Polls? How very 2016. Now Google Street View AI scanner can predict how people will vote What does a car say about its owner? Uni researchers have managed to accurately estimate income, education, race and voting patterns for US neighborhoods by looking at cars on Google Street View. Nationwide polls like the American Community Survey (ACS) reach out to millions of people and can cost over $1bn each year. A form …
Pious (aka 'Prius') owners not necessarily Demo-rats (or non-Republicans) unless it has a politically charged bumper sticker on the back... (it might just be someone who drives a lot and wants to save money).
A former 'down the street' neighbor of mine had a Pious, and the back had plenty of politically charged bumper stickers.
As for ME, _my_ car is in the GARAGE all the time. 'Across the street' neighbors park in front of my house (they have too many cars or something). So much for your photo-based demographics!
But if anyone's asking, it's an 18 year old Ford Mustang convertible (with the top down nearly all the time). And I'm a 'conservative libertarian' (who votes Republican because libertarian candidates tend to be tin-foil-hat flakes that just want marijuana to be legal, and they're nearly won that battle now, so what ELSE are they going to do?)
But I'd like to have a pickup truck... and a motorcycle... and a lamborghini [ok probably won't get that].
(need a 'wrench' icon)
I live in a high mountain town in the Rockies where the pickup truck and hunters are everywhere, but I almost never see gun racks in the pickups. Perhaps it's mostly a 'Badass wanna-be' phenomenon, enhanced by Hollywood movies, in which case you might see more gun racks in Hollywood than in Durango. Or it could be a regional affectation centered in Texas or Montana. I've lived in the Southwest all my life and gun racks just aren't that common. In fact they are so rare that I note it when I do see one.
I just picked Durango because it's a known city name. I actually live in Pagosa Springs, not far from the Toltec & Cumbres line. In this town, little Durango looks like a big city. We don't have a rail line right here tho; they ripped them out when the logging stopped a century ago. Now all we have is a lot of hot water and a nice river draining directly from the snowfields above the town.
It wouldn't work, because apparently yanks can't tell what a Russian flag looks like.
apparently yanks can't tell what a Russian flag looks like.
This was a hilarious prank. "Americans Take Action" handed out the Russian flags with "Trump" on them.
It was supposed to point out the odd relationship between Trump and Putin. It worked because--as you observed--Trump supporters are intensely parochial.
I wonder how Brits would score in the same situation. Or Mexicans or Russians. Actually that would not be a fair test, because the US flag is so memorable, not to mention garish as hell. Betsy Ross was nobody's fool.
The Russian flag is merely three bars of white blue and red, and nothing else. What was the Tsar thinking? And now look how easy it is for their flag to be mistaken for a simplified version of their main rival's flag!
They need something in the middle to break up the monotony. Hammer and Sickle? No, that's out. Eagle? Already taken by Mexico and several other countries. Rearing Cobra? No, wrong culture. Well, they could always go with a boring agricultural motif like so many US states, oy.
Maybe they're better off with the simple tribar after all...
"Liberal Texans are still Texans"
Just because an area has a socioeconomic trend and a vehicle trend, doesn't mean the two are correlated with each other rather than both correlating to the area. Any more than republican areas having lots of gun nuts makes all gun nuts republicans. Sometimes it's just Texas.
Right. Nice data farming. So did the photos only include vehicles in the driveway, or also parked on the curb? And if curb parking was in the mix, did they attribute that to the general area, or a specific address?
What about condos, and apartment complexes where there is no available data as to which vehicle goes along with which tenant? What about part-time lodging? Or is this just about generally collating data of vehicles in a specific area? Because that should not be able to take into account work vehicles (landscaping crews, service screws, utility crews, etc).
And in my neighborhood, how about the off-road enthusiasts who get their unlicensed dune buggies and ATVs out of the trailer on Friday afternoon and then again on Sunday afternoon and race up and down the city streets at high speeds? Are they counted too? And just exactly which demographic are they attributed to?
Hey Google - pay attention to the digit I'm showing you. Rate that.
That ain't the least of it:
https://www.theguardian.com/politics/2017/feb/26/robert-mercer-breitbart-war-on-media-steve-bannon-donald-trump-nigel-farage
It would seem that Google and Facebook are the monkeys and Cambridge Analytica is the organ grinder.
People in a given socio-economic are show their preferences by their lifestyle and vote accordingly. Last year's presidential election was thus correlated by the study. Interesting.
Now explain how the government changes sides every few elections and how to predict that despite the fact that people generally change cars less often than they vote for president.
"people generally change cars less often than they vote for president"
Well... It's a thing of averages, of course, on both sides.
And it made me think about my own car use versus my own presidential-election participation...
The first car I actually owned, I owned for about a year (but previously driven it while it belonged to my dad, for a total *use* time of about three years).
After that, a gap of about four years, then a car I bought in 95. I eventually dragged it, kicking and screaming, to a scrap yard in 2000, that is five years and 137000 miles later. (Even after that, it ran fine, if the weather wasn't too wet.)
I replaced it with one that lasted from early 2000 to early 2004, when I parked it with its roof on the ground and its wheels in the air. (I don't recommend this, by the way.)
