
Obama 53% (51%)
diff +10% (+12%)
Zogby tracking poll. 5/3–5/4 (5/2–5/3) (5/1–5/2) (4/30–5/1)
Clinton 40% (39%) (37%) (34%)
Obama 48% (48%) (46%) (50%)
diff +8% (+9%) (+9%) (+16%)
American Research Group. 5/2–5/4 (4/30–5/1) (4/26–4/27)
Clinton 42% (41%) (42%)
Obama 50% (52%) (52%)
diff +8% (+11%) (+10%)
Survey USA. 5/2–5/4 (4/26–4/28)
Clinton 45% (44%)
Obama 50% (49%)
diff +5% (+5%)
Rasmussen Reports. 5/1 (4/28)
Clinton 40% (37%)
Obama 49% (51%)
diff +9% (+14%)
Research 2000. 4/29–4/30
Clinton 44%
Obama 51%
diff +7%
Mason–Dixon. 4/28–4/29
Clinton 42%
Obama 49%
diff +7%
INDIANA
FiveThirtyEight.com's prediction:
Clinton 51.0%, 36 delegates
Obama 49.0%, 36 delegates
diff –2.0%, 0 delegates
Real Clear Politics average:
Clinton 48.8%
Obama 43.8%
diff –5.0%
Pollster.com average:
Clinton 49.7%
Obama 43.9%
diff –5.8%
____________________________________________________
Insider Advantage/Majority Opinion. 5/4 (4/30–5/1)
Clinton 48% (47%)
Obama 44% (40%)
diff –4% (–7%)
Zogby tracking. 5/3–5/4 (5/2–5/3) (5/1–5/2) (4/30–5/1)
Clinton 42% (41%) (42%) (42%)
Obama 44% (43%) (43%) (42%)
diff +2% (+2%) (+1%) (0%)
Public Policy Polling. 5/3–5/4 (4/26–4/27)
Clinton 49% (50%)
Obama 44% (42%)
diff –5% (–8%)
Suffolk University. 5/3–5/4
Clinton 49%
Obama 43%
diff –6%
American Research Group. 5/2–5/4 (4/30–5/1)
Clinton 53% (53%)
Obama 45% (44%)
diff –8% (–9%)
Survey USA. 5/2–5/4 (4/28–4/30) (4/25–4/27)
Clinton 54% (52%) (52%)
Obama 42% (45%) (43%)
diff –12% (–7%) (–9%)
Rasmussen. 4/29
Clinton 46%
Obama 41%
diff –5%
TeleResearch Corp. 4/25–4/29
Clinton 48%
Obama 38%
diff –10%
____________________________________________________
My own prediction: Obama wins NC 57%-43%; Clinton wins IN 52%-48%.
Consider this an open thread for your own predictions.
____________________________________________________
Update:
I've seen a lot of comments in DailyKos commenting on how we should ignore the results of some (inaccurate) pollster, and only put our stock in the results of some (more accurate) pollster.
Polls, by definition, are random samples. One week, a pollster might stumble across 500 people who are representative of the electorate, and the next week they'll just happen to stumble across a more Obama-friendly crowd, or a more Clinton-friendly crowd. That doesn't make them a bad pollster, it just makes them unlucky. It's silly to ignore certain pollsters or to put faith in others because of their past performances. Instead, look at the average of all the polls, and you end up with a huge sample size that has much less of a chance of being wrong... I'm not saying that individual pollsters don't vary in quality, I'm just saying that its better to take an average of all of the polls than it is to put all your eggs in one basket (even if it's a higher-quality basket).
Keep in mind, the only real say a pollster has in their results is: 'likely voter' screens, oversampling minority groups, and predicting turnout. Oversampling minorities is important for two reasons: 1) Most of the time, the urban working class are less likely to have telephones, less likely to have the time to respond to a telephone poll, and are more likely to be working during polling hours. So a random telephone survey tends to undersample this group; 2) As voting patterns obviously vary tremendously between demographics, it's important to have a decent sample size for each demographic group. a sample of only 50 people has the potential to be very inaccurate.... The way that pollsters account for these problems is to take a disproportionally large sample of the desired demographic, and then adjust that to the predicted % of the electorate they will make up. Not all pollsters do this, unfortunately (especially in state races).
And then there's predicting turnout. Which is pretty much guesswork.