A new electoral divide has emerged in America.
This divide is not rooted in race, geography, age or education. Instead, it is engagement in democracy itself.
I thought this article had some interesting findings that are especially relevant after the latest polls from 538 and the Economist suggested that Trump had a slight edge this coming November. Since 2020:
-
Biden has mostly seen an increase in support from demographics typically associated with high-turnout voters such as those who are college educated, live in suburbia, women, and even some older and white voters.
-
Trump, in contrast, has generally seen an increase in demographics associated with low-turnout voters such as those with no college education and men. He has also slightly increased his support amongst black and Hispanic youths.
Hopefully, this means Trump’s increased support does not translate into actual votes on November 5th. For Trump to win another term would be a complete disaster not just for the United States, but also the world.
You should note that while polling shows Trump’s edge to be slight, historically he outperforms his polls by about 8 points.
So an apparent tie should be considered ‘leading’ for Trump when you consider that a significant portion of people that show up to vote for him are under represented by standard polling data. This 8% was observed in both the 2016 and 2020 results.
Yes. I usually give Trump the edge by at least 4 points, depending on the state/poll in question. He has an annoyingly consistent ability to beat the polls on election day.
He also demonstrates how popular populist, anti-establishment candidates are in the US. Even if he does lose in November, I fear he will have a lasting impact on US politics and it will not be good.
Here is a visualisation I made of this from a while back:
Interesting. Which polls were included in your study? Was it the 538 aggregate?
I ask because some polling groups have better methodologies than others. Personally, I tend to disregard those polls done by those interest groups or organizations which have a clear political alignment in favor of others that strive to be neutral. An example I like to give is this Huffington Post poll from 2016 which gave Hillary Clinton a 98% chance of winning; the Huffington Post is not a neutral source. (Note - I am not saying you did this, I am only asking because I am curious.)
I also think it’s important to note that the projections are probability-based, and many of the victories Trump had were potential oucomes predicted by probability-based polling and algorithms. They just were not the most likely outcome. Nonetheless, experience from the last two presidential elections shows the margin of error with Trump is significant and typically in his favor.
These two articles talk a little about polling methodology and how pollsters are working to be more accurate.
- Polling isn’t broken, but pollsters still face Trump-era challenges
- What polling does — and doesn’t — tell us
Now, I suppose there is always the possibility that we are just in a very unlucky timeline and that there are 498 (or however many) universes out there that have seen Trump thoroughly trounced as predicted. But that’s all hypothetical and therefore moot, and I do agree there is reason to be skeptical of the polling (even as they work to be more accurate.)
Yes, that data set is from the fivethirtyeight aggregate 2020 data set. Polling average of Oct/ Nov, and then I cant recall the electoral outcomes data set, I think maybe UC Santa Barbara stats dept.
I don’t use their weights, just a raw aggregate. It is matched however, so it’s local polling versus local results. I had to drop DC because it was such a an extreme example, but also because the numbers were so low.
On polling in general, I agree with some of the points you are making, by my primary thesis is that candidates like Trump and Bernie engage cohorts that are still typically not sampled in most polls. It should be telling that the US has such low voter engagement, that when you engage a 2-3 % slice of novel voters, you can easily disrupt polling if only 35% of people are voting.
The knife cuts both ways however, where if a candidate like Biden is disenfranchising even a small percentage of non serviced voters, he may poll higher but his performance won’t show it when it comes time to vote.This is exactly what we saw in 2020, where his polling overestimated his performance by an average of 4%.
So I don’t think of polling as broken, but rather, it’s not ever going to be able to show you the whole picture, if the demographics of likely voters have shifted.
my primary thesis is that candidates like Trump and Bernie engage cohorts that are still typically not sampled in most polls
I think this is exactly it. For whatever reason, their support is difficult to poll and they activate voters who normally wouldn’t vote and are not captured in polls running up to the election.
In Trump’s case, he is pulling in some very dedicated voters.
-
