Cross Tabs End Notes

The hardest thing to do in polling it seems is to accurately guess who the voters will be. Pollsters get better and better at weighting samples to more accurately reflect observable characteristics of voters, which is why the forecasts keep getting close to the center of the bullseye. But there are still these harder to observe characteristics that it's almost impossible to model... unless you can figure out what they are.

So here's an idea - what if, instead of just throwing up our hands, we found a way to poll people who don't want to be polled?

Kabir Khanna of CBS News did just that - and he came on to talk to me all about what they did, and what they learned.

About the Guest

Kabir Khanna, Ph.D., is Director, Election Analytics & Technical Systems at CBS News. He produces stories on elections, polling, and politics, making sure they are based on best practices and innovations in quantitative social science. He ensures surveys and statistical estimates are representative and accurate, and breaks down results on air. On election nights, he projects races for the network and manages the Data Desk, generating insights into the electorate in real time, as well as estimates of turnout, how ballots are cast, and key voter groups.

Stuff We Talked About

"The voters Mamdani added to the Democratic coalition in New York: CBS News analysis"

"1 year in, Americans call for more inflation focus from Trump, CBS News poll finds"

Polling at a Crossroads, by Michael Bailey

Stay in Touch

📬 Subscribe to our newsletter at crosstabspodcast.com for new episodes, insights, and behind-the-scenes content.

📹 Watch video episodes on YouTube @CrosstabsPodcast

💬 Follow us on BlueSky: @crosstabspod.bsky.social and @farrahbostic.bsky.social

📍 Produced by The Difference Engine

📨 You can also email me directly at farrah@crosstabspodcast.com