Three Weeks Out: Is The Race Really This Close?

October 15, 2024

Source: The New York Times

Is the zone just being flooded with crap?

We have seen that a lot of pollsters are using recalled-vote weighting and that may be overestimating Trump’s support nationally and in some battleground states. There is also the “flooding the zone” problem.¹ Nate Cohn doesn’t think it’s making much of a difference in the polling averages, largely because Republican-aligned polling is not that much different than some non-aligned polling. But one problem might be that recalled-vote weighting and other methodological techniques produce similar results to what Republican-aligned voters want to show.

You may remember this phenomenon in 2022, and a bit in 2020, in which some suspect or just plain lousy Republican-aligned pollsters released survey results favorable to Republicans in an apparently effort to influence the polling averages to make it look like Republicans were doing better than they were. This phenomenon is not only back this year, but appears to be more intense and possibly even coordinated. In the past couple of weeks, we have seen polls from about 40 non-aligned pollsters, only one Democratic-aligned pollster that I am aware of, and about 30 Republican-aligned pollsters. What’s the point of this, especially if - as Cohn says - it is not moving the needle so much on the polling averages?

In 2022, the point was to influence the political narrative to make it seem like there was going to be a “red wave,” or big Republican victory. Those of us who saw what was happening and were watching the early voting data knew there was no red wave forming. In fact, Democrats significantly over-performed the conventional wisdom on a president’s first mid-term election. This year, the flooding of the zone is also about influencing the narrative to make it seem as Trump is doing better than he is, but there is another, more sinister reason afoot.

This is likely an attempt by Republicans to justify refusing to certify election results should Harris win in November. The current polling averages suggest Harris would win the national popular vote by three points if the election were held today. It’s possible she could be winning by five or six points for a number of reasons: the recalled-vote weighting problem; a typical polling miss of 2-3% in her favor; and an underestimation of the youth, women, and Black vote could all singularly or collectively mean that the current polling is missing a lot of the Harris vote. While the average of the polling might not be impacted much this year by these Republican-aligned polls, the public continues to hear about one poll after another with close results. It’s not just in having close polling averages; it’s in the amount of crap they flood the zone with. More polls equals more crap.

If the zone continues to get flooded by lousy polls that favor Republicans, the public perception on the eve of the election may be that the race could go either way. If Harris wins by five points and with over 300 electoral votes - something I think is a fair bet right now - Trump will scream that the election was fixed somehow. (Of course, he will do no such thing if he wins.) I am not someone who worries that people won’t work hard if they think their candidate will win; in fact, I think winning helps motivate people. So, I am not so sure Democrats should not all be talking about how Harris is likely to win this race by 3-6 points with their neighbors in an effort to manage exceptions against the Republican plot.

Seeing the methodological problems: a case study of one poll

Last week, TIPP Insights released a new poll in Pennsylvania that had Trump +1. This is one poll where you can clearly see the methodological problems. To be clear, I am not accusing TIPP Insights of participating in flooding the zone (it may be a less sinister, but no less problematic phenomenon known as herding). The New York Times Poll Tracker does not consider it a “select” pollster (which refers more to transparency than quality) and 538 rates it as just under 2-star pollster.

When we dig into the methods used, we discover that the poll assumes that just 1.4% of all voters in the state will come from Philadelphia. In 2020, 11% of votes cast in the state were from Philadelphia - of which 81% went for Biden. If only 1.4% of Pennsylvania voters are from Philly and Trump can only win the state by one point (which is within the margin of error, btw), then he must be in trouble. That’s not a good poll for Trump if you understand this methodological problem. The cause for this error is largely attributable to a failure of their likely voter screen. While they do ask voters if they plan to vote, that is only one factor they used. Factors that make a respondent less likely to be a likely voter include being young and being Black. As a result, in a panel of 803 respondents cleared as likely voters about 11 were from Philadelphia. However, there are some pollsters out there suggesting that it looks like TIPP started with the topline result and massaged the data until it got the result it either wanted or thought was more likely. Unfortunately, they may be correct.

Pollsters have been thought to do similar things in the past when they got results that seemed out of sync with the prevailing thinking, leading to “herding,” in which pollsters converge their results around a consensus. I am not sure that herding has ever been shown to actually exist - except maybe unintentionally by using methodological techniques that are more likely to weight the results in the expected direction. This poll does not seem to be herding since other pollsters have been finding Harris leading there more often than not. But it does bring us back to our old friend, recalled-vote weighting.

