(C) Daily Kos
This story was originally published by Daily Kos and is unaltered.
. . . . . . . . . .



You should Trust Polls. Including ones showing Biden behind or close to Trump [1]

['This Content Is Not Subject To Review Daily Kos Staff Prior To Publication.']

Date: 2023-09-20

There have been two recent blog posts about polls. Ignoring Polls and Explain this Polling, the second is where I borrowed the title image from. Both posts and comments for them suggested ignoring the polls, that the polls were off, the Biden had the advantage.

Doing these things: Bad idea. Really, really bad idea. Unskewing the polls in general is a terrible idea.

Which isn’t to say polls are god, no one can claim that. Some show Biden decently ahead, some show tied or close to it. I don’t have a full list, I’m not watching them too closely. But they aren’t likely all that off, and if they are wrong it could mean Trump is ahead.

The Recent Record of Polls

2016

The Year Polling failed. Hillary Clinton predicted to win, sometimes with high percentages by some aggregators, and instead we got Trump. Lots of people disappointed, shocked, lots of not so high quality people excited. Three midwestern states expected to go blue instead went narrowly red. This was, no question, a big polling failure.

2018

There are some polling failures here, Gillum generally polled ahead of DeSantis, only to narrowly lose a happens in many Florida elections. Bredesen in Tennessee and Donnelly in Indiana were expected to be competitive in many polls, turned out not to be close. But generally, polls did well. They predicted the house margin well, predicted the senate margin in most races, same with governors.

2020

Like 2016, the polls biased democratic, in a lot of the same ways. Seemingly the Midwest has a lot of hidden Trump voters, as do other places. In senate races, Gideon vs. Collins was a surprise Collins win, and the house was much closer than expected.

2021

Polls predicted MacAuliffe pretty well. Said governors race would be a narrow Republican win, that’s what it was.

2022

This is interesting. If you look at 538 and I think other aggregator sites, they did get things roughly accurate. Look at the details, not as much. Their senate prediction was too Republican, as were a number of individual races. House was a bit closer.

Did the polls screw up again? Nope. There was in fact controversy over this, 538 was including a number of polls created to deliberately skew aggregators. Here is a link. The time tested, honest polls got the race right: Fetterman comfortably wins, Narrow Republican overall, states vote about as expected. A bunch of questionable Republican polls, with issues spotted at the time, told a different story. Which sites like 538 and RealClearPolitics included, distorting their results.

2024 Outlook

So that’s two off years when Trump was running, and two accurate years, plus the generally accurate, within reason, years before 2016. What does it say for 2024?

-If you look in general, polls were accurate in non-presidential years, underestimated Republicans in presidential years.

-There have been special elections in the past year. These have leaned extremely democratic, and people use these as an argument the election is solid, that the polls are overestimating Trump. Which might be the case, but such a miss hasn’t happened before.

-Pollsters presumably have gotten more accurate at figuring out how to account for missing voters over time. Hidden Trump voters were a surprise in 2016, they are not these days. Pollsters will also know about Cell Phones and not answering, and will be gathering information on how to compensate for these. Remember, these surveys are not just used for news, they are used for political campaigns, marketing, and other purposes. People will want to know how to get them right.

If I had to guess, I’d say…..probably accurate, Biden and Trump probably are about tied if the election were happening in a few months and people knew about it. That is, polls are probably accurate. This is a gut reaction, take it as you will. The entire political world just seems tired, worn out, low energy. So Trump’s hidden voters raging at the world, plus the activists who would turn out voters for democrats, will both lack the same juice that they might otherwise have. So most likely anything that interferes with pollsters getting good answers evens out, and more likely than not they are about accurate.

And if they aren’t accurate, they could equally be in either direction. Biden actually ahead by 5 or 6 or so, or behind bu 3 or 4 or so are entirely possible.

How do Polls Work? Why do they miss?

A poll works by picking a subset of people, members of a larger group. They ask questions, and reporting the answers. With some work in the middle. For elections, this means taking a subset of voters, and reporting the answer. Surveys of other sorts do the same thing. I’ve used data from the American Community Survey and Health surveys for analysis, these ask a subset of everyone in the scountry. American community survey is for demographics: income, where they live, race, education, family size, things like that. Health surveys ask about exercise, foods eaten, vaccines, health status, and more.

If you randomly surveyed enough people, your results would closely match the larger group you were interested in. And it takes a surprisingly small amount of people asked to get close results, if the people questioned are truly random. You’d closely heath election results, population of areas, income distribution, amount of people who smoke, or whatever else of interest you wanted to know.

