I stupidly listened to “Hacks on Tap” the other night. After I heard David Axelrod say that “all the polls show Trump winning because he has a 20 point advantage with Hispanic and working class voters of color” I turned it off.
I’ve been around and in the business a long time. I’ve sat in the back room during Frank Luntz dial groups and have eaten way too many M&Ms watching focus groups (and written and fielded tons of polls). I know how this stuff works and cannot believe the constant stream of stupid that is being served up by pollsters and experts in the press this election cycle by those who should know better, like David Axelrod. I think it’s time to take a step back and set the record straight.
1. The data is clear — we are winning.
There is no evidence — outside of polling — that Trump is doing dramatically better with low propensity voters which is why he is supposedly winning and Biden is supposedly losing. We’ve had multiple elections since 2020 and Dems have improved our numbers with HIspanic voters and working class non white voters. We won in 2022 (in Arizona) because we outperformed with the Hispanic vote. We won in NY 3 in 2024 because we outperformed with the Asian American (AAPI) vote.
And there is evidence — from exit polls of actual voters in the primaries — that Biden is going to get a significant share of disaffected 2020 Trump voters (who voted for Haley). Remember, these are polls of actual voters. There is no better model for the voting day electorate.
Think about it — what has Trump done/offered in his politics to win more Hispanic voters and voters of color? Nothing. The border issues — outside of FL — helps us not them with Hispanic votes in battleground states (in AZ and Nevada). And, what has Trump done to alienate non MAGA republicans — a ton.
The polling data which Axelrod is citing (which is also variable, not shown in many polls) does not correlate to voting behavior or other evidence. Note — per Gaza — the analysts who believe Trump is winning do not believe he is winning because of the youth vote, it’s all about him somehow dramatically increasing his share of the Hispanic and working, non-white voters. (I.e., black and AAPI voters).
Additionally per Biden’s base not showing up, there is no evidence in actual election or non polling data of this happening. The Biden coalition showing up is why we are winning in all the special elections. The electorate has gotten more and not less favorable for us because of Dobbs — this is why we are getting more college educated voters (because of Dobbs). Dobbs also helps with other groups (like Hispanic voters who are even more motivated by Dobbs than many other voters). The prediction that inflation (and Biden’s low approval ratings) would depress turnout were 100% wrong (the red wave debacle).
In sum, it’s pretty clear Biden will get a significant share of the disaffected GOP vote and should retain most of his 2020 coalition. It’s not clear — outside of polls — that Trump will get more Hispanic and working class voters of colors to support him and it is clear that he is losing non MAGA republicans and independents. Trump’s vote floor is declining and Biden’s is increasing. And, Trump was already behind. The data is clear, we are winning, potentially by a lot.
2. The assertion that the 2022, 2023 and 2024 elections are not predictive for what will happen in November but the polls are predictive is statistically stupid
The experts in the media dismiss the salience of Democrats winning almost every election since 2020 for what will happen in the general, they say, the general election will be different because shy Trump voters will somehow emerge. This is just stupid and wrong.
Special elections are, particularly in large states, a sample of the likely electorate in the general. Yes, more committed voters — like college educated suburban voters — tend to turn out more in off year elections but the Hispanic voters and working class non white voters also turn up and vote. In fact, Dem performance with those voters is why we won in VA in 2023 (not because of white college voters from NOVA). It is incredibly dumb to hand wave over these results and ignore this data if you are trying to predict outcomes in November.
These voters — HIspanic, AAPI and Black voters — come from the same communities and neighborhoods as low propensity voters. The voters who show up are representative of the Hispanic voters and working class white voters Axelrod and others are discussing. If there was a major change in the support of these voters, it would show up in the voting data (some of which is very recent) but it hasn’t. Yes, it is different voting for Katie Hobbs versus Joe Biden but in this age of extreme partisanship, party ID at the ballot box is strongly correlated with later voting behavior. There is not a lot of partisan ship voting variance currently. If there was a massive change happening in the electorate — as indicted by the polls — it would show up in the off year elections and it has not.
3. Polling is broken and getting worse (and sloppier)
People don’t answer their phones, emails, texts or their doors anymore. It is very hard and expensive to reach and accurately model voters because you have to make a lot of contacts to get the necessary number of respondents. As a result, polling is getting less accurate and less robust. This is why the red wave was missed. The battleground, state polls, were inaccurate. The polls have been off by 4 to 8 points in the recent off year elections. Suozzi won by four more points than he was supposed to — according to the polling average. Trump’s actual vote total has underperformed the final polling average in almost every primary state by 6 to 20 points.
