Cars have long been status symbols – if you have a fancy car, people might think you’re rich.
While flaunting ritzy items is usually a practice of one-upsmanship to impress neighbours, a new study finds that the types of cars found in a community might also suggest very real facts about an area’s income level, demographics and an existence of inequality.
For example, the more foreign cars in a neighbourhood, the higher the average income in that neighbourhood. That might not sound too surprising – but the surprising part is that all this information was gathered not by hand but by algorithms combing Google Street View. This new approach has big implications for how large-scale income data can be collected in the future.
A Google Street View car in Vienna in July 2017 (Credit: Getty Images)
The study comes from Stanford University in California and was published last month. For the study, the team built an algorithm that combed through 50 million Google Street View images from 200 US cities. These algorithms recognised the year, model and make of every car in every image, even if they were obstructed by other objects or photographed from an odd angle. Next, those findings were compared with data from the American Community Survey, a $250m door-to-door study that rounds up data ranging from gender to education to unemployment, as well as voting preferences.
With this comparison, the researchers could accurately predict a host of demographic identifiers such as income, race and even political views – all from the types of cars found in each community.
American cars made domestically, specifically Buicks, Oldsmobiles and Dodges, were associated more with lower median household incomes
The study found that German and Japanese cars (Lexus in particular) were found in areas with high median household income. Meanwhile, American cars made domestically, specifically Buicks, Oldsmobiles and Dodges, were associated more with lower median household incomes.
Google Street View gives an in-person view of points on a map, but now it might help researchers predict income data for neighbourhoods (Credit: Alamy)
Using powerful AI to analyse images of cars on Street View could provide a faster, cheaper way to pinpoint where inequality might hit the hardest. For example, Chicago is starkly segregated by income level, with cheap cars and costlier ones separated in big swaths across the city, while Jacksonville, Florida saw the least economic segregation, judging by the distribution of cars.
Jacksonville, Florida saw the least economic segregation, judging by the distribution of cars
This is a big deal, too, because surveys and censuses that usually find out that kind of data (like race and political leanings) are labourious, time-consuming and expensive.
Timnit Gebru, lead researcher for the study, says that since the study has been published, the team has been contacted by political scientists who wanted to use the data.
“[They] could potentially augment the information we have with things other than cars, and get more accurate and timely results to perform demography,” she wrote in an email.
While the findings give credence to the fact that the type of car you drive may suggest certain things about yourself, such as wealth or class, Gebru offers caveats, too.
What I would like people to take away from this study is that social scientists could potentially use images to study culture – Timnit Gebru
“I don't think that all countries have the associations between cars and people that exist in the US,” she says. “In other countries, there might be a strong relationship between certain garments and other expressions of culture.”
Google also employs human "trekkers" to walk around roads that cars can't reach in order to capture images for Street View (Credit: Getty Images)
“What I would like people to take away from this study is not necessarily that there is a strong relationship between cars and people – it is that computational social scientists could potentially use images to study culture,” she says. The machine learning like the kind she and her team used could act as a complement to demographics surveys that already exist.
Outside of matters of money and inequality, other countries have been looking to use AI and public images, like the ones from Google Street View, to learn more about the people who live there in ways that traditional surveys or censuses could measure. For example, Canadian health researchers are also analysing Google Street View images to study the relationship between certain diseases and the greenery or pollution levels of certain neighbourhoods.
And as long as there are tech companies out there snapping photos of our lives, houses and even cars, our salaries and secrets may be more accessible than we think.
If you liked this story, sign up for the weekly bbc.com features newsletter called "If You Only Read 6 Things This Week". A handpicked selection of stories from BBC Future, Culture, Capital and Travel, delivered to your inbox every Friday.