Political Smokeout

Fun with crosstabs in PPP’s poll

The new poll from Public Policy Polling includes a public release of voluminous cross-tabs, so you can see how each question broke down by various demographic categories.

Among the fascinating things you can find by digging in (and keep in mind that some of these questions have fairly large margins of error):

• Mike Rounds leads Kristi Noem 43-39 among all Republican primary voters. But he’s up 60-14 among the 9 percent of South Dakota Republicans who voted for Obama. Among Romney-voting Republicans, it’s just 43-42 Rounds.
• There’s no big gender gap in a Noem-Rounds primary matchup — Rounds leads 41-36 among women, 45-42 among men — though more women are undecided.
• Conventional wisdom is Stephanie Herseth Sandlin has a problem with her base as a result of votes against liberal priorities like the Patient Protection and Affordable Care Act. This poll, at least, doesn’t show that. Among Democrats, Tim Johnson’s approval rating is 77-17, while Democrats approve of Herseth Sandlin 82-13. The two are basically identical among approval by people who voted for Obama in 2012 — 85-11 for Tim Johnson, 86-10 for Herseth Sandlin.
• 21 percent of South Dakota Democrats disapprove of Barack Obama. 11 percent of South Dakota Republicans approve of the president.
• Obama’s approval among white South Dakotans is 37 percent. Among non-white South Dakotans (7 percent of the sample), it’s 58 percent.
• This is probably insignificant because we’re talking about the preferences of about 20 to 25 people here in the polling sample, but among non-white Republicans, Noem has a huge 61-8 lead over Rounds in a potential primary.
• No surprise that Obama’s approval rating is highest among young South Dakotans, aged 18-29 — 50 percent approve of the president, versus 45 percent who disapprove. Less predictable? Obama’s second-best age category are people over 65, where 42 percent approve and 52 percent don’t. The most anti-Obama age group are people aged 30 to 45, among whom just 28 percent approve and 66 percent disapprove.
• In light of those numbers, I looked at the Rounds-Noem primary matchup by age. Rounds beats Noem among 18 to 29-year-old Republicans (44-39) and Republicans  65 and older (43-38), but is tied among those 30-45.
• Unfortunately for my curiosity, PPP didn’t poll about how Herseth Sandlin would fare in a primary challenge to Tim Johnson. That is very unlikely to happen, but wouldn’t the poll result be fascinating?

Again, you can view these crosstabs yourself here.

Poll: Rounds strong, Brendan Johnson unknown

Ah, sweet manna: We’ve got public polling for the South Dakota 2014 race.

Polling firm PPP, which is Democratic-affiliated but has a generally good track record, has polled 1,069 South Dakota registered voters about Mike Rounds, Kristi Noem, Tim Johnson, Stephanie Herseth Sandlin, Brendan Johnson and 2014.

Read their release here, and the full results here.

Key findings:

• Mike Rounds is popular. He’s got a 51-34 approval rating (which is actually lower than I’d expect), and in hypothetical 2014 matchups beats Tim Johnson 52-41, Herseth Sandlin 49-44, and Brendan Johnson 53-32.
• Tim Johnson is no longer popular, although he’s not exactly unpopular. About as many respondents approved of him as disapproved (44-45, though this question asked about his “job performance” and not about general opinions about him as it did for all the others), and he would lose to both Rounds and Noem in hypothetical matchups — 51-34 to Rounds, 49-45 to Noem.
• No one knows who Brendan Johnson is. Only 42 percent know enough about him to have an opinion; of those who do know enough, 25 percent are favorable and 17 percent unfavorable. So take these results with a big grain of salt — if Brendan Johnson campaigns people start to know who he is, these numbers will change quickly.
• Herseth Sandlin could be poised for a comeback. She’s more popular than either Johnson (with the above caveat, plus the additional warning that as someone who’s been out of office for a while, she’s less polarizing than she was a few years ago), with a 52-37 favorability rating. Herseth Sandlin loses narrowly to Rounds 49-44, and is basically tied with Noem, 48-47.
• Kristi Noem’s not as popular as Rounds, but she’s in a decent position — especially among Republicans. Noem’s got a 49-42 favorability, and would beat Tim Johnson 49-45 and Brendan Johnson 49-37, while being a point behind Herseth Sandlin 48-47. Though she’s worse than Rounds in every matchup, Noem trails Rounds just 43-49 with 17 percent undecided among Republican primary voters.
• Possibly because Brendan Johnson is so unknown, this poll shows Democratic voters preferring Herseth Sandlin 68-16.

Again, because this is so far out, before anyone has really campaigned, all this poll really shows is the starting point. These numbers will change a lot as various people get into the race and start meeting voters and running ads.

How accurate is this poll? PPP released their demographics. Their sample was 49 percent Republican, 37 percent Democrat and 15 percent Independent or other party. Actual registration is 46 percent, 36 percent, 18 percent. They asked voters who they picked in the 2012 election and got Mitt Romney 55 percent, Barack Obama 38 percent. Actual results were 58 percent and 40 percent.

Obama’s approval rating among these South Dakota voters was 38-57.

What’s the takeaway from this? Mike Rounds is in a good position, but hasn’t put this election away yet. Tim Johnson would face an uphill battle if her pursued reelection, and Stephanie Herseth Sandlin might be a better candidate for Democrats than the incumbent senator. As for Johnson’s son Brendan, it’s too early to say because most voters don’t know anything about him. That relative anonymity also will make the initial stages of a Brendan Johnson campaign very important — he’s ripe to be defined, and it remains to be seen whether he or Republicans will forge his public image.

And, again, remember that these are very preliminary numbers.

If you want to delve into the crosstabs, click here.

Who’s polling the legislative session?

A Twitter user is reporting she was polled yesterday about the upcoming South Dakota legislative session.

According to user @book3mom, the call lasted about five minutes, was conducted by a live interviewer who was from out of state (key tell: they mispronounced “Pierre”) and didn’t include any disclaimer about who conducted the poll.

Among the things she reports being asked:

• how she ranked the importance of various social and fiscal issues for the legislative session
• her opinion of President Barack Obama, Gov. Dennis Daugaard, Secretary of State Jason Gant, the state Democratic and Republican parties, and the tea party movement
• whether South Dakota should let the federal government run its health insurance exchange under the Affordable Care Act (but not whether South Dakota should expand Medicaid)
• whether she thought Pierre too partisan
• if she liked Nebraska’s system of nonpartisan legislative elections

The poll came from 503-575-7964 and came up as “Bernett R” on caller ID. A quick web search determines that number is tied to Bernett Research, a call center firm; that doesn’t identify who hired Bernett to conduct the poll.

Does anyone have any insight into who’s behind this?

Grading the Nielson Brothers

The only public polling released in South Dakota was from Nielson Brothers Polling, a controversial outfit that has pegged some races very closely but has a few misses, including an enormously far off call in the 2010 gubernatorial race.

This year, NBP released a poll on the eve of the election, interviewing more than 600 likely voters from Oct. 31 to Nov. 4. How did they do?

Well, if you go by just calling races correctly, they went 6-for-7:

But there’s more to polling than just calling the winners correctly. If you predict a race will be a 20-point win and your pick wins by just 1 point, you really blew the call, even though your poll had the winner on top.

So here’s Nielson’s polling by vote shares:

Some of these misses are pretty big, but a lot of them are surprisingly small given how big the undecided numbers were in some races. For example, Yes on 14 had 30 percent undecided; if you split that down the middle and add the totals to each side, you end up with something pretty close to the result. So here’s a look at the margins:

On four races, NBP were pretty accurate. Their predicted margins were off from the actual margins by less than 4 percentage points — or less than the poll’s smallest margin of error, 3.8 percent.

Two more were off moderately. They underestimated the share of the vote Romney would receive, and underestimated the annihilation RL16 would receive.

And then there’s the six-year PUC race, where NBP completely missed the boat, calling a dead heat when Fiegen won by 13.8 points. NBP’s numbers for McGovern and Clarke were very close to the final results — but Fiegen’s support was almost 13 points lower, greater than the number of undecideds.

On the candidate races, where NBP missed, it generally missed in favor of the Democrats. They thought Varilek, Nemec and Obama would lose somewhat less badly than they actually did, though in all three cases NBP predicted convincing victories by Republicans.

They underestimated the margin of defeat for all three ballot questions.

All that said, for downballot questions, voters’ opinions tend to be malleable and not very firm, so you’d expect less accurate races the less attention a race receives.

Sometimes polls are off consistently in a single direction. This is called, statistically, “bias.” If there is a consistent bias from a pollster, you can account for that. People like Nate Silver do this, identifying which polls generally show good results for one party or another and weighting their results to account for that.

There may be some of that with NBP. (National forecasters identify them as a Democratic-leaning polling firm, though the Nielson’s don’t voluntarily identify themselves that way.) But there’s also a second issue, of inconsistency, which can’t be so easily controlled for. In 2010, they nailed the U.S. House race in South Dakota while missing the boat on the gubernatorial race. In 2012, they got almost everything right but were way off on one race.

Getting most of South Dakota’s races right, and pretty closely in many cases, reflects well on the Nielson Brothers. The results of this election will probably bolster NBP’s credibility. But it won’t silence their critics or answer all the doubts about their accuracy, given a persistent erraticness.

Final Nielson Brothers numbers

The only public pollster in South Dakota this election cycle has released one more round of numbers.

The headline number is a dead heat between Matt McGovern and Kristie Fiegen for PUC, with McGovern drawing 41.9 percent and Fiegan at 41.6 percent. Libertarian Russell Clarke draws 6.7 percent and 9.8 percent are undecided.

Rep. Kristi Noem looks safe, up on Matt Varilek 53.9 percent to 41.2 percent with 4.9 percent undecided.

Chris Nelson is destroying Nick Nemec, 59.2 percent to 27.5 percent with 13.2 percent undecided.

If NBP is right, all three ballot measures may be headed for failure. According to this poll:

• Referred Law 14 has 27.6 percent support and 42.4 percent opposition, with 30 percent undecided
• Initiated Measure 15 has 38.9 percent support and 49 percent opposition with 12.1 percent undecided
• Referred Law 16 has 28.2 percent support and 53.8 percent opposition with 18.1 percent undecided

If that’s what actually happens on Election Day, RL14 and IM15 could lose despite only minimal efforts made by the opposition camps.

Tomorrow, we’ll know how accurate the Nielson Brothers’ final poll is.

The poll had around 650 respondents with a margin of error around 4 percent, and was conducted via interactive voice response (robopolling).

NBP: Referred laws in trouble, divided IM15

Some interesting news from the Nielson Brothers on the ballot measures. Again, the usual caveats about these pollsters apply; we can use these last-minute polls to judge the firm’s accuracy come Election Day.

NBP found:

• Initiated Measure 15’s support is within the margin of error over the opposition, with 43.3 percent in favor and 42.8 percent against. Compared to NBP’s past polls, support for this tax increase has plunged — during a time when Moving South Dakota Forward has dominated the airwaves. If this is accurate, we could be looking at a new benchmark for wasted political money in South Dakota.
• Referred Law 14 is trailing significantly: 24.1 percent say they support the law, 39.8 percent say they’re opposed. 36.1 percent are undecided. This is another measure where the side dominating TV ads is losing; the Democrats opposing RL14 didn’t even bother to form a ballot committee.
• Referred Law 16 is getting trounced. 24.6 percent support the measure, and a full 51.6 percent oppose it. 23.8 percent are undecided.

Depending on how things go, Nov. 6 could be a very bad day for Gov. Dennis Daugaard. If this poll is at all accurate, two of his signature policy initiatives could be in deep, deep trouble.

Nielson Brothers: McGovern winning, Varilek losing

We’re a few days out from the election, so this is the poll to judge the Nielson Brothers by once we get results on Nov. 6. Will they nail the race? Or will they be way off?

A poll just hitting my inbox, conducted from Oct. 28 to Oct. 31, shows mixed news for Democrats in South Dakota.

Their key findings:

• Matt McGovern leads his PUC race with incumbent Kristie Fiegen, 45-39, with Libertarian Russell Clarke drawing 5 percent and 11 percent undecided.
• Kristi Noem leads Matt Varilek 50-44 with 6 percent undecided.
• Chris Nelson trounces Nick Nemec 58-28 with 14 percent undecided.
• Barack Obama’s job approval in South Dakota is 47 percent, with 53 percent disapproving.
• Kristi Noem’s job approval is 55 percent, with 45 percent disapproving
• 46 percent of voters say they plan to vote for the Republican state senate candidate, while 35 percent say they’ll vote for a Democrat and 19 percent are undecided

Obama’s approval rating seems really high here. On Aug. 1, Gallup showed 38 percent of South Dakotans approved of Obama. If you were to assume that Gallup poll reflects how South Dakotans currently feel about Obama (a big assumption), that would suggest the NBP sample is overweighted with Democrats, and would show more support for Democratic candidates.

That sample is about 650 (between 634 and 671) likely voters with a margin of error of around plus or minus 3.8 percent.

But in terms of order, this poll seems plausible — that McGovern would be the best-performing Democrat, Varilek behind but not embarrassing himself and Nemec being blown out of the water.

A strong McGovern candidacy could explain why the South Dakota Republican Party went negative on him today. 

Plausible, of course, is not “accurate.” It’d be great if we had another public poll to rely on.

Maybe the Nielson Brothers are right, and maybe they’re way off. Nov. 6 will tell.

Is Noem-Varilek tightening, or not close?

Two pieces of information came across my desk late last week, at least superficially contradictory.

