AI could soon tell you the future value of your home

• The future of home prices will shape the financial fates of many Americans.
• An AI tool from MIT can accurately predict how home prices would change over time.

If you're one of the 63% of US adults who own a home, your future wealth and retirement age could hinge on one question: How much will your home be worth in the future?

According to an MIT researcher, AI might be able to provide you with an answer.

Writing for the Financial Times, Carlo Ratti, an urban technologies and planning professor at MIT, detailed the capabilities of the real estate bot he and his fellow researchers have been developing. To date, Ratti said the AI has been trained using Google Street View images — in particular 20,000 houses around Boston — and data on how their prices have shifted over the years. The researchers used a deep learning model to help the bot identify connections between homes' visual features and their values.

The results: The bot can look at a house and predict its price with "striking accuracy," Ratti said. But it might be more than just an improved Zillow, whose Zestimates have come under greater scrutiny in recent years.

"When other variables were added — such as structural information and neighbourhood amenities — our algorithm was able to make very accurate predictions of how prices would change over time," he wrote.

In 2020, home equity accounted for 28% of US household wealth, per the Census Bureau. Only retirement accounts had a bigger share at 36%. For those nearing retirement age, home values are even more important. Roughly 80% of Americans age 60 and over are homeowners, per a February Vanguard report, with housing accounting for 48% of their median net worth. By selling their home and relocating to a cheaper housing market, the median homeowner over 60 can cash out roughly $100,000 in equity, the authors estimated.

But these wealth gains are only possible if one lives in a home whose value rises considerably over time. In the future, AI tools like the one at MIT might help prospective homeowners predict where the biggest gains are likely.

These technologies could come with some risks, however, Ratti said. For one, they could have a set of biases, perhaps undervaluing homes in predominantly black neighborhoods.

Someday, if more is learned about what visual features produce high home values, homeowners could begin making a series of seemingly inconsequential changes to their homes — all to appease the bots.

"Imagine a dystopian future where everyone repaints their house a specific colour to game the system and impress the Google Street View Car," Ratti wrote.