Polymarket was right. I was wrong.

In my last post, I talked about the potential for polls and betting markets to be misleading when it comes to predicting elections, and I singled out the crypto-based, foreign-based betting site Polymarket for its bullish betting odds in favor of a Trump win in the 2024 US presidential election. When I wrote the post, the betting odds could be interpreted to say Trump had a 66% chance of winning, while most “likely voter” polls had things neck and neck. While my main premise—that bad actors can goose betting markets—is still technically correct, I should point out that the big fish who bet $30M on Trump is just a French guy who thought polls were underestimating Trump’s support. And he was right. You could say over all that Polymarket was wiser about the election than the polls. Whether this would be true in other circumstances remains to be seen. Nonetheless, when money was on the line, people made the smarter assumption that Trump support was being underestimated, and nothing nefarious was at-play.

My assumption at the time was that polls were underestimating likely high turnout of women in support of Harris. It seemed, after all, unlikely that there would still be shy Trump supporters after all of these years or that polling firms had not adjusted for their existence after having underestimated them in 2016 and 2020. Pollsters will argue that their results were often within the margin of error of reality, but a few were dead wrong—most infamously the final Selzer poll that showed Harris up in Iowa. When the Selzer poll came out, someone at Rasmussen said something to the effect that it was too bad they didn’t have time to prove the poll wrong. At the moment, I thought, hah, that’s proof that a biased pollster can manufacture an outcome that helps Republicans counter the narrative that Harris was running away with things. But all they were really saying is, hang on, there is a lot of tacit Trump support that is still not being captured in polls. And they were right.

The good news? Polls were generally right that it would be a close race. What people got wrong, including myself, was in telling the right stories about what the numbers could mean. See JD Deitch’s LinkedIn post today about this. So yes, data always comes with a narrative. And while data could be used to manipulate political outcomes, I don’t think that really happened this cycle, and if anyone was not telling the whole story, it was me.

Thanks for reading.

-CW