They say surveys are bullshit! (And other advice)

Two recent events prompted me to write this. One was a post here on Daily Kos and the other was Steve Kornacki’s latest stylings. Let me take them one by one.

The Daily Kos post was this:

www.dailykos.com/…

It attempts to present a slightly optimistic argument for Democrats taking back the House of Representatives. This immediately struck me as bizarre: Democrats narrowly lost the House of Representatives in 2022, and they did so by completely ignoring some seats in New York (George Santos?) that they should have won. They should take them back. You will lose some through gerrymandering and gain some through ungerrymandering. But no party could screw up House leadership and governance worse than the GOP has in the last two years. With that in mind, expect them to come out on top in competitive districts.

The above post is a rather preliminary argument for optimism, but it makes an analytical error right at the start. Here is the comment I left there:

“This is a sincere post and it obviously takes a lot of effort to publish it. But right from the start, a claim is made that is highly misleading. Let’s just run through some numbers.

The claim is:

“The race ratings are historically correct. Cook’s 2024 analysis of his own ratings since 1984 found that more than 90% of the races for governor, House, Senate and president were accurately predicted. And the more secure the assessment, the higher the accuracy.”

At first glance, “over 90%” sounds pretty good. But think about it and just use the House of Representatives as an example.

There are 438 seats in the House of Representatives. According to the ratings here, 367 of them are either “Solid Republicans” or “Solid Democrats.” These seats can be predicted without looking at a single poll: just look at actual voting history. If the district lines haven’t changed and it’s gone decisively for one party or the other in the last 6 cycles, then call it and you’re done. That prediction will be more than 99% accurate right there. Riots in such districts are big news and are usually accompanied by some sort of scandal. If there were more than four participants in an election, we would know.

OK: So 71 seats remain. Less than 20% of the total. Here’s what you can do with them: Flip a coin and make your prediction. Certainly this method will have an accuracy of about 50%, so there are now over 400 correct predictions out of 438: over 90%. Completed.

So the statistics cited do not in the least support the idea that Cook’s analysis is somehow a good source of additional information beyond the obvious. In fact, if you take away the gimmes, it more or less proves that Cook is no better than a monkey throwing darts. No better than a random coin toss.

All the fundamentals are in favor of the Democrat this year. Run the GOTV and it should be a blue wave election.”

What about Kornacki? He’s concerned about a poll showing a sudden 5% swing toward Trump (www.mediaite.com/…). But just stop and think. There was one major event this entire cycle: Biden’s debate, which was so alarming that it led to him dropping out of the race. This worried me because the Democrats would be able to completely screw up the replacement (Open Convention!), but it was handled brilliantly and smoothly. Harris was great too. Surveys showed slow and steady improvement. OK: That makes sense. But a 5 percent swing to Trump based on anything in the last few weeks? There’s just nothing that could cause this. Nothing. So this result just proves how unreliable the surveys are. It’s a magical thought to think that they are watching a huge event that none of us even noticed!

Surveys are inaccurate for a variety of reasons, some are unsystematic (random noise) and others are systematic. The unsystematic problems can be improved by larger sample sizes. The systematic (biased samples) cannot do that. Furthermore, many polls, particularly on the right, do not attempt to eliminate systematic biases: they exploit them and create them (e.g. through likely voter screenings). For them, polls are not about gathering information, but about creating a propaganda narrative.

A systematic bias arises because no true random sample of the population is carried out. In a laboratory, in a randomized controlled experiment, you can really randomize. But for surveys, that’s just not possible. Who do you contact? How? When you call, which groups are more likely to answer? (I never answer an unknown phone number). You can try to “unbias” the sample by “rebalancing,” but that is based on pure speculation. How do you decide who is a likely voter on the Likely Voter screen? (Surely all the voter registration surge following the Taylor Swift endorsement won’t make it through any likely voter screen.)

What do we actually know? We know that we’re starting from a baseline where seven million more Americans voted for Biden than Trump four years ago. 7 million. What would make many of them change their minds? As has been pointed out here again and again, we know that Democrats have exceeded projections since Dobbs, particularly where abortion itself is on the ballot, even in red states. We know that there was a surge in registrations associated with the Harris acceptance and the Swift endorsement. We know who these are intended for. We know about the recent complete upset in the Anchorage mayoral election. We know that early voting is strong and heavily skewed toward Democrats. In other words, there are all signs of a strong blue wave. On top of that, Trump and the entire Republican Party are completely devastated at this point. These are just facts and are more reliable and important than surveys.

What can you do if you are (and you should be) anxious?

1) Don’t despair. That’s what they want.

2) GOTV of course. If you live in a battlefield, sensible people definitely remind you to vote. (If you know irrational people who tend to sit this matter out, let sleeping dogs lie.)

3) If you have the energy, time, personality and are in the right field, advertise. The acquisition diaries here were great.

4) If you have the energy, time and personality, call the bank wherever you are.

5) Show support online. Even small things like a recommendation or a thumbs up for a good post promote esprit de corps.

6) Write helpful comments or encouraging posts online.

7) This is a difficult question. If you have the time, energy, and personality for it, hit back at right-wing sites in the comments section. That means: facts, facts, facts. The right-wing websites (think Fox or RCP) are used to having comment sections that are just cheer sections: someone posts the nonsense of the day from the GOP propaganda factory and everyone gives the thumbs up. They are simply not prepared to be factually rejected and shown that the claims are simply false. Have the facts ready. As they say, don’t feed the trolls (you’ll get all sorts of nasty comments), but don’t ignore them either. Call her. If they use juvenile slurs (you, Harris, or Biden), don’t respond immediately, but point out what they are doing and that it is juvenile. Also: go meta. Sometimes you need to push back on the content, and sometimes you need to point out how the site does its propaganda. For example, RCP often rewrites the titles of articles in the links, which is extremely misleading. Most people who never click the link still get an impression of the content, and it is often wrong. Or if one of these places publishes an article along the lines of “The Internet goes wild after Harris says…” that just summarizes a lot of predictable nonsense from right-wing operatives, point out that it’s not news. Of course, Republican operatives will attack Harris no matter what she does. Yawning. Or if suddenly everything Harris says is called “word salad,” no matter how coherent it is, point out the trick they’re using. Propaganda is like a magic trick: once you see how it works, it stops working. Point out what they are doing.

We are on the right track and heading in the right direction. Keep it up and we will be smiling on November 5th.

For the record, here are my predictions. Harris wins easily. We know election night. The House of Representatives becomes Democratic. The Senate…that’s a real concern. We lost WV and Tester seems to be in real trouble. I know it would be a joy to defeat Cruz and/or Scott, but I believe that in a wave of elections, the best chance lies in Nebraska. But if they take the Senate, it won’t be a catastrophe. That means two years of good legislation passed by the House that gets nowhere, followed by a midterm election where the map is tilted against Republicans, so two years of getting things done.

And we must move forward with the National Compact so that all this nonsense about only caring about seven states ends and the only fair system – one person, one vote, equal – is put in place.

Oh, and once we have the House, Senate, and Presidency, give DC the congressional representation it deserves, and Puerto Rico if they ask for it.