Unfortunately, random sampling isn’t perfect. So pollsters need to do some extra work. Sometimes subgroups are really small. To get consistent, accurate results, you need groups of a certain minimum size. So surveys will sometimes deliberate ask more people in a particular group than is proportional to get accurate answers, called oversampling. In a presidential election, a poll might do this for, say, Native Americans since random sampling would get so few responses, if they were interested in how that group was voting.

Weird results for smaller groups is why looking at crosstabs is often a bad idea. Smaller groups might have gotten odd results, but these cancel out in the total result.

Sometimes, one group responds more than, or less than, proportionally to the larger group. Pollsters can compensate for this using other information, like demographics. So if black voters picked up the phone less than expected, pollsters would see that the population answering questions is less black than the census or ACS or other data show that voters should be, and overweight the black responses to get a properly portioned sample.

Some polling misses are randomness. Randomly sample enough populations, and sometimes odd things happen. Polls are never exact, sometimes they are very, very inexact.

Polls missing in 2016 and 2020 is another problem: a nonrepresentative sample the pollster does not know is nonrepresentative. They make a prediction to the best of their ability, but some people were missed and not properly weighted, they vote, and the election result is far different than expected. In 2016 and 2020, these were shy Trump voters. Most likely, they demographically matched people answering the pollsters, but have personality differences or life history differences that pollsters never needed to check before, and voted accordingly.

Another problem is likelihood to vote. Polls report “registered voters” and “likely voters”. These phrases are vague, actual likelihood to vote needs to be a fixed number to be useful. Pollsters can estimate this, but it will be inexact. And different motivations may lead to different likelihoods of voting.

For 2024 polls to be off, there must be a population of voters pollsters missed. For them to be off in Biden’s direction, they have to be more democratic than expected. People arguing special elections or abortion assume these voters aren’t picked up by pollsters, but there’s no preexisting reason to assume this. And shy Trump voters could equally still exist.

Why a Dem Leaning site has people unskewing polls

Most of you laughed at Romney supporters in 2012 doing it, I’m sure. In fact, polls underestimated Obama. Now you yourselves are questioning the polls and arguing they are wrong. Why do this, and what to do in response?

Forced Positivity

This Comment gets it right. Its a trend on this site. Biden and Harris are the greatest leaders ever (exaggerated for effect.) Republican are all bad. Its a heavy emotional investment. And if Biden is the greatest, they must be winning, if they aren’t…….no, no, can’t be possible, no way. So polls must be wrong.

Distrust of Numbers and Knowledge and Expertise

Antivax and anti mask: most of you agree these are bad. You shouldn;t ignore disease like that. But its not just COVID, or other conservative bugbears like global warming being fake. Its a general trend, to view expertise, data, research as an enemy if it doesn’t tell you what you want to hear. Moderates do this as well, or uncommitted people. How often are police funding cuts blamed for crime increases, even though such cuts often didn’t happen, and crime changes aren’t often associated with police funding, or different types of prosecutors anyway? I remember back in 2019 simply mentioning to a family member that universal health care problems tend to be less expensive. I’ve studied health economics at the PhD level, done some research, I’d know this better than this person. But they still got tensed up and stressed at the idea. Also argue with me over things I similarly would know better than they would.

So, expertise, data, analysis, lots of anger at it. Polls are numbers and data, they are no exception. Republicans insist that polls must be off, even an election in 2020. And people who support Biden are doing the same thing.

The way social media talked about polls even before this year showed laziness at understanding how they work. Lots of cherry picking a single poll for optimism, choosing to believe one and not another based on the results you want, that sort of thing. Its not hard to read polls: look at the aggregate and range of results as a guide, and assume anything close is random and could go either way. Close enough to get involved in the election if you have time.

Tiredness and Laziness

If polls are anywhere near close, and you want dems to win, you should get off your ass and start talking to people. Go to work on them. That’s the implication, even if they consistently showed something like Biden +3. But politics in general seems burnt out and tired at the moment, easier to just assume the best and not bother with anything.

The tiredness is probably why we have Trump vs. Biden again. Choosing a default option, not bothering to check if someone better could do it. And getting angry if asked to look at a different choice, as I’ve seen on the democratic side at least.

Hatred of Progressives

I wrote about the reactions to Elizabeth Warren in my first post on this site. They were often angry, abusive, dismissive when she ran for president or argued for a policy. Same goes for others, squad members arguing for things or running for office or supporting others, new prosecutors, congressional primaries. Moderates were angry, dismissive, excited if they didn’t win, blaming in general elections. Was….not sure what’s allowed in this site. Bad words, whichever ones you like. I was not expecting this kind of anger and abuse going in, certainly not from people who professed to support the goals these people were running on.