The replacement for that lasted from early 2004 to early 2009, when I *sold* it.
So, the average is a bit over four years.
The analysis falls down when you consider that I've never voted for or against the president of anywhere, because although I've lived in two different countries that have presidents, I've never been a citizen of either.
I completely respect the fact that you have an average vehicle ownership time of "a bit over four years". Obviously, in your particular case, my analysis does indeed fall down.
Now, for my analysis to "fall down" globally, you'd have to decide that most people have the same average ownership time. You will allow me to doubt that, if only for the fact that I know people who change vehicle every two years. In my particular case, I have owned an Open Kadett for 11 years, a BMW 330d for six and a half, and my latest is an Audi A5 which I bought in 2012 and I am nowhere near thinking about selling it. That is among a few other cars that have passed in my hands more briefly, I freely admit. So my average is significantly higher than yours, even considering the Laguna 2 I only had for 3 years.
I will not, however, consider that my ownership times are common either. I base my opinion on the fact that are more poor people that own cars than there are rich ones, and if the rich people do indeed have the means to change cars regularly, poor people don't. Rich people can choose how long they want to own their car(s), poor people make do with theirs until they don't have the choice any more.
I do believe that that tends to bring ownership times to significantly over a presidential mandate. I will be quite happy to be directed to some actual statistics drawn from a population study that demonstrate that this belief is wrong.
“But what if there was a way to get the information without having to deal with mountains of paperwork or bothering people at all?”
Easy: just make it up. Be about as accurate and informative, wouldn’t it?
But what if there was a way to get the information without having to deal with mountains of paperwork or bothering people at all?
Either I, or the article author, have a fundamental misunderstanding here. This analysis doesn't replace things like the ACS survey, it builds on them!
The only way the guessing of things like income, race and voting preferences from vehicle choice, house style, size of garden, location, etc work is because of surveys like ACS providing the correlation information! Sure, this analysis of street view images might help for spotting some changes (such as gentrification of an area) earlier than waiting for the survey but if the surveys didn't happen the guesses would rapidly get out of date!
I have not read the paper, only the article. I would hope this point is explained in the paper.
Will it really go out of date?
The current data was used to build the algorithm, but if the demographics change in the future, well, that's why the machine is learning patterns and not addresses, right?
I mean, if in the future it sees a cluster of beaten-up pickups it will guess Republican, even if the image comes from Silicon Valley.
Apparently the Trump voting campaign knew exactly which neighborhood to target with which slogan, on what channel. They could have got this targeting accuracy by letting 150 analysts go over the most recent ACS results for a month; or by letting maybe 5 people do it in a few days, based on multi-source information, like cars and tweets, with the help of serious computing. They chose the latter and got successful.
To me though, it's eerily "The Circle"-like; it's as though anyone with enough data analytics knew you like your long-time neighbor does, maybe better; and you don't even know who they are, or why you keep reading what they want you to read on Google news. Kinda scary.
Here is why: https://www.theguardian.com/politics/2017/feb/26/us-billionaire-mercer-helped-back-brexit
Apparently Trump paid 6M to a spinoff (from a 3 letter agency, most likely CIA) specializing in subversion of the democratic process with extensive experience in "altering election results the way we want them". That is where he had the data, methodology and instructions to act upon.
I wonder:
1. What were their references. How many elections did we interfere with - successfully or unsuccessfully?
2. Which retarded cretinous monkey decided to spin-off this as a private venture. What's next? Spin off of the nuclear weapons? Or the old disused USA VX gas production line? Or maybe the SmallPox stocks kept by the NIH?
The only reason we know about this is because the same spinoff worked on behalf of the Leave campaign "for free" in the UK BrExit referendum and this counts as an undeclared donation of services. Though I bet it will be swept under the carpet instead of being investigated (because of 1 above).
So as many other posters have pointed out - it determines the neighbourhood based on what cars are visible on the street; whilst the local DMV has (legal) vehicle ownership all in its computer (as do insurance companies, various marketing organizations, etc).
As for people's income, and say charitable/political giving, mortgage, property taxes paid - it's all in the IRS database; some of which is provided as statistical geodata (every year when I file, the tax software shows me how I compare for the county and state for each 1040 section and gives me warnings if my entries are significant outliers).
Amazon and google probably have a pretty good idea as to what car I drive from the parts I buy and web searches against every maintenance / repair job needed.
Hmm, this household, despite having a net worth above $1Mil, has never exchanged vehicles less often than six years, and have kept several for around a decade. Come to think of it, that might be WHY we are millionaires. Buy solid no-frills modest cars, keep 'em maintained. Current vehicles: 2008 Scion, 2009 GM compact, bought new and kept up. The record was a Honda Civic wagon that went fourteen being driven nearly 100 miles a day.....
Same here. Current cars are a 2005 Mazda 3 and 2014 Honda CRV. My attitude has always been buy the cheapest/best vehicle for the job, pay cash and drive it till it drops. Our house is modestly sized and in a modestly priced middle class neighborhood. My father-in-law was a multi-millionaire and the most expensive car he ever bought was a Toyota Corolla. I think a good many people with engineering backgrounds think that way.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
