Data Digest: Have You Tried Unplugging Polling and Then Plugging It Back In?
Polling averages are showing a much better landscape for Democrats these days, but people of both parties have felt burned by polls before.
Welcome to Codebook's Weekly Data Digest! Data Digest is a great way for premium subscribers to start the week.
Quick housekeeping note: on Wednesday at 8 pm eastern, I will be doing an Instagram Live! I'll be doing "Hot Sauce/Hot Take", a new thing where I make a hot sauce in my kitchen and take questions about polls and such. Some of the best questions may make their way into next week's Ask Away! Tune in, friends.
The election draws near, more and more people start tuning into what you've been monitoring all along through this little newsletter: the polls!
And when these folks tune into the polls, they are finding they are occasionally surprised at what they see. It is, after all, a midterm year where Republicans were cruising to an easy win...and now suddenly there are a million Republican strategists giving not-for-attribution freak out quotes to eager political reporters.
Enter Nate Cohn's latest at The Upshot on whether we can trust polls this year.
I believe a healthy skepticism of the polls is good. However, I not a Poll Truther. (I'm literally a pollster.) I do not think there is some big conspiracy to create Fake News Polls to bias the landscape. I believe this not just because I am a pollster but because I believe in capitalism, markets, and incentives.
Do you know why I try to produce very accurate polling and insights? Because I think it is the right thing to do, because I take pride in my work, because I genuinely want to know the state of play, and also, let's be honest: it would be very lucrative to be the pollster who gets it exactly right!
I have not asked him, but I assume Robert Cahaly of Trafalgar has many more clients post-2020 election than he did before, because he got a lot of states closer than many public pollsters.
So I keep chuckling at this tweet...
Readers, I have to confess: I had not thought of that before! What if instead of distrusting the polls we just fixed them? WOW! What an idea. (As one of my colleagues dryly noted to me, "have we tried turning the polls off and then turning them back on?", hence the title of this newsletter.)
I'm being sarcastic, of course. But an awful lot of effort has gone into fixing the polls! AAPOR has a whole conference about it and it is generally a real hoot! I feel like I'm losing my mind here.
The problem of "the polls might be wrong" is that you cannot know ahead of time if they will be. We pollsters get stuck fighting the last war an awful lot. In 2012, Republicans weren't calling enough cell phones. Boom, fixed. Then, in 2016, the cell phone thing was fixed but now we weren't calling enough people without college degrees. OK, by 2020, fixed that thing...but then the election happens and you still have a lot of states where the results were wonky. AAPOR's big report digging in suggested that the problem is a mix of things, including the actual materialization of a "shy Trump vote"-ish problem, largely driven by low social trust voters.
One of the ways that pollsters are trying to fix it THIS time is by ensuring that your sample is matched back to some kind of prior political event: primary turnout, vote history, and so on. We won't really know if this worked until after the election ends. Even then, if the result is a mixed bag, the answer might not be satisfying.
That's why in Codebook I am always cautioning people to consider polling thoughtfully. Do not trust it blindly and do not reject it as trash. Take it in the spirit in which it was generated and don't ask it to do something it isn't built to do. I still believe polls can give us very meaningful information about trendlines, even if I'm skittish about letting it "predict election results."
Thank you for coming to my mini-TED Talk.
On to the rest of the data...
Presidential Approval
The current Real Clear Politics averages for the President's job approval are:
Overall job approval: 43.1% approve (up 1 points from last week, up 2.8 points since August 6)
Economic job approval: 38.4% approve (up 0.1 points from last week)
Biden's job approval sure is building back better, eh? (Go ahead and groan.)
This week's bump is driven by two polls, IBD/TIPP and, hilariously, still Rasmussen. I cannot get over this!
But I think this screengrab from the Reuters/Ipsos poll page is illustrative of the trend we've been seeing. Republicans are not deciding that Biden is maybe not so bad. Instead, it is Democrats who are just a teeny tiny bit deciding that Biden is maybe not so bad.
Source: Screengrab from Reuters/Ipsos
2022
Because Real Clear Politics and FiveThirtyEight calculate their "generic ballot" in different ways, I'll include each here each week for context:
RCP: Democrats +0.4 (GOP advantage down 0.3 from one week ago, down 0.5 from August 12)
538: Democrats +1.3 (Democratic advantage up .3 from one week ago, up 1.0 from August 6)
This, from Ally Mutnick at POLITICO, sums things up well:
In all, Republicans need to net only five seats to win the gavel. And while Democrats may be poised to mitigate some losses, Republicans say there’s still little chance the party’s summertime surge can overcome the stacked map.
If Republicans only need to net five seats, and there are 32 seats listed as "toss-ups" by Mr. I've-Seen-Enough, Cook Political Report's Dave Wasserman, even a fifty-fifty split of the tossups gives Republicans the House. Incumbents are often harder to unseat, but seven of the Democratic "toss-ups" are open seats anyways. Plus, district lines are all new.
2024
The WSJ poll I looked at last week has Biden +6 over Trump in a national rematch. But. If you take a peek at the latest few Emerson polls out of swing states, they show Trump up...despite the Mar-A-Lago raid, despite Dobbs, despite J6 Committee.
In the last month, they have Trump...
Up 3 in Arizona
Up 5 in Georgia
Up 6 in Pennsylvania
Up 14 in Ohio
Imagine a *big eyes emoji* right here.
COVID-19
Here's how this week's COVID-19 picture looks, via The New York Times.
COVID-19 7-day new case average: 66,159 (14 day change of -25%).
COVID-19 7-day hospitalizations average: 34,931 (14 day change of -9%).
It is really remarkable to watch the weird convergence of the case data and hospitalization data. Likely, this is because cases are just not being reported to or by public health officials. I mean, if I am in the CNN green room tomorrow and I take a test and I'm positive, I'll rush out of the building in a mask and try to isolate from my daughter. (Please do not let this happen, universe.) But you know who I probably won't tell? The DC government. Not because I'm trying to stay off the grid, but I don't tell the DC government if I get a cold, or if I have a stomach bug, or what have you...and COVID has begun to feel like that.
You have to wonder if case data is valuable at all at this point.
The Economy
Here's how this week's economic picture looks:
AAA's National Average Price of a Gallon of Gas: $3.72 (down six cents from last week)
Unemployment (August 2022): 3.7%
Inflation (July 2022): 8.5% year over year (next release: tomorrow!)
University of Michigan's consumer sentiment data tells the story neatly. Essentially, people feel better today than they did in July. But they feel a lot worse today than they did one year ago.
And remember, don't believe the hype: the economy remains THE big issue. Here, again, that piping hot fresh Reuters/Ipsos poll...
Odds and Ends
In the wake of the death of Queen Elizabeth II, follow fellow Bulletin writer Elizabeth Holmes on Instagram for her "So Many Thoughts" coverage of all things royals these days.
Alex Albon! Please get well soon! (Albon, a driver for Williams F1 team, had appendicitis this weekend and had complications during recovery.)
Matthew Berry's 100 Facts column remains one of my favorites. How that you know how Week 1 has gone (mostly) for your fantasy football team...how's it looking?
Wednesday at 8 pm! Hot Sauce/Hot Take! Instagram Live! Be there!
(Cover photo: Westend61/Getty)
Thanks again for reading Codebook! If you have any feedback, leave it here in the comments - I read them all and value your opinions! - Kristen