The people behind the SDSU Poll, who claim they match the most sophisticated methods of conducting a survey are on Public Broadcasting today claiming that their poll shows the Governor’s race has a tight margin. Unfortunately, their claims seems a bit ..dubious at best:
The race over who will live in the Governor’s Mansion is tightening less than four weeks before the Nov. 8 election, according to a poll released from South Dakota State University on Tuesday.
Gov. Kristi Noem, who’s running for her second term, against Democratic challenger Rep. Jamie Smith is leading the race 45% to 41%, with a margin of error of 4%, according to the SDSU Poll, which is part of the School of American and Global Studies at the university in Brookings. The poll also states 14% of voters have not yet decided.
Read the entire story here.
I can’t help but find the claims of a margin of only 4% between the two candidates questionable. And let me explain why.
There hasn’t been much in terms of public polling being released about the race in general, which makes a lot of political junkies hot to hear what’s going on. But that doesn’t mean we should hoist the first carnival barker up as a confirmed truth-teller. I find wild claims about a close race pretty sketchy without more evidence. And I just don’t find it here.
The same SDSU poll throws a dart at another race in the State – the US Senate Race, where they make a claim of Thune leading in the US Senate race by 53% to 28%. Why do I question it? Again, it seems slanted ridiculously low against the Republican candidate.
In 2016, Senator Thune completely smashed Jay Williams on a 72%-28% vote. Yet 6 years later a poll is released showing the Senator down a significant amount from those prior results by nearly 20% against Brian Bengs, a candidate who Democrats barely acknowledge, and who has had only a negligible campaign presence which even Jay Williams had eclipsed 6 years ago by this time? I think they woefully underestimate the vote.
Especially since they give themselves an “out” by claiming “When undecided Republicans are ultimately forced to choose, negative partisanship makes it highly unlikely that they would vote for Democrat Brian Bengs.” Am I the only one who thinks this stated caveat is their way of claiming oops, “negative partisanship” is why we were way off is a bit of a cop out that they will fall back on when the actual results show a distinct difference?
When we fall back on results we do know, there seems to be a lot of difference to what information we’re gleaning from other places.
Earlier today, we heard from the Morning Consult group about how popular Governor Noem is in South Dakota, with a job approval rate of 57%. So when a poll being conducted about the same time makes the claim that she’s only up from her Democrat opponent by 4 points, you can’t help but to question it.
In a little insider information I’m hearing, Voter ID calls to non-Republican recipients are showing hard percentages in favor of Governor Noem which exceed anything we’re hearing in the SDSU poll. Add that to the overwhelming support she enjoys from Republican voters, and while everyone anticipates the race will narrow, it doesn’t resemble anything that correlates with the SDSU numbers, even beyond their caveat that they “expect that a majority of these undecided respondents will break toward the governor.”
Why are these numbers so weird? Well, one thing you have to look at is the SDSU Poll’s methodology.
To do the polls, the researchers purchased a data file with the addresses of every registered voter in South Dakota from the Secretary of State. They randomly choose from that list and send letters inviting voters to go to the Question Pro website and enter the identification number listed in the letter to take the survey online.
“We are the only poll (in the state) that uses this particular method,” said Wiltse, pointing out most polls are telephone surveys.
Read that here.
They’re basing these results on an on-line survey? Given who uses the internet, I have a strong feeling that these results skew very young. And that’s going to have a big effect. According to the US Census..
Voter turnout also increased as age, educational attainment and income increased. Voter turnout was highest among those ages 65 to 74 at 76.0%, while the percentage was lowest among those ages 18 to 24 at 51.4%.
Read that here.
A poll using a response mechanism that’s most likely to be used by the youngest group (who votes the least) and unused by the oldest group (who votes the most) is skewing to the left? How many seniors do you know who respond to on-line surveys? As I inexorably crawl through my 50’s, I’m not even nice to telephone solicitors anymore, much less finding myself wanting to waste my time with an on-line survey. The methodology alone raises significant doubt in my mind about the accuracy this poll to gauge election results.
Sorry, but I’m not buying what they’re selling. Because a lot of it seems tenuous, especially since all of their results come with a setup where they can shift the blame to “negative partisanship.”
