When The Argument‘s editor-in-chief Jerusalem Demsas and I first began talking about building a polling program, we were both obsessed with one problem: Issue polling was dominated by interest groups.
So many of the public opinion polls you have probably seen over the past few decades have been written and released by people who have both financial and professional incentives to reach a predetermined answer. That means they will often write questions in biased ways and even withhold negative findings from the final published results.
This has polluted the information environment. Agencies put out surveys with loaded wording on everything from climate change to the Middle East, and prominent media outlets uncritically run with them.
We wanted to do something different. So, we made three decisions:
We were going to create a panel of experts across the ideological spectrum to review our language for accuracy, bias, and effectiveness.
We would release everything we polled — no matter how the results impacted our preferred political ideology.
We would dive deep into issues, testing from multiple different angles so we could try to figure out what was really going on.
Over the last year, we’ve conducted 11 national surveys on topics ranging from artificial intelligence to personal health (you can see every single one of those surveys here). These surveys provide us with an immensely rich dataset on how Americans really think, which forces us to recalibrate our view of public opinion and reevaluate bad priors.
In that vein, I wanted to share some of the findings that surprised me the most over the past year.
If you have found any of this work useful or informative over the past year, become a paying subscriber and help support our work!
1. Everyone’s moved on from COVID-19
I came to The Argument thinking that Democrats still had to reckon with a deep, persistent set of COVID-induced scars for the electorate.
In some cases, this might be indirectly true — for instance, Democrats face a staggering deficit on crime with critical portions of the electorate, and part of that is likely due to the COVID-era crime spike. But when it comes to topics like school closures and lockdowns, there’s little evidence that Democrats are paying a sustained penalty for any perceived excesses of the early 2020s. In fact, when we polled this late last year, people gave Biden and the Democrats better grades than they gave Trump and Republicans.
It’s not as if there were no lasting effects from COVID in public opinion, though. Public health experts have taken a deep and sustained hit, with 45% of voters saying that they “often hide or distort information for political reasons.” While nobody tested our exact wording pre-pandemic, the closest counterparts we have suggest a deep decline from the pre-COVID consensus; in a 2019 survey done by Pew Research Center, trust in medical scientists was at 87%.
Anti-vaccination sentiment is also worryingly high, with 42% of registered voters claiming that the side effects of vaccines outweigh the benefits. That shift is mainly thanks to a deep decline in Republican support for vaccinations. And voters in our survey did not want another set of COVID measures in the event of another pandemic; they believe such measures may have been necessary the first time, but they also think that once was quite enough.
But that’s different from whether Democrats are paying a political price for their COVID policy. And there’s little evidence to support that hypothesis. In fact, even after all the hullabaloo over mandatory vaccinations and masking, Democrats enjoyed a large and clear edge over Republicans on personal health. That is not what you would expect in a world where Democrats are continuing to experience an asymmetric partisan backlash to COVID.
It’s not like this topic has gone away from the news; recently, former National Institute of Allergy and Infectious Diseases director Anthony Fauci’s top aide, David Morens, pleaded guilty to concealing COVID records. And yet, Democrats haven’t dipped at all on the generic ballot since then — their lead has grown from D+6.2 to D+6.5 in Nate Silver’s tracker.
It doesn’t seem to have changed public opinion on the issue either; after Fauci’s July 29 Senate testimony, The Economist/YouGov asked Americans “Which party do you trust more to handle vaccines and public health?”. Democrats led Republicans 40% to 26%.
Some people might be surprised by this. But there’s a long history of voters moving past scandals and evaluating parties based on the here and now, rather than what happened in the past. After all, Americans moved past January 6, Watergate, and Iran-Contra fairly quickly, and Democrats were never as underwater during COVID as Republicans were during said scandals.
There’s no reason politics should only work in one direction.
2. Twitter really is not real life.
Before removing social media from my phone and decreasing my posting frequency by a factor of 10 in 2026, I used to get a lot of my news from Twitter and other social media applications. Removing them from my life has done an immense amount of good for my mental well-being and my ability to understand what the average person thinks about anything.
I don’t think we properly understand just how bad social media is — and what a skewed picture it provides of the world. As an example, in the immediate wake of the January Minneapolis shootings, there was only one cohort in which positive views of Immigration and Customs Enforcement (ICE) broke 40%: people who got their news from Twitter.
It’s possible, of course, that these are pure selection effects and that Twitter is not actually causing people to think differently — unsurprisingly, Twitter is also the platform that has the most favorable view of Trump. It’s actually the only news source where his approval splits evenly; everywhere else, he is noticeably underwater (including cable TV, local TV, and podcasts).
But we also have evidence that Republicans on Twitter are very different from Republicans on other platforms, which compounds the information/“vibe” distortion you might experience on the platform.