White Airbnb Hosts Earn More. Can AI Shrink The Racial Gap?
By Lane Lambert for Forbes
White people who host rental properties on Airbnb earn significantly more per year than Black hosts, but a “race blind” pricing algorithm could help close that income gap, new research shows.
Black hosts who rely on Airbnb’s algorithm to set enticing prices instead of manually choosing rates increase their occupancy rates significantly, bringing their earnings more in line with the higher rental incomes of white hosts, according to a study coauthored by Shunyuan Zhang, an assistant professor in the Marketing Unit at Harvard Business School.
Zhang’s findings come at a critical time for the travel industry. People eager to shake off their COVID-19 cabin fever are gearing up to take vacations, raising Airbnb bookings by 52 percent last quarter from a year earlier. Harnessing artificial intelligence to reduce racial economic disparities might help more property owners benefit from pent-up lodging demand.
The research results could also have broader implications for promoting racial equity at a wide range of businesses, from financial investment programs and medical services to e-commerce platforms like eBay and Uber. “If other companies invest in building similar race-blind algorithms, that may even the playing field for users,” Zhang says.
Without AI, white hosts earn more
Airbnb has faced criticism for disparities in revenue earned by white and Black hosts. For example, a 2017 study found that across 72 predominantly Black New York City neighborhoods, Airbnb hosts were five times more likely to be white. And the same study found that white Airbnb hosts in Black neighborhoods earned an estimated $160 million, compared to only $48 million earned by Black hosts.
Indeed, Zhang’s research found that prior to launching the algorithm in 2015, Airbnb’s white hosts made $12.16 more per day than Black hosts, according to the study, Can an AI Algorithm Mitigate Racial Economic Inequality? An Analysis in the Context of Airbnb, which will be published in Marketing Science this summer or early fall.
Properties owned by Black hosts had 20 percent less demand than similar white-owned properties, hurting occupancy rates and adding to the imbalance, say Zhang and her coauthors, Nitin Mehta of the University of Toronto, and Param Vir Singh and Kannan Srinivasan of Carnegie Mellon University.
This difference in the monthly occupancy rate “suggests that Airbnb guests may be systematically biased against renting from Black hosts,” the researchers write.
Based on detailed neighborhood and city data from 660,000 property listings in the United States, six years ago Airbnb began offering a free tool to help all hosts set reasonable rates. The real-time, “smart pricing” program uses machine learning to automatically adjust the property’s nightly fees by considering a rich set of factors that influence demand, such as location, amenities, and seasonal fluctuations.
Zhang and her coauthors were curious whether the algorithm might help reduce racial inequalities by optimizing nightly rates for both white and Black hosts, so everyone could benefit financially.
Narrowing Airbnb’s racial income gap
Zhang’s team analyzed a two-year span of nightly rental prices and daily income for 2,118 properties in seven US cities between 2015 and 2017. The researchers found that among all hosts who used the algorithm, the real-time price adjustments increased demand and revenue. Although hosts who used the tool saw nightly prices fall by 5.8 percent, higher occupancy rates helped lift average daily revenue by 8.6 percent.
Black hosts, who owned 9 percent of the properties the researchers analyzed, booked significantly more renters when they used the algorithm, shrinking the revenue gap between white and Black hosts for comparable properties by a whopping 71.3 percent.
By relying on intuition and other methods to set rates, many hosts of all backgrounds were overpricing their properties. The research team found that the algorithm’s price recommendations were not affected by the host’s race, so both white and Black hosts saw similar price corrections.
“One thing we can say: The algorithm is race-blind,” Zhang says. Overall, Black hosts who used the algorithm saw even more gains than white hosts who used it. “Black hosts do benefit more,” Zhang says.
Adoption is a challenge
However, the research team uncovered a problem that prevented greater racial equity gains: Black hosts were 41 percent less likely than white hosts to take advantage of the smart pricing application. And for the Black hosts who didn’t use it during the study period, the racial income gap persisted.
“This is a well-meaning algorithm,” Zhang says. “It provides flexibility. There is opportunity for a (Black) host to improve revenue, but only if you use it.”
It’s unclear why fewer Black hosts use the tool. Asian and Latino hosts also appear to use the algorithm at a lower rate than white hosts, though the gulf for those groups is less significant.
The research was partly inspired by Zhang’s own experiences with Airbnb, and by a previous study in which she analyzed whether adding professional web touches, such as expert property photographs, produce more rentals. (They do.)
If Black hosts appear more hesitant to embrace “smart pricing” in the latest study, Zhang says Airbnb’s best strategy is persuasion, rather than revising a clearly successful algorithm. “Encouraging hosts to adopt it might be more practical,” she says.
What else can Airbnb do?
Perhaps the bigger challenge comes in reversing the racial biases of prospective guests. Airbnb could consider preventing guests from knowing the hosts’ race, possibly by masking their profile photos until transactions are completed, the researchers say.
That could backfire if guests make stereotyped inferences based on a host’s name or home décor, the researchers warn, but they say it still may be an idea worth testing.
Airbnb underwent similar corporate soul-searching in 2016 after a study by HBS associate professor Michael Luca revealed how often guests with distinctively Black names faced discrimination (pdf). In response, the company conducted an internal study (pdf), modified aspects of its booking tool to prevent bias, and vowed to become more inclusive.
Airbnb certainly isn’t alone. Many companies are bound to encounter the complexities of addressing biases and racial equity challenges related to the use of algorithms, the researchers say.
“As AI and algorithms are gaining critical importance in enterprises, at least as this study attests, they can potentially be valuable tools in mitigating racial biases,” says Srinivasan. “However, a great deal of care is necessary to carefully ensure that indeed is the case.”