Question 1): Who should we trust? The Michigan electorate is bombarded weekly by a smorgasbord of polls on various political campaigns, from Governor and U.S. Senate on down. Which ones are credible, and which aren’t?
This past week, a top Lansing subscription newsletter (MIRS) backed away from publishing the results of a poll by a prominent Michigan pollster after the newsletter’s editor received major pushback from the campaign of Democratic U.S. Senate candidate Mallory McMorrow.
The poll, conducted by Mitchell Research & Communications, found McMorrow at just 6 percent ahead of the state’s pivotal Aug. 4 contest, far behind former public health official Abdul El-Sayed at 42 percent and U.S. Rep. Rep. Haley Stevens at 33 percent.
McMorrow has trailed the other two candidates in a number of recent public surveys, but 6 percent would mark a new low — a sign her campaign for the critical Michigan Senate seat may be mired in third place. McMorrow’s campaign told POLITICO’s Adam Wren that the polling methodology was faulty.
A McMorrow campaign spokesperson said in a statement that “Voicing concerns about a poll isn’t a pressure campaign. They (MIRS) chose not to publish a survey that is deeply flawed.”
POLITICO observes that the poll was conducted through a methodology known as text-to-web, in which random voters are selected to receive a text message link to a survey to fill out. That allows pollsters to ensure they are reaching an appropriately wide-ranging group of voters. But the McMorrow campaign said all respondents received the same open-access link, which would allow anyone with the link to take the poll — potentially multiple times.
The McMorrow flack said Mitchell’s poll “was conducted through an open SurveyMonkey link sent over text, meaning anyone who received this poll could vote multiple times or send the link to friends and supporters to impact the results. This is fundamental polling malpractice.”
The Mitchell poll also found that Michigan Secretary of State Jocelyn Benson holds an 8%-11% lead over all three top GOP candidates for Governor. In the U.S. Senate race, former Congressman Mike Rogers has a 4%-6% lead over Haley Stevens and McMorrow, but trails Abdul El-Sayed by 5% “according to a statewide poll of 827 likely voters in the November 3 general election for Governor and U.S. Senator conducted June 11-13.” The survey has a margin of error of + or – 3.41% at the 95% level of confidence.
So, what about this, anyway? Was Mitchell’s poll flawed? How does Mitchell compare with Michigan’s other four major pollsters, right now and over time?
Answer 1): Actually, it turns out Mitchell looks pretty good compared with its competitors, according to the two major evaluators of all polling firms nationwide. In fact, it’s been the MOST accurate pollster in Michigan based on the most recent election results studied in 2024.
Mitchell was ranked 13th most accurate pollster of 136 survey companies in the country that polled in 2024, according to ActiVote. Mitchell ranked higher than all Michigan-based pollsters and higher than the New York Times, Washington Post, ABC News, CNN, Quinnipiac, and YouGov.
For purposes of comparison, other Michigan polling firms were ranked as follows: Glengariff Group, 66th; Marketing Resource Group, 87th; and EPIC-MRA, 105th. A fifth outfit, Target Insyght, wasn’t ranked because apparently it didn’t conduct a sufficient number of surveys to be included in the list.
However, another evaluator, ABC News/538, covered a much larger number of polling firms and included Target Insyght. Again, Mitchell came out #1 in Michigan. Here are the ABC News’s findings:
- Mitchell Research & Communications — Ranked #59 (2.4 stars / numeric grade).
POLLSCORE: -0.2 | Transparency: 7.0 | Polls (total/time-weighted): 84 / 39.0
Solid mid-tier performer with decent historical accuracy.
- EPIC-MRA — Ranked #99 (1.9 stars / numeric grade).
POLLSCORE: -0.7 | Transparency: 2.2 | Polls: 65 / 21.3
Lower transparency hurt its score despite some Michigan-specific polling presence. - Glengariff Group — Ranked #175 (1.5 stars / numeric grade).
POLLSCORE: -0.2 | Transparency: 0.9 | Polls: 17 / 9.3
Lower-tier rating, limited polls factored in.
- MRG (Marketing Resource Group) — Ranked #242 (1.1 stars / numeric grade).
POLLSCORE: 0.6 | Transparency: 3.5 | Polls: 35 / 9.9
Below-average performance in the empirical metrics. - Target Insyght — Ranked #258 (1.0 stars / numeric grade).
POLLSCORE: 0.3 | Transparency: 1.7 | Polls: 7 / 5.7
Near the bottom of rated pollsters with very limited data.Steve Mitchell, the CEO of Mitchell Research & Communications, has mounted a vigorous defense of his most recent poll’s results and of his methodology, which he says have been the same he has always used.
Here’s how Mitchell describes the principle reasons for the U.S. Senate Democratic primary poll results that McMorrow & Co. found so offensive: “One of the reasons for (McMorrow’s) seeming collapse is the fact El-Sayed had received a large amount of unpaid media because of endorsements (that El-Sayed received) from Vermont Senator Bernie Sanders and Hasan Piker, the anti-semitic podcaster. Meanwhile, Haley Stevens had an outside organization spend more than $6 million on her candidacy,” Mitchell wrote. However, he added: “Our poll was conducted June 11-13, which coincided with an ad buy of at least $5 million on behalf of McMorrow that started just the day before we began our polling. Therefore, McMorrow’s ads did not have enough time to impact our results.”What’s more, Mitchell has gone beyond that simple statement and issued a press release “setting the record straight” in his own words, but no news outlet has printed it. TBR has it, however, and here it is:
FOR IMMEDIATE RELEASE: CONTACT: Steve Mitchell 248-891-2414Friday, June 18, 2022
10 Reasons Why Open-Link Polling is Secure
“There has been a great deal of misinformation about the methodology used by our company and others to collect data through SMS – Web polling.
