SC-CBS/YouGov: D: Clinton 59% Sanders 40%; R: Trump 42% Cruz 20% Rubio 15% Kasich 9%

[Shadows]

"Margin of error of 9" doesn't mean "It could be [reported result-9] just as it could be [reported result+9]." It just means there's a 95% probability that the result will fall somewhere within that range.

It is Plus & minus 9. It does mean that.

What you are talking about is Confidence Interval. Most surveys are conducted with a confidence interval of 95%.

[cinyc]

"Margin of error of 9" doesn't mean "It could be [reported result-9] just as it could be [reported result+9]." It just means there's a 95% probability that the result will fall somewhere within that range.

It is Plus & minus 9. It does mean that.

What you are talking about is Confidence Interval. Most surveys are conducted with a confidence interval of 95%.

No, it doesn't.  The probability that the result is + or -9 is much, much lower than it will be exactly right or + or -1, for example.  Every poll can be turned into a probability forecast, which is what 538 essentially does.  The actual probability that it is + or -9 or worse is likely around 5%.

[IceSpear]

The black numbers are actually pretty similar to ARG's poll (73-26 vs. 74-19.) The main difference is that whites go 54-35 Clinton in ARG and 63-36 Sanders in this. Bold prediction: It will fall somewhere between these two extremes.

[muon2]

"Margin of error of 9" doesn't mean "It could be [reported result-9] just as it could be [reported result+9]." It just means there's a 95% probability that the result will fall somewhere within that range.

It is Plus & minus 9. It does mean that.

What you are talking about is Confidence Interval. Most surveys are conducted with a confidence interval of 95%.

No, it doesn't.  The probability that the result is + or -9 is much, much lower than it will be exactly right or + or -1, for example.  Every poll can be turned into a probability forecast, which is what 538 essentially does.  The actual probability that it is + or -9 or worse is likely around 5%.

Tony and cinyc are mostly correct. The confidence interval defines the the range that the actual result falls within at some given probability, which is usually 95% for political polls. The MoE is usually defined as half the confidence interval, and that is what political polls report. To the extent that the statistics fall in a normal distribution the MoE and confidence interval are directly correlated. For skewed samples they may not be completely correlated, but that is not as likely in the Dem race with just two candidates both polling more towards 50% than towards 0% and 100%.

[○∙◄☻¥tπ[╪AV┼cVê└]

All 49 voters under 30 said Bernie is honest and trustworthy.

[Panda Express]

"Margin of error of 9" doesn't mean "It could be [reported result-9] just as it could be [reported result+9]." It just means there's a 95% probability that the result will fall somewhere within that range.

It is Plus & minus 9. It does mean that.

What you are talking about is Confidence Interval. Most surveys are conducted with a confidence interval of 95%.

No, it doesn't.  The probability that the result is + or -9 is much, much lower than it will be exactly right or + or -1, for example.  Every poll can be turned into a probability forecast, which is what 538 essentially does.  The actual probability that it is + or -9 or worse is likely around 5%.

Tony and cinyc are mostly correct. The confidence interval defines the the range that the actual result falls within at some given probability, which is usually 95% for political polls. The MoE is usually defined as half the confidence interval, and that is what political polls report. To the extent that the statistics fall in a normal distribution the MoE and confidence interval are directly correlated. For skewed samples they may not be completely correlated, but that is not as likely in the Dem race with just two candidates both polling more towards 50% than towards 0% and 100%.

muon2 are you supporting trump?

[muon2]

"Margin of error of 9" doesn't mean "It could be [reported result-9] just as it could be [reported result+9]." It just means there's a 95% probability that the result will fall somewhere within that range.

It is Plus & minus 9. It does mean that.

What you are talking about is Confidence Interval. Most surveys are conducted with a confidence interval of 95%.

No, it doesn't.  The probability that the result is + or -9 is much, much lower than it will be exactly right or + or -1, for example.  Every poll can be turned into a probability forecast, which is what 538 essentially does.  The actual probability that it is + or -9 or worse is likely around 5%.

Tony and cinyc are mostly correct. The confidence interval defines the the range that the actual result falls within at some given probability, which is usually 95% for political polls. The MoE is usually defined as half the confidence interval, and that is what political polls report. To the extent that the statistics fall in a normal distribution the MoE and confidence interval are directly correlated. For skewed samples they may not be completely correlated, but that is not as likely in the Dem race with just two candidates both polling more towards 50% than towards 0% and 100%.

muon2 are you supporting trump?

That would make this a pretty strange thread.

[RBH]

Hillary goes from 60/38 to 59/40 in about 3 weeks, with 13 days left.

YouGov polls have some sampling oddities, but it doesn't look like much changed in their universe.