A New York Times/Siena College poll provoked intense debate recently because it showed President Biden in pretty bad shape against his likely Republican opponent Donald Trump.
Democrats and analysts questioned the sampling and methodology. I read their posts. It’s a useful debate, just not that illuminating about what may happen in November.
Every poll is at best an approximation; and I am far from the first person to learn that I can do better by looking at polling averages. Hopefully the imperfections of various polls cancel each other out. As of March 5 it’s a little early even to find a good polling average; but some recent surveys posted on the 538 website look like this:
Sometimes Biden is a little up. Sometimes Trump is. Sometimes it’s tied. And sometimes, apparently, the pollsters pushed every single respondent to pick one of two candidates! This seems unrealistic, as I’ve heard from voters who are unsure what to do, and it’s likely that they will have more than two choices. I think that pushing people to answer this way may still indirectly reveal something about the electorate, but the final tally is very unlikely to be 52-48.
In short the campaign is close, and we still have eight months of campaigning, and eight months of world events, before Election Day. It’s hard to know more than that.
Yesterday I was talking with my colleague Krishnadev Calamur, the incisive head of NPR’s Washington Desk, who reminded me that NPR has changed the nature of its polling in recent years. NPR focuses much less on who’s up or down, and more on what’s on voters’ minds. More questions measure voter sentiment on issues such as the economy, crime, immigration, or Ukraine. What matters most to people, and why? This approach adds insight to the horse race.
It also matches my experience with interviewing random voters across more than 20 years of elections—I have periodically gone to politically interesting neighborhoods to knock on doors, and whoever answers is the story. I learn a lot about voters’ lives and concerns. Sometimes “Who will you vote for?” is the least interesting question. I usually save it for last.
To be clear, I’m still going to look at horse race polls! But I will watch the averages, and keep in mind how much they don’t tell me.
Thanks for reading Differ We Must, my companion to the book of the same name. Differ We Must tells Lincoln’s life story through his meetings with people who disagreed with him. This Substack tries to offer a few behind-the-scenes glimpses as I try to report on our modern-day differences.
I’m happy to say that, in addition to the hardcover version of Differ We Must, the audiobook has been selling well. I care so much about this story that I will read it to you myself! Have a look, or listen.
Steve, loved the book, and this substack. How come we hardly ever hear from media how off (or flat out wrong) polls have been once a particular election is done? Most were off in 2020, 2022, and this past Tuesday 2024. But we never seem to hear the media learn from that. They just go onto the next election and start presenting those polls as if they are accurately predicting what voter sentiment is and what's going to happen (based on past results?).
Separate, but related issue: how come most media lets pundits, politicians etc flat out lie during interviews with hardly any push back or fact checking? Sorry to say my most recent example of this was you talking to Asa Hutchenson this morning, and he went off on "failed" Biden policies like we used to be energy independant under previous administration but no more. Factually untrue, US producing more oil and NG now then ever before. I think letting one side consistently lie with hardly any fact check/pushback is aiding in this negative sentiment among some of the electorate.