Donald Trump is reportedly considering announcing his 2024 bid tonight during a rally in Ohio with Republican Senate candidate J.D. Vance, just hours ahead of Election Day.

It was reported last week that Trump was considering announcing his 2024 bid sometime this month, most likely in the week after the election. That timing is key in signaling just how quickly the national conversation will switch to 2024 post-midterms. Republican New Hampshire Gov. Chris Sununu summed up reports of Trump’s plan to announce before the new Congress is even sworn in: It’s a “terrible idea.”

Today reports surfaced that Trump is actually planning to move up his announcement to tonight. And Republicans are now reportedly even more in sync with the “terrible idea” assessment.

With the election one day away I wanted to take a look at the latest polls. They still don’t tell a totally clear story. Big picture what we see is still much better for Republicans than what we saw in late summer or even as recently as October. All the key Senate races are more or less tied. That means anything from a one or perhaps two seat pick up for the Dems to a four seat pick up for the GOP is entirely plausible. But with all this sobering news we’re not seeing the kind of late polling breakout I might have expected. The generic ballot averages have actually ticked slightly back in Democrats’ direction over the last couple days, though this could well be statistical noise.

There are going to be big political winners and losers on Tuesday. But there are also going to be big polling winners and losers. Through this cycle different classes of pollsters have been seeing a very different race. Wednesday morning we’re going to know who was right and wrong. But here’s an aspect of polling that doesn’t get talked about enough. It’s not just accuracy. There’s another part of this.

We know that polling has gotten both harder to do and more expensive to do as fewer people respond to polling phone calls. If one out of twenty calls gets answered, that’s a much more expensive proposition than if one out of three does. So non-response has been driving up the costs of polling, and that’s overwhelmingly hitting the pollsters who use live callers. Live calls are generally considered the most accurate, though it’s far from certain whether that’s still the case. Non-response also puts accuracy under growing strain because pollsters need to make sense of which political groupings are more or less likely to respond. If non-response is identical across all political affiliations it’s not a problem from an accuracy perspective. But that’s almost certainly not the case.