by Yanni Loukissas and Firaz Peer,
Program in Digital Media; School of Literature, Media and Communication
Every day, Zillow.com, an online housing marketplace, reassesses the value of your home. In fact, the company is continuously updating its ‘zestimate’ — the name for Zillow’s proprietary approach to generating a market-based estimate — for about one hundred million houses and apartments nationwide, whether or not they are for sale. Their business relies on data from public records as well as privately held multiple listing services. Because of the widespread availability of these data, Zillow doesn’t need complex financial models of the housing market to assess your home’s worth. They can use simple algorithms to fit the details of your property to comparable listings in the same area. Zillow doesn’t create these listings or sell access to them. Like most successful web companies, Zillow survives off of advertising revenue—a kind of surplus value created when it aggregates existing data into a new form of context for understanding the housing market.
The value of property has long been assessed in relation to its context; however, the scale and visibility that Zillow provides is unprecedented. Their zestimate is what Nick Seaver calls an “algorithmic system.” It combines computational modeling with human steering from both experts and the crowd (you can also update the details of your home’s listing). It is one of many examples of how data might take a more direct role in shaping perceptions of property value, and thus development. From Zillow’s perspective, the future of housing — and of cities — will be shaped by Big Data.
In Atlanta, this version of the future is not going uncontested. Trent, an Intown real estate agent, confronts an uncertain outlook for his job. How can he continue to justify the cost of his services (commission in Metro Atlanta is typically 6%) at a time when almost anyone can access listings for sale and rent online? “I can’t hold data hostage,” he jokes. But Trent’s situation is serious, and one he equates with the circumstance of the travel agent a few decades ago. Orbitz, Travelocity and Expedia, among others, have all but put an end to that vocation. “In the past, someone needed my services,” Trent recalls of his early days in the business just ten years ago. “Buyers and sellers wouldn’t know what houses were on the market without agents.” Today, Trent must find leverage elsewhere. It is no longer access to data that realtors provide, he argues. Rather, it is context: “the context necessary to understand what it might be like to actually live in a neighborhood or an apartment complex.” From Trent’s point of view, access to data isn’t going away — but local agents will play an important role in interpreting it.
Both the developers of Zillow and the agents that resist its encroachment into real estate believe that making sense of housing data requires an understanding of context. But they disagree on what context means. In terms outlined by Paul Dourish, Zillow’s definition of context is “representational,” relying on statistics and algorithms to assess the housing market. In contrast, Trent’s definition is “interactional.” His sense of what a house could go for on any given day is contingent on the dialog he is able to establish between buyers and sellers. In an interactional model, writes Dourish, “context isn’t something that describes a setting; it is something that people do.” As such, the interactional context of the housing market can vary enormously depending on who you talk to and when.
Oscar, an organizer based in Atlanta, talks mostly to people of color, renters who have been driven out of communities they grew up in. These residents are being priced out in the immediate sense and, ultimately, pushed out by financial speculation and gentrification. In Atlanta, there are almost no regulatory policies that protect low-income residents from the inevitable outcomes of a market on the rise. “A crisis is hitting renters. We need data to declare a renters’ state of emergency,” asserts Oscar urgently. He sees a broader context for the data available on what is for sale or rent; they are only part of the picture. “Whenever any information or data is created, it is created in the interest of a group,” he explains. “Zillow serves the wealthy.” But a critical reading of existing data isn’t enough to change the tide. Oscar needs “counter data” to fight gentrification. “How can we produce data to serve the oppressed?” he challenges. Oscar contends that collecting data on “the truth of the system” can give rise to a new sensibility for Atlanta’s development. “We need data on how many people are being displaced. We need data on their mental, emotional, and physical health. Who’s being displaced and what is the consequence of that? We need data to show that there is mass displacement that is causing great suffering.” Part of Oscar’s work is filling in that missing context, which shows that the 2007 housing crisis is not over.
While technologists and realtors are working to define the context in which their clients might make the best possible choices in a local market, Oscar seeks to reveal another condition: one which calls the logic of the market into question. These three ways of approaching the question of context — aggregating Big Data, interpreting that data using local knowledge, and generating counter data — are competing strategies for imagining the future of Atlanta. All of these strategies implicitly accept that data are now a necessary medium for understanding urbanism, which has reached a scale that would be difficult to contemplate otherwise. Nevertheless, data have become a site of contestation, which will determine how cities evolve and for whose benefit.
With support from the Georgia Tech Center for Urban Innovation, the Local Data Design Lab is intervening in this contest by prototyping new tools for thinking critically about data and their role in urban change. Our tools are meant to serve education, journalism, activism, and even art — practices that have the power to reshape the social image of the city. In a course hosted by the lab in the fall of 2015, students worked with our new code library to reflect on what gentrification in Atlanta looks like through data. They developed projects to test a variety of approaches: analyzing implicit discourses on gentrification embedded in Zillow listings, visualizing patterns in home values affected by the new BeltLine, and gaming out the implications of luxury development for the sustainability of low-cost housing. As our underlying toolset is refined, the lab will work with communities in Atlanta to question housing data and the context for their interpretation. It is not enough for data “to be free” — a platitude opined by Stewart Brand in the late 1960s and appropriated by enterprising technologists ever since. The Local Data Design Lab is building new capacities for data literacy, so that not only housing experts but the broader public can confront the social and political implications of this medium for shaping perspectives on the future of the city.