I’m writing a couple posts about why people hate progressives. was supposed to come out last Saturday, its delayed, in the process of being written along with some other ones on this site and some personal stuff elsewhere. Will be out when it is out, hopefully this weekend along with the others.

Over time, moderates got a lot of what you wanted. No big healthcare changes, just obamacare with some tweaks. No supreme court expansion. Filibuster in place, no voting rights expansion and reduced gerrymandering. Lots of police money. One global warming thing, which still has lots of emissions coming out and the problem getting worse. No big childcare program. Little things like the post office and other departments having Trump appointees still there. The very slow Garland now Smith investigation.

And in return, the candidate, the biggest moderate of 2020 is….losing. According to many polls. Or barely winning nationally, which could be a loss depending on the electoral college.

I’m sure if Biden wins by a lot, we’ll hear that moderates were right all along, if you want progressive policies tough luck, no medicare for all or public option or anything, Palestinians are screwed, no court expansion, the works. If Biden loses or there’s a narrow win, it was progressive’s fault and moderates were right along, and its the first group’s fault that Palestinians are getting screwed, no one has health care, the court is more Republican, out of control Global Warming, that’s how this sort of blame works. None of which deals with current polling issues right now.

How to Actually Respond to these polls

Assume the polls are accurate, and if they are close enough small error is a lost election, go to work. Which does not mean blaming progressives or asking if they let the perfect be the enemy of the good or any of that nonsense. I saw one post on here, forget which one or where to look, where the writer did this, than brought up how Manchin or moderates or somesuch are helping Republicans. Which tells you where the problem lies: talk to interested groups, older peopl,e people voting Republican, get them to do differently.

Simple as that.

Explaining some polls

The Harris poll at the top was a bit weird, I agree. Biden polls are also close, and often show differing Republican candidates ahead or behind. I am a person with opinions and a little knowledge, not a marketing or public opinion firm: these are best guesses as to where patterns come from.

-How Trump is close or ahead: Lots of people like the guy. For some reason, the guy does nothing for me. I can provide explanations, guesses, that upcoming “why people hate progressives” post will have some. But you still have to trust election results, whatever the reasons are, a lot of people like Trump. The election will be close no matter what.

--Republicans are committed: Conservatives make being conservative an identity. Its not, its a choice a person makes, but the ones doing it are invested and committed.

-Other Republicans: Sometimes these are ahead, sometimes behind. To me, it look like first impressions, what’s in the air, what media is saying. DeSantis was rated a lot higher in polls earlier this year, than the finger eating thing and stories about him being unlikable came out, now he polls worse than Trump from what I’ve seen. Less well known candidates

-Why Biden might have less support: Could be a few reasons. His approval started around 50’s to 40’s, now is a lot less. Fill in your explanation why that is. Some mix of less educated people shifting Repuublican and the lack of much passed by democrats (voting rights court reform, the works), doesn’t help. Abortion is supposed to be a motivator, but the fact it was banned a lot of places, along with slow Trump prosecutions, adds to tiredness apathy, etc.

--Biden age: Yes, Trump at 215 pounds and an ability to slowly walk down stairs isn’t in as good shape as the bicycling Biden. Yet, a lot of people think the opposite, or say the opposite. It is weird, Trump gets excused a lot. Chalk some up to excusing things Trump does, maybe the fact that he talks fast in interviews, previous impressions of Trump. or media repetition.

-Harris vs. Biden: In most polls, Harris does worse than Biden, at least that I’ve seen. Harris….I’ll write about her in one of that block of posts I mentioned. Speculating here would make that post awkward. I’ll stick with saying Harris doesn’t seem well liked for a few reasons.

Thing about these is: You have to take them seriously. Not comment about how they are obviously wrong, not lecture progressives or younger people about how Biden is obviously the best option and everyone must show up. Not repeat that Harris and Biden are the best politicians ever and no one else could do better. Instead, actually engage with, do work on, moderates. And older people. And people who aren’t paying much attention. And maybe activists will do the work for you, but they might be burned out, or already doing said work. Maybe show some appreciation and thanks, in a meaningful way like listening and supporting better politicians and policies.

Face the polls head on, solve the problem. Take the rewarding route, not the easy one.

[END]
---
[1] Url: https://www.dailykos.com/stories/2023/9/20/2194332/-You-should-Trust-Polls-Including-ones-showing-Biden-behind-or-close-to-Trump

Published and (C) by Daily Kos
Content appears here under this condition or license: Site content may be used for any purpose without permission unless otherwise specified.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/dailykos/