Study Finds Craigslist Rental Listings in Poorer, Nonwhite Neighborhoods Contain Less Information about Homes and Neighborhood Amenities

An article in Housing Policy Debate, “Housing Search in the Age of Big Data: Smarter Cities or the Same Old Blind Spots?,” finds that the amount and type of information contained in Craigslist rental listings varies with neighborhood demographics. Compared to listings in lower-poverty neighborhoods, listings in higher-poverty neighborhoods contain more information about renter requirements and less on the rental homes or their amenities. Listings in predominantly Black or Hispanic neighborhoods, regardless of poverty level, contain less information than listings in whiter neighborhoods. The authors argue that limited online information about poorer and nonwhite neighborhoods may preclude some home-seekers from learning about them, contributing to neighborhood segregation.

The authors analyze two collections of Craigslist rental listings: a dataset of 1.4 million geolocated listings created between May and July 2014 and a dataset of 1.7 million geolocated listings created between May 2017 and February 2018, both restricted to the 50 largest U.S. metropolitan areas. The authors’ previous research on the 2014 database showed that relative to the expected volume of listings that vacancy rates would predict, over half of majority white neighborhoods are overrepresented but less than a quarter of majority Black or Hispanic neighborhoods are overrepresented. Overrepresented neighborhoods have average incomes $21,000 higher than underrepresented neighborhoods, so Craigslist rental listings are more likely to concentrate in wealthier neighborhoods.

The authors used computational text analysis to examine differences in information provided in the listings. Listings in communities with more Black, Hispanic, or poorer residents used fewer words or were more likely to focus on tenant requirements and qualifications than on unit amenities. Nationally, listings in poor Black neighborhoods contain 33 fewer words per listing than listings in poor white neighborhoods. Listings in nonpoor Black neighborhoods contain 32 fewer words per listing on average than listings in nonpoor white neighborhoods. In Seattle, the trend is reversed: 63 more words on average in listings in nonpoor Black neighborhoods than in nonpoor white neighborhoods. This discrepancy, however, might be explained by a Washington state law that required landlords to state all tenant requirements in listings. The metro areas with the greatest differences between the length of listings in nonpoor white neighborhoods and nonpoor Black neighborhoods were Salt Lake City, Riverside, San Diego, Las Vegas, and Providence.

The authors also analyzed how often landlords supplied information in Craigslist’s filterable fields. Home-seekers, for example, can exclude all listings that do not explicitly provide an address, indicate whether a washer/dryer is available, or whether pets are allowed. The authors hypothesize that home-seekers with greater means may filter out listings that do not specify this information, so if listings in poorer neighborhoods do not contain this information, those home-seekers may simply never see them. Relative to listings in nonpoor white neighborhoods, listings in poor Black neighborhoods are 7% less likely to contain an exact address, and listings in poor Hispanic neighborhoods are 9% less likely to include an exact address. They are also less likely to include washer/dryer availability or pet policy.

The authors hypothesize that information asymmetries in online housing platforms may function like other steering mechanisms. As Craigslist users with greater means filter out listings in poorer neighborhoods, they are less likely to find homes there. To the extent that differences in information track other neighborhood-level demographic characteristics, this filtering is likely to reproduce racial and ethnic stratification and perpetuate residential segregation.

Access the full paper at: