By Adam Pagnucco.
Here is the kind of conversation I hate.
Political source: “OMG. Have you heard the rumor?”
Me: “I hear lots of rumors, including a really juicy one about you last week.”
Source: “Really? I returned the money. No charges were filed.”
Me: “OK, I was kidding when I said there was a rumor about you, but let’s return to that later. Anyway, what do you have?”
Source: [Now irritated]: “Did you hear about the poll? Politician X is down by a lot. They’re toast, right?”
[Editorial note from me: the election is nine months away.]
Me: “Puh-leeeeeze. Now let’s get back to that money you grabbed.”
Source: “What money? Did I say anything about money? No, I definitely did not say anything about money. That’s my mistress calling, I gotta go.”
Let’s set aside the alleged money and mistress(es) because I’m always interested in those rumors. I hate rumors about horse race results in early polls. Repeat after me: BE CAREFUL WITH EARLY POLLS. Why?
I have had personal experience with them.
Let’s go back to the most recent – and final – campaign I worked on, David Blair’s 2022 run for county executive. Our campaign polled four times. All four were conducted by Fred Yang, one of the nation’s best pollsters. My previous discussion of polling in the race relied on published polls by three different pollsters, each using different methodologies in a tight window of time. That raises more comparison problems than reviewing four polls by the same pollster – a really good one – over a longer time period. Note: I have received written permission from the campaign to publish this data. Without it, I would never publish it.
Here are the Democratic primary horse race results from our four polls. Pay attention to the time periods and the context in which they were taken.
Horse Race from Poll Number One: September 2021
Marc Elrich: 41%
David Blair: 11
Hans Riemer: 10
Tom Hucker: 5
No preference: 33
Blair was down by 30 points. If this poll had been published, the so-called intelligentsia would have concluded that we had zero chance to win. Elrich’s margin was due to sky-high name recognition and a solid approval rating tied to COVID. Blair had lost most of the name recognition that he had built in his 2018 campaign due to the passage of 3 years.
This very early poll had one number that turned out to be right on target: Elrich got 41% support in this poll and 39% of the vote in the primary 10 months later. Think about this: he is an incumbent who spent a million dollars and actually lost 2 points of his support. I can’t think of another campaign in MoCo history that won (by 32 votes out of 141,598 cast) and was anywhere close to this terrible.
Finally, this was ten months out. No meaningful campaigning had yet occurred.
Horse Race from Poll Number Two: February 2022
Elrich: 41%
Riemer: 12
Blair: 11
Hucker: 5
No preference: 31
Elrich’s margin over Blair was still 30 points. No candidate had any statistically significant movement. (The margins of error of these polls were roughly 4 points in each direction.) Why? There was still no meaningful outreach to voters. Of course, nothing had changed. And yes, once again the intelligentsia would have said that we were doomed.
Horse Race from Poll Number Three: June 2022
Elrich: 36%
Blair: 22
Riemer: 11
Peter James: 1
No preference: 30
Elrich’s margin over Blair had declined from 30 to 14. What had happened since February? The teachers union endorsed Elrich in April but he still dropped by 5 points. (Again, his campaign was epically bad.) Blair went on TV in mid-May and was hitting the mailboxes. The Post had not yet endorsed. Still, the intelligentsia would have said we were dead because we were down by double digits with a month to go. But the campaign was not over.
Horse Race from Poll Number Four: July 2022
Elrich: 29%
Blair: 28
Riemer: 19
James: 1
No preference: 23
The Post had endorsed Blair but it came with less than a month left. He was pumping serious money into TV ads and mail, some touting the Post endorsement and others going negative on Elrich. The campaign was now fully developed and voters were making up their minds, but it was all happening very late. Three days before election day, Congressman Jamie Raskin endorsed Elrich.
Final Election Result: July 2022
Elrich: 39%
Blair: 39
Riemer: 20
James: 2
Elrich wins by 32 votes.
Think about the story told above. Until the very end, folks who believe early horse race results without considering their context would have said we were dead. They would have been VERY wrong. We had a candidate who got up at 6 in the morning to do campaign-related work and regularly went to events after dark. We had outstanding vendors like Fred Yang and Bill Knapp. We had some damn good campaign staff who knew what they were doing. We had outside support from the real estate industry to offset the Apple Ballot. And we had a greater than 6-1 resource advantage over Elrich. (It was 3-1 over Elrich plus Riemer.) Add that to Elrich’s campaign converting exactly zero swing voters over his September 2021 numbers and we almost won. But it’s historically very difficult to beat incumbents and that along with Raskin’s last-minute endorsement were probably the biggest factors in rescuing Elrich from defeat.
Let’s understand the lesson from the above experience. I am not claiming that any of our polls were inaccurate. Rather, they were snapshots of the electorate at the time they were conducted and captured the circumstances of those moments. Early polls taken before campaigning starts can’t take into account the impact of campaign activity, so they primarily reflect starting name recognition and early incumbent approval ratings. Once major campaign events take place – mail, TV, field, digital, the Apple Ballot, the Post and intensifying media coverage – then the horse race numbers become more predictive of the outcome.
The Blair example is not unique. I knew a candidate who was getting crushed in an early internal poll but later posted a solid win. I knew another candidate who was on the verge of victory in an early internal poll but later got annihilated. I won’t name these candidates because I don’t have permission to discuss their internal polls but trust me: comebacks and collapses are more common than you might believe. Repeat after me: CAMPAIGNS MATTER.
So how do you handicap a race in the early going? Here are a few suggestions.
Review the prior electoral results of candidates if they’re available. Look at their performance by detailed cuts, especially with regards to geography and racial group. (Given residency patterns in the county, the latter can be inferred by precinct distribution.) This enables assessment of their strengths and weaknesses.
Look at fundraising – of course.
Consider the advantage of incumbency. Last time I wrote about this, I found that MoCo incumbents had a win rate of 90%.
Try to forecast the distribution of big endorsements, like the Apple Ballot and the Post. (I have a series coming up about this soon.)
Compare candidates’ message and issue mix to any polling on issues, like the Banner poll. Issue measurement and message testing are legitimate uses of early polls. (Yang is really good at this.)
Look at any wild cards, especially the likely actions of outside groups with money.
Miscellaneous bits of data can be fun, such as bullet voting in the council at-large race.
And it’s OK to say, “I don’t know.” It’s typical for data points to contradict each other in the early stages of competitive races. That’s what I expect to see. Often, the most intelligent approach is to construct scenarios and estimate their likelihoods. So Politician X isn’t necessarily destined to win, but if factors A, B and C come to pass, X may be likely to win. Otherwise, Politicians Y or Z might have a shot.
Above all, rely on data – the more, the better. Don’t base your thinking on hunches, rumors, what your friends say, what pundits say, what social media says or pure opinion. If you want to rely on faith alone, then get out of politics and into the priesthood where you belong.
Priests, God bless em, don’t blindly use horse race numbers from early polling.