New Harvard Poll Shows Harris With a 31-Point Lead Over Trump (with young voters)

[RiverStone]

CAMBRIDGE, Mass. — A new national poll from the Institute of Politics at the Harvard Kennedy School shows Vice President Kamala Harris with a 31-point lead over former President Donald Trump in a multi-candidate matchup among likely voters aged 18-29. In a head-to-head race Harris leads 64% to 32% and the numbers continue to rise as more young people become involved. (full article below)

https://www.thewellnews.com/opinion-polls/new-harvard-poll-shows-harris-with-a-31-point-lead-over-trump/

[Fiendish Thingy]

1. This is a massive poll with a fairly low MOE

Compared to most polls, whose subgroups of young voters are often 200 or less, with a corresponding high MOE.

[WarGamer]

2. What were the numbers in 2020?

I googled it myself... in 2020, Biden got 60% of the 18-29 vote.

If this poll comes to fruition it's a nice bump from 2020.

[Wiz Imp]

3. Strong evidence that the recent Quinnipiac poll is not to be trusted.

Now their youngest demographic was 18-34, so slightly different, but they had that demographic at just 48% Harris to 45% Trump. For both to be accurate would mean that age 30-34 would have to be near 100% Trump which we know is not true. So for those who may have freaked out over the Quinnipiac poll, this is some strong evidence casting doubt on its accuracy.

[Self Esteem]

4. No. It just means crosstabs are not to be trusted.

There's a difference between a poll specifically sampling a demographic, as the Harvard Poll does here, and the crosstabs of a poll sampling that demographic.

Crosstabs generally have a high MOE because they're a subset of the population polled. Quinnipiac polled 1,728 likely voters.

I don't have the total breakdown of what percentage of those polled were 18-34 but in 2020, 18-29 made up 17% of the electorate and 30-44 23%, so, let's just assume 18-34 was 20% of the poll.

That means it's just 346 total from that group.

Which likely has a significant MOE.

Couple that with the fact Biden did worse among 30-44 (+6) than he did 18-29 (+24), his lead among 18-34 was probably well below the +24 he received...likely closer to +15 or so, which is likely where Harris is at nationally. Add the MOE (which is frequently +/-10 on crosstabs and her lead could be +13 among this group ... which is a lot closer.

The Quinnipiac Poll might be an outlier but the fact other polls show the race tightening makes me thin it probably is closer to what we're seeing than not.

[Wiz Imp]

5. If the crosstabs are not to be trusted then the top line is not to be trusted either

If the crosstabs show they are underestimating Harris support from younger voters, then that means the top line is likely underestimating Harris support overall. There are plenty of other reasons to question this poll including their PA and Michigan polls from a few days ago showing 6 & 5 point leads. If Harris is leading by that much in those 2 states then there is no way she is trailing nationally. (And before you say it - I know they are different polls with different samples - but the fact that different polls from the same pollster show incompatible results within a couple days causes me to question the reliability of that pollster's latest poll(s).

And other polls do not really show the race tightening. Most pollsters show a bigger lead for Harris in their latest poll than they did in their previous poll. That includes Quinnipiac and the NY Times.

[Self Esteem]

6. This is just a complete misunderstanding of how polling works.

I already told you why crosstabs are different than polling: they tend to have a larger MOE because there's a much smaller sample size among specific demographics. That has nothing to do with the topline and the topline's MOE.

[Wiz Imp]

7. I don't need you to tell me how polling works

I know how polling works (I worked as a statistician for 33 years) and I know when the results of a poll are questionable. This poll is highly questionable.

[Self Esteem]

9. Do you? Then you'd know that crosstabs are not a reliable indicator of anything...

Because they have a much higher MOE than the actual topline. Here's a good article (even if it's The Hill) that explains it:

https://thehill.com/opinion/4344490-mellman-the-problem-with-interpreting-poll-crosstabs/

So, no, your premise is not correct. Crosstabs are too volatile due to the limited sample size to be taken as anything but a very marginal foundation for a point.

[Wiz Imp]

10. That article doesn't show what you think it does

[republianmushroom]

8. This will upset the convicted felon and leave it's mark.