The TIPP Insights poll was in the field at the same time that New York Times/Siena was. The latter’s poll for Pennsylvania found Harris up by four points. If you recall, Siena does not use the technique. It’s not clear whether TIPP Insights does either, but they have an odd likely voter screen and they say they weight by prior elections. In any case, Trump +1 is the kind of result we might expect to see in Pennsylvania with recalled-vote weighting assuming nothing has changed there in four years. It does appear that TIPP Insights expected a Trump +1 result and massaged the data and methods to make it work. I am not saying they did that, but it sure looks like they did. In any case, they should have thrown out a poll that only had 1.4% of their likely voters coming from an area that we know will provide over ten percent of the actual vote share. If they were really weighting to past elections, one would assume they would have considered that.

Is Harris really in trouble with Black voters?

We keep hearing stories about how Harris is in trouble with Black voters, particularly Black men. We know Harris surrogates, such as New York AG Letitia James, have made behind-doors pitches to Black men. As it should be: you should ask voters for their support, not assume it. It’s not only the polite thing to do, but it can have real practical consequences.² We also know Obama made a more public charge - scolding might be a better description - to Black men, which probably not the most productive way to handle it. This suggests to us that the Harris campaign agrees with the pundits that she is in trouble with Black men. I don’t know what the campaign actually thinks on this issue, but I don’t think they have reason to be concerned because we have seen Black support for presidential candidates depressed in fall polling before. It happened in 2020, and some folks thought Biden was in real trouble with this key Democratic constituency. In the end, 90% of Black voters went for Biden.

I suspect the same will happen again. It’s a phenomenon known as “coming home.” It is not unusual to see blocks of supporters withhold support in polling only to “come home” in November and vote for their preferred party or candidate. In 2020, a lot of folks thought now-Democratic National Committee chair Jaime Harrison had a real shot to beat Sen. Lindsey Graham (R-SC) because of the high number of undecided voters in the polling. But, Republicans were angry at Graham and were expressing it this way. They eventually came home and Graham won easily.

I don’t know if anyone has ever studied the phenomenon to figure out why this happens, but I have some ideas. One is that in our political system this is the only way to productively express dissent or dissatisfaction with a candidate after the nominees are decided. I think it may be fair to think a number of Black men are not happy with Harris’ promotion of herself as a prosecutor, but in the end most will come home to her. There is a similar phenomenon happening with voters concerned about Palestine, but it is a much smaller effect.

In the end, voters in our presidential system have three options - and only one is productive. They can vote for the major party candidate that best aligns with their politics, the can vote for a minor party candidate who will certainly lose and perhaps help to give the major party candidate who does not align with their politics a better chance to win, or they can not vote. Those are every voter’s choices; it’s not a normative assertion, it’s a descriptive one.

Lawn signs don’t vote … but that’s not the point  

In 2010, Worcester County Sheriff Guy Glodis ran for Massachusetts State Auditor against former state Labor Secretary Suzanne Bump. I knew both candidates; having gone to college with Guy and worked for Suzanne. Glodis was a former state senator whose father had also been a state senator. Bump had been a state representative before going to work for Gov. Deval Patrick (D). Both were Democrats and faced off in the state primary. There is an old saw in politics that “lawn signs don’t vote,” yet campaigns continue to make a priority out of getting them to supporters and getting them placed in high-traffic areas. By all accounts, Glodis was creaming Bump in the lawn sign battle. Glodis signs were everywhere in some parts of the state and in others there still appeared to be a lot more for him than for Bump. Bump supporters were often heard worrying about what this meant for the election. Well, lawn sign don’t vote and Bump won convincingly. But here’s why lawn signs do matter.

Lawn signs are part of a suite of activities known among campaign activists as “visibility.” People holding signs at the corner or at a public event are also visibility. The point of visibility is to not just to familiarize voters with a particular candidate – although building name recognition is important, particularly in local races – but to demonstrate support in the community. As we have seen from the 2010 Democratic primary for the Massachusetts Auditor, more lawn signs does not mean more votes. There are all sorts of reasons folks have lawn signs in their yards: doing someone a favor, for instance, or landlords putting signs up on their tenants’ properties. And some folks just won’t put one up even if asked, so the lack of lawn signs doesn’t always tell us much either. But these factors are often more true in local and state elections than presidential ones. While landlords might put up signs on their tenants’ properties where those tenants are not supporting the presidential candidate on the sign, there are likely much fewer persons putting up signs as a favor for a friend or the candidate. As such, it is much more likely that when you see a lawn sign for a presidential candidate, that home includes at least one voter for that candidate.

