We’ve discussed repeatedly in recent months how poll results aren’t just “the numbers” in some hard, incontestable sense. They include a set of assumptions about the nature of the electorate. For most TPM readers, this is a fairly straightforward point that doesn’t require much convincing or explanation. But this post by a professor at Vanderbilt provides a really helpful real-world illustration. Josh Clinton takes sample data and shows that by using different reasonable and good faith assumptions about the electorate he can get results ranging from Harris +.9 to Harris +8. Don’t pay attention to the fact that these results are all still in her favor. The point is that the assumptions baked into the poll can yield results 7 points apart. It could as easily be Trump +3 to Harris +4. Again, it’s one thing to understand this in the abstract. But the specific explanation and the concrete outputs tell the story in a different way.
If nothing else, this is why that 7 point spread is just a bright flashing neon light that many of us are disregarding or not even seeing while we’re obsessing about win or loss margins of like half a percentage point.
