Thread by @niketasays

Hi y’all, tonight this former-math-teacher-turned-public-policy-junkie is going to break down why really expensive polling is going to tell us jacksh*t about this upcoming election. [A thread.] chicago.suntimes.com/news/mayoral-r…

The absolute most important part of this poll (and most others) comes in the final page of the memo issued by polling organizations, the nerdiest of the sections on methodology.

If I were using this to plan a mini-lesson on statistics, this is where I’d introduce a game of “bias bingo,” in which we’d find all the different ways the 644 voters polled were not an adequate snapshot. But let’s start with a few basic reasons.

1) Polling registered voters.

In a same-day registration city, polling registered voters skews results even more than usual, esp if insurgent candidates are trying to mobilize new voters.

See also: @AmaraEnyia’s voter registration event on a Sat night at a bar in Wicker Park.

2) But let’s be clear. Those 644 people who responded to calls aren’t just REGISTERED voters, they’re deemed LIKELY voters.

What establishes “likelihood”? Past voting record.

Usually in campaigns that number is unlocked by voting in the last FIVE out of FIVE elections in a row.

Ever notice you start getting aggressive mailers at a *certain* age? For me it was 29, when I finally voted in enough elections for campaigns see me pop up as a worthwhile target. I wasn’t an unlikely voter before — the algorithm just didn’t see me until that magical age.

3) Then let’s deconstruct what margin of error means.

It does NOT mean they’re only off by +/- 3.88 points.

It means there’s a 95% chance they’re only off by 3.88 points.

What about that other 5%, you ask?

That is a 5 percent chance that they are off by more or less than 3.88 points ...

For known, LIKELY

and REGISTERED voters

who trust polling enough to answer the phone.

Now add up all that bias. Who doesn’t show up?

- Young folks
- Transient folks
- Folks who historically don’t trust random calls asking personal info
- People who don’t vote EVERY election but show up when the election is hot
- People who will register on the day of

Given all this, given Millennial turnout out in Nov, given FBI charges, given how much can happen in 30 days,

for the first time in a really long time,

we don’t really know anything about what’s going to happen.

And that’s kind of a really cool thing.

(End)

UPDATE: There are lots more math nerds out there caring bout the polls. YAY! So here are a couple things I glossed over last night:

50% of respondents had landlines 😳😳😳!!!!

And as if to know that their sample was so skewed, they didn’t publish how many of each age/race/gender they polled.

So while they had a poll of 644 ppl, we have no idea how many ppl within each demographic they polled to make predictions of how “women” or “Black people” feel.

And in case you don’t think this is important, check out how one 19 year old Black male Trump supporter in southern Illinois influenced national polls in 2016: nytimes.com/2016/10/13/ups…

For all these reasons, don’t trust the polls in this race.

Instead, vote for what YOU want our city to look like! Our process is openly biased towards racial equity, but we’re using this to build a citywide vision for our future.

chicagoforall.org // polls close THURSDAY!