But there is an interesting twist this year. Trump supporters are well known to be highly visible when it comes to their support. In fact, it sometimes goes from support to intimidation. Trump supporters don’t usually have just a lawn sign. They will fly a Trump flag and have other pro-Trump decorations on their property. It’s been pretty easy for the past eight years to spot Trump supporters in neighborhoods, and not just at election time. For that reason, I think dashboard surveys³ of neighborhoods this year might just tell us something they would probably not tell us in a typical election. And reports of such activity on social media have been suggesting that in a lot of places where Trump had a lot of visibility in 2020 there is little evidence of support today. This suggests to us that – at the very least – perhaps there is no hidden Trump vote like what pollsters missed in the past two elections. But there could be other reasons.

Is it possible that even Trump supporters are suffering from Trump fatigue? Yes it is. It’s really hard to keep up that level of cheerleading over a number of years – even if you otherwise love the guy. There are a lot of football fans in America – but even most of them do not fly their team’s colors all the time. However, they typically do on gameday – especially if gameday is the Super Bowl. And when it comes to politics, the presidential election is the Super Bowl. So, my guess is that any decrease in Trump visibility is not due to Trump fatigue.

More likely, I think, is that this is a colossal failure of Trump’s ground game. While some people will make their own lawn signs, the vast majority of people get their signs from the campaign. There has been a lot of reporting on Republicans expressing concern about the lack of a ground game. Trump has opened far fewer offices in battleground states than Harris has. And the Trump campaign has outsourced its canvassing (door-knocking) operations to outside vendors. I am hearing that these outside vendors are doing a very poor job of getting the work done. (I have a suspicion we’ll see that people close to Trump and perhaps Trump himself are financially benefiting from this arrangement.) If the decreased Trump visibility being seen in some states is not indicative of Harris outperforming the polling, then my money would be on the campaign’s inability to deliver what it needs to voters on the ground.

Yes, these dashboard surveys are anecdotes, not data.⁴ And there really is no way for me to verify the claims other than to report I am hearing them over and over again from Democrats in battleground states. They may be seeing what they want to see, but considering Democrats’ talent for seeing the glass as half-full I would expect to hear a lot more concerns about losing the lawn sign battle once again. Assuming there is some truth to these anecdotes, they can help us to groundtruth data. And that gets us back to whether the polling is capturing the 2024 electorate accurately or not.

Polling frequently misses shifts in the public, especially when they are not big. But big shifts do not have to happen to influence presidential election outcomes these days. Aside from 2008, we have had close elections every cycle this century. Not always razor-thin, but close. Nate Cohn had a chart showing how closer the election would be if the polls this year had a 2020 miss and a 2016 miss. Harris starts with +3, and it goes down from there. But what if it’s a 2022 miss? That’s interesting.

Source: The New York Times

With a 2022 miss, Harris wins by +2 nationally, but she wins bigger in the battleground states. In fact, she wins 25 more electoral votes with a 2022 polling miss (it varies according to state polling) than if the polling is spot-on, despite losing a point in the national popular vote. This is further evidence that Harris may be able to win the electoral college with a very close national popular vote win.

The anecdotal evidence out there that there is less Trump visibility than there was in 2020 might give us a clue to the direction the polling miss is headed this year. The lawn signs - or lack thereof - might be telling us something.

With three weeks left and millions of people already voting, the most important thing you can do is help to get out to mobilize and protect the vote. I don’t think donations alone suffice this year; think about what you can do, and lend a hand. Sometimes, helping out a campaign has nothing to do with actual election work - like cooking food for volunteers - so asking how you can help might get you a surprising (and perhaps welcome) response. Whatever it is, this is the time to start doing it.

1. This can be traced back to a call to right wingers to “flood the zone with shit” by fascist propagandist Steve Bannon several years ago.

2. The late former Speaker of the House Tip O’Neill (D-MA) used to tell this story, which may be apocryphal about asking people for their vote: It goes something like this (it’s been told slightly differently over the years). Tip’s neighbor told him that she was going to vote for him in an election even though he did ask for it. “I’ve lived across the street from you for 18 years,” he said. “I cut your grass in the summer and shovel your walk in the winter. I didn’t think I had to ask for your vote.” Her response: “People like to be asked.”

3. A dashboard survey is a look around the neighborhood for certain things, often while driving. It can be informal and just to get the sense of things or it could be more data-driven by counting, say, the number of fire hydrants in a certain area.

4. I have not heard anyone reporting that they have taken a count of homes with signs and compared that to homes with signs in 2020. That would not be anecdotal, but actual data. I call these social media reports anecdotal because people are largely giving their impressions of what’s going on around them and how they remember what happened four years ago rather than reporting data.