Thanks for finding the numbers. They make a big difference.
The OP makes it sound like the numbers are 95% confidence intervals rather than a small margin, as if, say, only 5% of liberals owned a fishing rod or some such. Instead, there is just slightly more chance a that a rod owner is conservative. Statistically significant, but not the massive predictor implied in the original article.
This surprised me too. I say this as a liberal who just returned from having lunch at CFA, having discussed whether we should go to Arby’s the next time we get fast food. As with most statistical models, there is a non-zero error rate.
And the fast-food brand most closely associated with being conservative? Arby’s.
Roy Rogers has got to be a close second.
So, how do we know that the AI isn’t just doing automatated P hacking on a massive scale? Finding random anomalies rather than sound trends? Or is that actually fine because marketers don’t care about causality, only correlations that give them even a slight edge in targeted marketing?
This. It’s just ‘fishing tackle’. Adding to the scores with different types of fishing gear is like saying ‘has a vehicle with 1 or more wheels’; ‘has a vehicle with 2 or more wheels’; etc, and scoring each one separately/additively. It’s not just machine learning that needs improvement here.
The original paper is linked to right in the article:
What is an English Muffin? I’ve always wondered.
I think it’s more of design flaw. Like a lot of posters are pointing out, many of these things are pretty tied to geography, and most of the markers for liberals are lack of things they’ve tied to conservatives.
We already know that liberals are clustered around cities. You know what isn’t, oceans and streams and abundant storage space. So is owning a fishing rod a marker of conservativism? Or are conservatives clustered in areas where fishing is more common/practical?
Or hell. Fishing is an expensive hobby, and wealthier people are both whiter and more conservative.
I think the best equivalent in Britain would be a crumpet.
The ones sold in supermarkets are uniformly awful. To get a good one you have to know a muffin man.
It’s a little round yeast roll cooked on a griddle.
A lot like a crumpet but browned on both sides.
It can be. Yet, here in UK coarse fishing was (some time ago, but probably still true in some segments) often a pastime of the working class more than others.
I found it strange they included multiple entries for fishing gear but nothing about guns. I know quit a few liberal folk that love fishing but are anti hunting.
Marketers don’t care about correlations either. They care about sufficiently impressive sounding woo to attract clients.
Now you’ve done it.
Or, you know, a muffin.
…which to me is nothing like a crumpet. Crumpets have all the little holes in. I don’t want those in a muffin.
Yeah, looks like gun ownership is way more predictive than fishing poles. Something seems off about the selected stats.
No. No, no, no. Same old garbage.
THESE BRANDS ARE MARKETED TOWARD PARTICULAR GROUPS TO BEGIN WITH.
They speak of these things as if everything is just thrown out there by chance. How about MyPillow? How much you want to bet ownership of a MyPillow product is an indicator of conservatism? Well, duh. I bet owning a MAGA hat is an excellent indicator.
American culture has generations of conditioning through racist and predatory advertising techniques. Franchises have rules about where and how and by whom they are opened, and are advertised to particular groups, too.
How can anything like this really seem illuminating? At best it is predictable, at worst it’s just more bias. It’s like a snake eating its own tail. They’ll probably sell this data back to the advertisers and franchisers who created their results to begin with.
Is that what they are? Muffins like they were when I was a kid, before that day when they mysteriously turned into small cakes with blueberries or chocolate or whatnot?
I never liked those. Bland and stodgy. Crumpets are way better.
So do English muffins as sold in Canada. One characteristic is that they can be split into two even halves with a fork. Are crumpets or muffins in England the same?
Yup. Or said another way, “garbage”
“Studies” like this are so useless that they make all of us actively stupider by reinforcing stereotypes. You can’t “predict” anything without a causal mechanism established. Finding tiny p-values-worth of correlation significance in huge noisy data sets is masturbatory psychology bullshit that does nothing except generate clickbait headlines.
I really wish media outlets wouldn’t give press to worthless science like this. I hate to even call it “science”.
We’ll put it this way. I’m a working class person who owns multiple fishing poles. I have not used them in like 5 years.
“Expensive” doesn’t neccisarily just mean cost. I can catch my own bait and I already own the equipment. I know the good spots fish from shore. I do not have the time to regularly go fishing.
And I’ve run into a lot of self identified working class people who throw fishing around as an identity markers. “Working class” people who own multiple homes, a $140,000 pickup truck and more than one boat.
So traditional associations of this sort run into the same sorts of correlation mess. Fishing likely was often a pass time of tenement dwellers and textile workers of the past. And even today it can be a cheap way to get a meal or pass an afternoon. Most of the recreational fishering you do see in NYC (for example) is immigrants and people living below the poverty line.
But population wide? Who will more often have the time and money to regularly fish (or take up fishing)? And are they more likely to be in dense urban centers or less dense places on bodies of water?
