Dearest friends,
I will have a full length post soon enough, I think. I’m working on one. Economics based.
In the meantime, there is something on my mind. The problem with polls.
I got polled again today.
According to Pew, you have a 1 in 154,000 chance to be selected for one of their polls. Fair enough. Here’s the problem though. I’m polled every month or two. Either I severely, severely underestimate the amount of polls conducted in this country, or something is rotten in the state of Denmark, though probably not intentionally rotten.
Supposing I experience a 1 in 3 chance to be polled each month, that suggests there are something like 70,000 polls conducted a month. That seems improbably high to me, but perhaps it’s true. Otherwise, there’s some sort of strong sampling bias going on.
I can think of two major reasons why this might be so.
1. People don’t like to talk to pollsters. I like to talk to pollsters. A pollster will dial until they reach a halting point, which often means they’ll halt at me or a person similar in character to me. That’s not going to be representative. That’s going to be the kind of weirdos who love talking to pollsters.
2. The difficulty in building samples of certain demographics. I’m not just polled an unusual amount. I’m also invited to focus group new products for Big Bank and Fancy Corp. Why? If you’re constructing a sample and one of the components has to be people in their 20s of a certain net worth, then you’re not dealing with a lot of people! Clearly, too many of the yoofs are avocado toast loving wastrels. Polls try to fractionally distill into groups like income >$250k, Asian, Wiccan, Republican, Space Wizard, but these are not independent categories, but interact and form weird archetypes, such that if you’re looking for rich people, the kind of rich people you find will return radically different results. I’m not sure I’m being very clear, but people are archetypes, are types, are kinds of people. The man who owns a dealership may be just as rich as a BigLaw partner, but brother, they are going to be very, very different. So if you try to construct samples and you pick rich people, you can create really weird and bizarre answers depending on the arbitrary rich person you found that day.
Weep woo. Bean burritos.
A strange thing. And yet, polls are held up as scientific truth.
Call me maybe,
Monsieur le Baron
An update: Got polled again today.
Very random, much sample.
Maybe it’s because I don’t pick up my cell phone often for strange numbers and don’t have a home phone at all, but it’s been 8 years of adulthood and no polls. No jury duty even though I’m a registered voter. I’m not very good at being randomly sampled.
That’s quite the streak. How can polls be representative like this?