First, a reader forwarded over an email sent from someone at The Gula Graham Group, Rep. Kristi Noem’s fundraising consultants. The email — which I can’t certify hasn’t been edited beyond the redaction of the recipient’s name — appears to be a plea for donations for Noem because “things in SD are closer than we’d like,” citing the latest Nielson Brothers poll:

It’s also a peek behind the window of how congressional fundraising works. Donations to a candidate usually give the donor access, a chance to make your case. This isn’t unique to one party or the other. It’s also not really a secret. I discussed this a year and a half ago in a Rapid City Journal article:

“It will oftentimes make it that much easier to get your phone call returned, or get a meeting with a member of Congress that you’re seeking to chat with,” said (Dave) Levinthal (of the Center for Responsive Politics). “They may not be the be-all and the end-all when it comes to exacting a desired result, but it gets you in the door, it gets you in the game. That’s a heck of a lot more than the average person can usually say.”

This email is also striking in how seriously it treats the Nielson Brothers poll. Republicans in public almost uniformly dismiss NBP as biased and inaccurate, but this fundraising appeal isn’t dismissive at all.

Does this suggest the public GOP contempt for NBP is a smokescreen and the much-maligned pollsters actually have an accurate picture of the race? Or is Noem’s camp simply trying to stir up donations from supporters by pretending to worry about a race that isn’t that close?

Late last week I also spoke to Stuart Rothenberg, editor of the nonpartisan Rothenberg Political Report. Analyzing congressional races is what this guy does full-time — identifying which races are close, which are blowouts, and which are too close to call.

Rothenberg said “Democrats continue to talk about (Noem-Varilek) as kind of a dark horse upset special,” but that they so far haven’t done much more than talk.

He looked into his database for polling and found the Nielson Brothers polls.

“If these polls, at least two of the three, reflected reality, there’d be a lot more buzz in Democratic circles,” Rothenberg said. “We’re just not hearing (about) the race.”

I mentioned my understanding that the Democratic Congressional Campaign Committee had polled Noem-Varilek.

“If it was good, they’d be sending out a release,” he said. “We get releases from crappy races, where ‘If he gets the money, he could win.’”

Nothing, however, from the DCCC on Noem-Varilek.

That said, Rothenberg acknowledges that he and the campaign professionals can sometimes get caught by surprise.

“Are there races around the state where I wonder, I worry that both parties are missing the boat, they’re not playing?” he said. “I don’t know what’s going to happen on Nov. 6.”

It seems pretty clear that Noem’s probably not approaching 70 percent landslide territory, the heights reached by Tim Johnson in 1988, John Thune in 1998 and Stephanie Herseth in 2006 — their respective elections for their second full terms in the House. (It’s not impossible she’ll close strong and reach that area, but the buzz around the race is not that of a landslide.) But beyond that? Who knows.

Nielson Brothers: races tightening

In July, the Nielson Brothers showed one-point leads for Kristi Noem and Kristie Fiegen. (That’s bad for the Republicans/good for the Democrats!)

In September, they showed nine to 10 point leads for Noem and Fiegen. (That’s good the Republicans/bad for the Democrats!)

Now in early October, the Nielson Brothers are showing somewhere in between.

A new poll was conducted Oct. 1 through Oct. 5, with 730 to 762 respondents (it varied per question) and a margin of error between 3.55 and 3.63 percent.

In it, they show the following:

• Romney +10.5 (51.6 Romney-41.1 Obama)
• Noem +5.7 (49.3 Noem - 43.6 Varilek)
• Nelson +26.4 (55.1 Nelson - 28.7 Nemec)
• Fiegen +2.1 (38.6 Fiegen - 36.5 McGovern - 8.5 Clarke)

Obama’s job approval is 43 percent. Noem’s is 53 percent. Prior NBP polls of the PUC race did not include Libertarian Russell Clarke.

So far the Nielson Brothers have provided the only publicly released polling of South Dakota races this year. They have their critics in the press and especially among Republicans, due to a short and uneven track record and the ties of one of the Nielson brothers, Paul Nielson, to the South Dakota Democratic Party.

I’d say definitely take these numbers with a grain of salt. But until I see other polling suggesting I should, I’d hesitate to dismiss NBP’s polling out of hand as some people encourage me to do.

Someone’s polling Buhl-Miles

And on the cheap, too.

The Argus Leader’s education reporter, Josh Verges, seems to get polled all the time. See here and here.

This one was short and interesting. Verges said there was just one question (paraphrased):

For District 15 Senate, do you support Kathy Miles, Angie Buhl, or are you undecided?

And that was it. No demographic questions or anything.

Without any understanding of the demographics you’re polling, a poll’s numbers are of limited value — and less accurate, without the ability to do any sort of likely voter screen or adjust your sample to fit the population.

To me, this sounds like a state legislative candidate who wants a quick-and-dirty snapshot of the race for as little money as possible.