The "sharing economy" is a burgeoning business model in which people offer their personal belongings and personal services to others, usually through online marketplaces that facilitate the transactions.
It seems absolutely egalitarian at face value: Anyone who owns a sometimes-unused thing—an apartment, a car, a boat—now has an easy way to advertise and share it. And anyone with the time and skills—driving, running errands—can find customers who need these services. But new research shows how online marketplaces can work in ways that are anything but egalitarian: They can inadvertently fuel racial discrimination.
In Digital Discrimination: The Case of Airbnb.com, Benjamin G. Edelman and Michael Luca investigate the possibility of racial discrimination against people who advertise properties on Airbnb, a popular online marketplace that lists temporary rooms and homes in some 34,000 cities across 192 countries.
The researchers found that black hosts were charging approximately 12 percent less for rental properties than were nonblack hosts.
Analyzing a data set that focuses on New York City (the company's biggest market), the researchers found that black hosts charged approximately 12 percent less for rentals than nonblack hosts—even when the properties were equivalent in terms of location and quality.
How did the researchers find out the race of each host? The same way potential guests do—they looked at the hosts' profile pictures on the Airbnb website. In fact, large profile pictures in online marketplaces were the inspiration for the research.
"In the early days of the Internet, online marketplaces like eBay were relatively anonymous," says Luca, an assistant professor in the Negotiation, Organizations & Markets unit at Harvard Business School. "You just saw a seller's ID with a seller score, or a buyer with a buyer rating."
These days, however, profile pictures are commonplace. It's a phenomenon that Edelman attributes to Facebook. In fact, many sites require customers to use Facebook for a login mechanism—not just marketplaces but media outlets as well. For instance, ESPN.com and USATODAY.com both ask readers to sign in through Facebook in order to post a comment on a news story. "If you log in to a site through Facebook, they get your profile picture, among other things," says Edelman, an associate professor and Marvin Bower Fellow at Harvard Business School. "Now, rather than standing on its own, your comment is juxtaposed with a picture of you."
While pictures have the potential to foster a sense of community in an online marketplace, they also enable racial discrimination. Thus, the researchers set out to find out whether potential discrimination was affecting the online marketplaces of the sharing economy.
A Marketplace Leader
Edelman and Luca focused their research on Airbnb for a couple of reasons. One, the San Francisco-based company, founded in 2008, is among the most successful sharing-economy marketplaces in the world, with 500,000 listings and counting. (By way of comparison, hotel industry giant Marriott offers around 530,000 rooms worldwide.) Airbnb, which has been valued at roughly $2.5 billion, recently closed a $200 million round of financing, according to a February 6 SEC filing.
Two, the site requires hosts to post large profile photos, which are displayed next to the pictures of the rental properties. "You can't be a host without a profile photo," Edelman says. "Sure, you could put a picture of your dog or the Eiffel Tower instead of a picture of you. But if you have a photo of a dog where your face should be, it might deter people from renting your property."
The researchers' data set consisted of all Airbnb New York rentals available on a randomly selected day in July 2012. For each listing, they recorded information including the asking price of the rental, the characteristics of the apartment or room, and the average rating each host had received. (The site's structured rating system lets guests rate the hosts and properties on location, check-in, communication, cleanliness, and accuracy.)
The researchers also hired a team of workers through the crowdsourcing marketplace Amazon Mechanical Turk (AMT), a popular source of data processing assistance. The workers examined the property photos on each listing and rated them on a seven-point scale from "This is a terrible apartment; I would not stay here at any price" to "This is an extremely nice apartment; I would stay here even if it were a lot more expensive than a nice hotel room." With these ratings, the researchers controlled for apartment/room quality as seen by a user examining an Airbnb listing.
Other AMT workers were hired to look at the public profile pictures of the New York hosts, coding them into one of the following categories: white, black, Hispanic, Asian, unclear but nonwhite, multiple races, unclear/uncertain, or not applicable (because the host had posted a nonhuman picture of, say, a dog or the Eiffel Tower).
Even controlling for factors including perceived quality and location, the researchers found that listings with nonblack hosts earned roughly 12 percent more than those with black hosts for apartments with similar ratings and property photos. "Moreover, black hosts receive a larger price penalty for having a poor location score relative to nonblack hosts," the researchers write in the paper. "These differences highlight the risk of discrimination in online marketplaces, suggesting an important unintended consequence of a seemingly routine mechanism for building trust."
To Airbnb's credit, the company's terms of service forbid content that "promotes discrimination, bigotry, racism, hatred, harassment, or harm against any individual or group." Hosts are responsible for setting their own prices for the properties they advertise on Airbnb, so it's not as if the company is explicitly biased.
But the "pricing" section of the website advises that new hosts set rates with market demand in mind. For example, Airbnb tells new hosts they "may want to charge lower than average rates to attract travelers comparing your place with existing reviews. Once you have a review or two, adjust your rates as needed."
The researchers interpret black hosts' lower prices as evidence of a similar, but permanent, disadvantage. In particular, they maintain that their findings show how black hosts have a harder than average time attracting travelers, and therefore must keep their prices lower than average in order to compete successfully.
"Airbnb's pricing suggestions implicitly acknowledge that price is a market-driven outcome," Luca says. "In a market, prices move until supply meets demand. And the data indicate that discriminatory preferences are factoring into prices."
Advice For Sharing-economy Companies
Thus far, Airbnb has declined the researchers' offer to visit the company and discuss possible mechanisms to facilitate trust without encouraging unintentional discrimination. But they do have recommendations, not just for Airbnb but also for any online marketplace in the sharing economy.
For starters, Luca and Edelman suggest evaluating whether profile pictures provide any necessary information to the customer. "For instance, ask yourself what useful information you learn from looking at the Airbnb host's face," Edelman says. "The last time I was an Airbnb guest, I never laid eyes on the host except on the website. There was a neighbor facilitator who dropped off the keys and picked them up."
In any case, the researchers recommend putting profile pictures in a less prominent place, perhaps on a separate page from the picture of the product or property. (On Airbnb, the host's profile picture is the second-most prominent element on any listing page, displayed next to the default picture of the property.) "Research has shown that when something is more salient, you respond more to it," Luca says. "This is something a platform should think about in its design decisions. Is a host's physical appearance really the information Airbnb most wants to emphasize?"
More generally, the researchers recommend that sharing-economy companies take website design seriously and to question early on whether their system design encourages racial bias. "These questions are too important to be left to engineers, and too important to be left to graphic designers," Edelman says.
As they continue this line of research, Edelman and Luca are studying some of the online ride-sharing services of the sharing economy. Such services are especially relevant to racial bias studies, they explain. Booking a cab online has the potential to eliminate the possibility that a driver will decline to pick up a black person who is hailing a cab. But racial bias will still be a problem if the customer is required to provide a profile picture when signing up for the service.
"In so many contexts, online information systems can be much better than the real-world alternatives than they replace," Edelman says. "But they can also be much worse. The burden is on us, as system designers, to get it right."
Note to readers: Edelman and Luca are interested in conducting experimental studies with firms that operate online marketplaces.
"Online marketplaces are a great innovation with extraordinary potential, but they also present important challenges, as explored in this paper. We hope companies will design their sites to facilitate transactions and build trust while avoiding unintended consequences such as discrimination," Luca says. "There's a great opportunity for an interested company to take the lead on publishable research in this area. We're willing to help implement changes based on behavioral economics principles, including finding ways to reduce discrimination without hurting the bottom line."