Lakshya Jain hosts weekly Substack Lives with election analysts so you’ll know all the latest midterm developments. We wanted to make sure you saw this one, but be sure to sign up for The Mag to get them in your inbox every Wednesday!
VoteHub’s new election forecast currently gives Democrats an 84% chance of winning the House and a 47% chance of winning both chambers of Congress.
Zachary Donnini, the head of data science at VoteHub, joined The Argument’s Director of Political Data Lakshya Jain Wednesday to discuss what data the model covers and what it predicts.
While Hispanic voters are expected to swing heavily Democratic again after supporting Republicans in 2024, this only matters in a handful of places:
“A lot of these Hispanic votes that Democrats will gain are wasted votes, right? Like, going from D+49 to D+65 in the Bronx doesn’t help Democrats at all. It doesn’t help them win House seats. It doesn’t help them win Senate seats,” said Zachary. “Where it does help the Democrats is in some of these newly competitive districts in Florida … It’ll help shore up some of these districts in New Mexico, Arizona, Nevada, and California that have been problems in the past.”
Controversially, VoteHub’s new model uses data from prediction markets like Kalshi.
Zachary defended the choice, arguing that markets can compensate for blind spots in both polling and fundamentals.
Lakshya wasn’t sold.
His concern was reflexivity: If forecasters lean on prediction markets, and traders in those markets are themselves moving prices based on what forecasters say, the system can end up amplifying its own assumptions rather than picking up new signals.
“How do you guard against improper feedback loop generation?” he asked.
To hear Zachary’s defense of this decision, watch along.
