
If you have read my articles before, you know I enjoy digging into polls. Today I want to revisit some popular myths about polls, and focus in particular on one especially amusing clown show.
Let’s start with the big myth you see promoted all the time in media: That the polls tell us who is winning the election at a certain time.
The problem with that notion, is that a candidate wins an election only if and when he/she collects the needed votes to win. Mid-to-late summer is campaign time, which is important for attracting support, but no, no one has even started to win the election just because a poll comes out saying someone is ahead. A really glaring example would be Joe Biden. For years since he took office, Biden was the focus of polling discussing his re-election chances. All the way through the winter and spring and early summer this year, Joe Biden assured everyone he was running for re-election, he would win a second term, and no one should doubt him.
Then on July 21, suddenly Biden was old news, and gone. And all the polling about Biden was worthless, completely irrelevant despite literally hundreds of polls covering him.
The media pivoted to a new fib – that Vice-Resident (you have to do something for me to call you President or Vice-President) Harris had surged from behind to lead Trump in the polls. But if you have been paying attention to the details, you will know that Harris enjoyed the benefit of being the new shiny object for the media, and the top Democrat running always gets additional points in media support. While Harris is losing some of that support now that the media rush is fading, she did not score any points, no matter what the headlines say in the polls.
Some readers have said that polls are worthless, garbage, but that’s not completely the case. Properly done, a poll is like a bathroom scale, in the say that you may weigh different on someone else’s scale than you do at home on yours, but if you use the same scale every day, you’ll know from the reading whether or not your diet is working. In the same way, a consistent poll, even when biased, will reveal changes in public mood and focus on certain issues and candidates. Even though I disagree with some of their weighting, for example, I appreciate the media polls which include the crosstabs and show us how many people are counted for each demographic group. That includes YouGov, which always goes to the trouble of publishing every question, some of the their demographics and showing how each group responded to the questions. Ironically however, YouGov also made an incredibly foolish mistake and exposed their hypocrisy beyond doubt, but I will come back to that in the last part of this article. For here, I will affirm that most political polls are trying to give the truth of how the race is going, as they see it, and they work hard to be consistent in that effort. That’s the public service I mentioned in the first part of the headline.
The second part, about con games, requires me to discuss again how polls are designed to work in election years.
In modern elections, there are basically four types of political polls:
- Internal polls taken by candidates’ campaigns
- Push polls meant to sell a preferred candidate
- Persuade polls meant to promote a position or issue, and
- Polls meant to promote the agency reputation
Internal polls, despite what you may hear in the media, are not taken to see if a candidate is winning, but are commonly used to help target opportunities and test ideas and tactics. While selected data may be released to the media to promote an advantage for a candidate, internal polls are used to develop campaign plans, which is one reason they are kept close to the campaign.
Push polls are a common part of polling nowadays. They exist to help sell a candidate, and are commonly created to help drive support in a close race. Clues to push polls can be found when the poll fiddles with demographics and weighting to make a poll appear to have more voter support than actually exists.
Persuade polls are similar to push polls, but exist not so much to promote a candidate as to generate interest and support in selected topics and issues. Polls from 2021 to now which obsessed on the January 6 riots, the move to suppress selected individuals or groups for ‘Hate Speech’, and the artificial anger whipped up against the Supreme Court are examples of groups working in long-term efforts to shape voter opinion ahead of future elections.
Finally, all polls need to establish and maintain a certain level of credibility to stay in business. That means that their final polls need to be somewhat close to the actual results, and they should not be wild outliers relative to other polls. This is why some polls can be absurd at times, but completely reasonable at others.
And then there are polls like the most recent offer from YouGov. I want to start by complimenting YouGov, despite it’s left-lean, because YouGov usually shows its demographics and crosstabs unlike many other polls.
That said, YouGov lost my respect when it started using different weightings in different polls, trying to create the impression of moving voter mood which was not real. That’s fraud by any reasonable definition.
But then there is what YouGov did this week.
If you look up the polls, you will see a poll presented by The Economist this week, as well as one presented by Yahoo. The thing is, neither The Economist nor Yahoo actually does any polling (just as NPR, The New York Times, The Washington Post, ABC, CBS, NBC. FOX News and so on do not actually do the polling they put out under their logo). Both The Economist and Yahoo used YouGov to do their polling for them, and the two polls came out just a day apart from each other. Cutting to the chase, YouGov used the same staff to contact basically the same respondent pool, asked similar questions, and then did some cherry-picking when deciding what to package and send out to the public.
With that in mind, let’s take a look at what YouGov packaged in those polls. First, a look at their published demographics:
Males: 713 Economist (45.9%), 870 Yahoo (48.7%), historical norm* 47.2%
Females: 839 Economist (54.1%) 918 Yahoo (51.3%) norm 52.8%
Whites: 1053 Economist (73.9%), 1123 Yahoo (68.8%), norm 73.7%
Blacks: 181 Economist (12.7%), 223 Yahoo (13.7%), norm 12.0%
Hispanics: 191 Economist (13.4%), 286 Yahoo (17.5%), norm 9.7%
Asians: none reported in Economist or Yahoo, norm is 2.8%
18-29 yo: 301 Economist (19.4%), 375 Yahoo (21.0%), norm 17.8%
30-44 yo: 344 Economist (22.2%), 451 Yahoo (25.2%), norm 29.2%
45-64 yo: 543 Economist (35.0%), 583 Yahoo (32.6%), norm 35.0%
65+ yo: 364 Economist (23.5%), 379 yahoo (21.2%), norm 18.0%
Democrats: 546 Economist (35.2%), 585 Yahoo (37.3%), norm 37.7%
Independents: 594 Economist (38.3%), 465 Yahoo (29.7%), norm 27.8%
Republicans: 412 Economist (26.5%), 518 Yahoo (33.0%), norm 34.2%
(historical norm is the average from 2000-2020 Presidential Election Exit Polls by the New York Times and CNN)
The Economist overvalued Females, Hispanics, 65+ year olds, and Independents and undervalued Republicans. Yahoo overvalued Males, Hispanics, 18-29 year old, and undervalued Whites.
But it’s also interesting to note the demographic categories not mentioned by either poll, including Married/Unmarried, Urban/Suburban/Rural breakdowns and geographical areas (Northeast, South, Rust Belt, West, Southwest, etc.), and Veterans. These are all demographic categories commonly covered in polls, and YouGov has published such categories in the past, but not now.
It should be interesting, for example, to consider whether Married Females have similar support for Harris right now as they did for Biden in 2020, or whether Harris has lost a piece of their support, as Clinton did in 2016. We’ve heard a lot about how Arizona may not be secure for Trump, for another example, so it would be interesting to see if Trump has lost some support in the Southwest, but YouGov left that part out.
In the end, it’s important to understand that political polling is not a hard science nearly so much as a recipe. Polling groups are paid to produce a report that does what the buyer/sponsor wants, so if the first draft does not say what they want, the poll group will just add more responses or fiddle a bit with weightings until it gets a report that is acceptable to the sponsor. If the sponsor wants the poll to simply say how things are going, that’s useful to understanding the race, but some polls exist to sell a story, not share the facts, something that needs to be kept in mind during an election season. YouGov is selling soup, not facts.