Let me set the record straight.
Our company (and I presume others who do MMS – Web Polling) uses ten protocols to prevent fraud and to prevent someone from hijacking our survey and providing us with skewed information.
Let me walk through those protocols:
1. A random sample of 100,000 voters with cell phones is drawn proportionate by area, gender, age, race, and party affiliation to the likely turnout in the next election.
- The list is randomized so that if you take out smaller subsets, each will include the same demographics.
- When text messages are sent to cell phone users they’re generally sent out in batches of 20,000, or 30,000 texts at a time. Because they are randomized, each list is a valid subset of the total sample.
- Every time a batch is sent, we use a different link to a copy of the same survey.
- We normally deploy three, four or more different links to the survey.
- After we have completed, the survey will examine each of these links to determine whether or not the data collected is consistent.
- For example, if we see that two of the links, or three of the links, have John Smith leading Mary Jones by 4% to 6% we know that these are consistent and similar, and therefore the information is highly likely not to be skewed.
- If, however, one of the links has Mary Jones leading John Smith by 5%, the entire opposite of the other subsamples, then we presume that someone hijacked that link, and we don’t use that data.
- Instead, we will send out another batch of text messages using a different link.
- We limit the time in which the survey is open. If we send a survey out at 10:30 in the morning, we’ll close it up by two in the afternoon. After the data for the analysis is taken out, we may open it up again, but no new data is used in the analysis.
- SurveyMonkey has a number of proprietary safety guides to prevent this type of fraud or hijacking of data from occurring.
- SurveyMonkey integrates AI in several ways relevant to survey creation, analysis, and data quality control(including detecting bad or fraudulent responses). It’s particularly useful for polling/consulting work.
- AI for Fraud/Bad Data Detection in SurveyMonkey
- SurveyMonkey uses proprietary machine learning models as part of its Response Quality feature.
o It automatically scans responses for signs of low-quality or suspicious data, such as speeding (unnaturally fast completions), straight-lining (repetitive same answers), gibberish, profanity, or other red flags.
o You can then filter these out to protect your dataset.
- Once we have the information in Excel format, we remove all duplicate IP addresses to assure that no one has taken the survey more than once.10. Finally, if we examine data and if it looks at all suspicious, we simply do not use it.
We have had only two instances when someone has successfully gotten around the safeguards and hijacked a poll. We spotted it right away because of our protocols and simply did not use the data.
We did not become the 13th most accurate pollster of 136 companies ranked by ActiVote in the 2024 election cycle ignoring potential attempts to hijack a poll. Our protocols protect our data and help us achieve the accuracy and success we have had. We ranked higher than the NY Times, Washington Post, ABC News, CNN, Quinnipiac, SurveyUSA, Marist and others.
I hope this clarifies and answers questions regarding open-link surveys.”
*********************************************************
When a Democrat swears to uphold the Constitution, they are lying.
I don’t think the big Orange Liar-in-chief is a Democrat.
Interesting article on polling. I suspect the best way to find out which pollster is most accurate is to wait for the election outcome. But I do understand the fear that some candidates and voters will have if they see their candidate trailing that their voters will either stay home or go vote for someone else to stop someone they simply don’t like. But that is part of the fun of politics.
During my 4 decade career in Lansing as a chamber of commerce professional, I had the opportunity to work with a large number of both in-state and out-of-state pollsters. Here are my thoughts about this situation. First, every public opinion poll is a snapshot in time. Second, accurately polling the changing status primary elections is especially difficult. And last, but not least: it was my privilege to work closely with pollster Steve Mitchell on numerous projects over many years. Steve is a Michigander & person of integrity with a well-established track record of accuracy. I strongly believe the recent, politically motivated, criticism of his work is total BS!
“Susquehanna Polling and Research, Inc, rated the most accurate firm in the USA by Real Clear Politics for its polling in numerous battleground state in the 2020 election, and #2 in the nation in RCP’s composite ranking of pollsters from 2014-2022…..” published the following polling results of likely Democratic primary voters with respect to the upcoming 8/4/2026 Democratic U.S. Senate Michigan primary election shortly after the recent MIRS/ Mitchell Research polling results had been controversially released:
All likely voters polled: Abdul El-Sayed 22%, Haley Stevens 20%, Mallory McMorrow 9%
Likely female voters polled: Abdul El-Sayed 23%, Haley Stevens 18%, Mallory McMorrow 9%
Although not completely aligned with the earlier controversial Mitchell/MIRS poll results, Susquehanna’s polling data tends to corroborate that Dr. Elsayed is in the lead and McMorrow has taken a precipitous decline from the two leading candidates in the Democratic primary.
For anyone to take seriously the whining of a two-bit trollop like MM is beyond the pale. She chatters up a storm but utters not a word of substance. It is truly pathetic that the demoncrats could not find a better candidate to run for office, but their picks of late have been pathetic. By comparison, MM makes Kamala Harris appear to be intellectual. Imagine waiting fourteen months to make a false complaint against a fellow legislator for purported sexual harassment. A truly stale claim that turned out to be unfounded. She said in her book that she felt his “eyes assess every inch of my body and score me in his mind like a purebred at a dog show.” This conclusion cannot possibly be correct, as she is nowhere near purebred status. More like a typical ditzy dame seeking attention. When your campaign is faltering, blame the pollsters. Looks like the editor of MIRS needs to get a backbone and a dose of testosterone. He might even consider a rabies shot.
Kyle Melinn says
(Edit)It took a lot of backbone to do what was right, pal. Read the polls. Talk to a pollster. Do some research. Something wasn’t right with that poll. We don’t buckle under political pressure.
Pollsters know that they are only judged on their polls released close to an election date. They manipulate results of earlier, publicly released polls to influence pundits and prospective voters.