Hi all,
Happy February! Nate Silver (for the uninitiated — the biggest name in the world of quant elections analysis) released his latest pollster rankings. And it went great for us!
Silver looks at polls conducted in the immediate run-up to the election and does a ton of math to see who is the most accurate. This is no easy task: He attempts to account for a ton of relevant factors (e.g. how tough was this race to poll, how close was it to the election, does a pollster publish appropriately messy numbers or hide behind the average, etc etc). Then, every cycle, he ranks them.
The 2024 election was my first cycle as a pollster — and NPI’s first cycle with our new team in place. Here’s the firm’s stats prior to 2024 and with this cycle included:
We overhauled the firm’s strategy on pre-election polling — and got big returns! Some basic tenets of what we did:
Settle on a sound methodology ahead of time and stick with it.
Trust the process — don’t get too hung up on a result that “looks weird.”
Go into races we think are interesting or tough. Not just over-polled swing states.
Don’t pay that much attention to what other people are publishing (though it’s impossible to completely tune it out)
And it seems like it worked! We jumped about 90 ranks (!) and went from a positive predictive +/- to a negative one (negative is good in this system). Our overall grade rose significantly. And our herding penalty dropped dramatically (more on what herding is here).
So maybe my initial self-assessment was too negative. According to these metrics, 2024 went really well!
As always, I appreciate all of you!
David
Congratulations to you and your firm for the good work you are doing and receiving some validation of that work from Nate Silver.