In addition, polling is getting sloppier and much less rigorous. Back when I was more involved with the stuff, we were not allowed to analyze cross tabs with a low number of respondents because the margin of error in doing this is so high. Now, Nate Cohn and others are writing headlines for voter segments predicating national outcomes for voting groups based on the answers of a few hundred respondents (just 205 non white no college age votes in the poll that has driven so much coverage). It is mind blowing to me that the pollster, Siena College, is allowing this to happen. The result? 15 point plus swings in the responses of demographic groups from poll to poll (which is what happens when you have small sample sizes and a margin of error over 10%). This crosstab data that the Times and others (like Axelrod) are basing predictions that the voter electorate is making massive changes is basically worthless. This is why there is so much inconsistency in the polls.
Finally, pollsters are increasingly weighting responses to get the results they want. The Times poll — for some reason — weighted the voter poll to make it go from a +1% Democratic electorate to a +3% GOP electorate. The electorate in 2020 was +1% Dem and the trend is that the electorate is getting more Democratic and not less. The Times did not explain why they did this but presumably believe that the Trump electorate will increase from past historic levels of turnout (in 2020 and 2016). It’s a massive assumption that was made somewhat casually. Unfortunately, this kind of weighting to
result seems to be happening more and more as pollsters try to work with models and responses they don’t view as ideal or that correspond to the results they want or believe is correct.
In addition, while it has gotten more expensive to do good polling, it has gotten cheaper to do bad polls (via robocalls, email, internet response). Polling has also now become widely accessible (whereas in the past it was a specialized field). You or I could do a very bad poll right now via existing online tools. The margin of error for polls done cheaply (and with small samples) is huge and not predictive. Unfortunately, the prominence of the polling averages — a valid way to reduce margin of error when the average was of a lot of high quality polls — and the low cost and availability of bad polling has meant that political actors can and are flooding the polling averages with bad polls designed to support their candidate (and the narrative they are pushing to further that support). You just need to create a seemingly credible group name for the pollster and you are in, like this one from “Mainstream Research” that polled only 250 respondents to make claims about North Carolina. It’s absurd.
So, polling, while another point of data, is a much diminished tool and is no longer nearly as predictive of actual electoral outcomes as it was in the past.
4. “Election” experts who are actually not experts are increasingly using poll data to make predictions in support of narratives that benefit them.
Despite the diminishment of polling as a predictive tool, more and more “election” experts are citing polling to support the narratives they want to push.
This is because polls come out frequently and are relatively easy to analyze. Anyone can read a cross tab “wow look at the support among voter X.” What is hard is to understand the statistical significance of the finding and to weigh it along with other data (from voting) in election analysis.
Most of the election experts in the media are not people who run campaigns or understand polls or elections. Axelrod, for example, made ads for Obama (he was not in charge of modeling the electorate). The Pod Save America guy are mostly speechwriters. Other political experts have even less experience — they are journalists who cover politics (like Ezra Klein) — not people who run and win campaigns. But, they are rewarded by more clicks, views and attention and, as a result, are greatly incentivized to promote narratives that get them attention from target media consumers.
So, they use polls to inform narratives that get them attention and help them achieve more prominence as experts. Think about, if you are an expert and say — it’s a toss up — you won’t get the clicks or views you need. This negative impact of this incentive structure is compounded by the material benefits of concern trolling progressive voters. The New York Times entire current business model — the money they get from their 10 million digital subscribers — was built by massive, concerned engagement following Trump’s election.
So, you have experts who are not really experts who need data to help them get attention and are not able to use anything else but polling to easily provide support for the articles, videos, they need to write. And, these experts often don’t really understand how to properly characterize and qualify polling an actual predictor of electoral outcomes. This, combined with polling breaking, means that there is a ton of bad and questionable information flooding the zone. It also means that — when the narratives these experts are flogging are questioned — they defend these narratives and attack other points of view (because they care about their narratives not the data).
5. Trump winning is predicated on him still being an “outsider” to voters when he is, in actuality, a former President.
The theory that Trump can win is predicated on the idea that he is suddenly — even more than in 2016 and 2020 — a beacon for disaffected voters mad at the whole system. It is not a logical argument. Trump is well known to voters and is a former, infamous President. He is not new or interesting to anyone outside of his core base. He is an incumbent. His turnout — with his base — will be strong again. But, it’s not clear how or why he will suddenly attract new or previous Biden voters to exceed his 2020 numbers (which he needs to do). There is no evidence of this new electorate which is why the experts who want to sell this narrative are quick to jump on small poll crosstabs with high margin of error that support this point. And, again, in contrast there is evidence that Trump’s base is eroding not expanding.
6. Don’t get me wrong, I am still worried and think we should all be worried. But, we should also know that we are winning and can, if we do the work, win by a lot.
I am worried about the impact of RFK Jr, No Labels, AI, October surprises many things but it is clear if we do the work we can win, and potentially win by a lot.
To quote Simon Rosenberg, I’d much rather be us than them. Let’s do this thing. And, when you see an another bad poll, remember: we are winning!