I reached out to ABC News and Morris, its Editorial Director of Data Analytics, with two basic questions: Why was the network’s model more favorable to Biden than other election forecasters? And what did the network think about how the model was being touted by Biden allies intent on keeping him in the race?
Though the network insisted it stands by its model, ABC News was not willing to make anyone available to speak on the record to answer this question. It did provide a well-placed source willing to answer questions off the record, and later agreed to share some of the answers from that conversation on background anonymously provided that we got approval of which unnamed quotes to run —an unusual arrangement, but one we reluctantly agreed to, identifying the person we spoke to as a “well-placed source.”
AD
The enforced anonymity was particularly surprising because Morris has widely discussed and defended his model on X, and because the forecasting business relies on shared, public data. But television networks are cautious by nature, and ABC finds itself in a strange position: Way out on a limb.
That is not, by the way, to say that Morris is wrong. Sometimes, outlying predictions prove correct, and longshots pay off big.
“If you’re trying to predict an election, you have to do two things: pick the factors that matter and then assess how much those matter,” the well-placed source said. “There’s room for disagreements within that step. That adds uncertainty about how the model is going to perform in the future. Very reasonable differences can lead to different results because the future is so uncertain.”
Silver, in a recent post on his new Silver Bulletin, made the same point more sharply: “At best, all it’s really saying is that Biden will probably win because he’s an incumbent,” he wrote. “At worst, it might be buggy.” (Silver also called Morris’s attempts to explain elements of his model “jargony,” and sniped that “he claims to be too busy to provide a longer explanation.”
The well-placed source at ABC News said the differences between the current 538 model and others such as Silver’s model and the Economist’s came down to several factors. 538’s election forecast model is a prediction of election data based primarily on economic growth, polarization, the president’s approval rating, and, importantly, whether an incumbent is running for reelection. Those fundamentals give Biden a 3-point lead in the current model for two main reasons, according to the well-placed source:
- The 538 model more heavily weighs incumbency.
- The 538 model is suspicious of state polls further out from the election.
“We treat the polls with a lot more skepticism,” said the well-placed source .”We include a higher amount of potential polling bias, which pushes us back down towards 50/50, at least right now.”
I specifically wanted to ask Morris about other 2024 election forecast models because of his firsthand knowledge of some of the rival models. As a former senior data journalist at the Economist, he helped build its election forecast model. Yet the current Economist model shows Trump with a 75% chance of winning the election in November.
I asked the well-placed source why the Economist model which Morris helped build showed such different results, but the answer was not among those approved by ABC News. I also asked about what the well-placed source thought of personal rivalries between Morris, Silver, and other former 538 staff who have been at times critical of his model, but the answer was not among those selected to be shared attributed to the source.
The best answer I could get was that a well-placed source said that 538 tested various models, but ultimately settled on the current model because it produced the best results historically when tested against previous elections. Ultimately, the source said the primary difference was the distrust the model has in current polls.
“The final source of bias in the polls is they just might change between now and November. We think that there’s just more volatility in the campaign than these other forecasters. We are four months away from the election. Historically polls move around by about 7-8 percentage points on average based on historical average between now and Election Day. Biden is down 3 points in the tipping points states. Biden is up three in our model nationally. So it’s no surprise to me that if the model saw that information it would say ‘Hey, there’s a lot of uncertainty here,’” they said.
As for how 538 felt about Biden allies taking a rosy view of 538’s results, the answer from the well-placed source was also not permitted to be shared. Despite his model’s relatively positive results for Biden, Morris’ personal commentary has been less forgiving. “Even the bull case for Biden is pretty bearish,” he wrote on X on July 11, while also accusing Biden and his supporters of “hand-waving away the polling.”
On Wednesday evening, Morris deleted a tweet in which he expressed skepticism of the Democratic National Committee’s since-abandoned push to certify Biden as the nominee before the convention next month. Still, on Twitter, Morris publicly pointed out that “even the worst forecasts for Biden right now still give him a 1-in-4 chance of winning.”
