Consider a specific home such as 583 Warner Avenue in Los Angeles. Suppose that a white family sells this home to another white family and a few years later this family sells it to a black family. A researcher who observes the same home sold at several points in time can estimate a "repeat sales" model to see whether blacks pay more for the same housing than whites. The answer according to this study is that blacks pay 4% more than whites for the same home. Is this a lot of "discrimination"? That is for you to decide. Note the key to this empirical design is observing the same home sold to whites at some point in time while it sold to a black family at a different point in time. If a given home was only sold to black families or to white families, such a home would offer no "within variation" and a repeat sales approach could not be used to study for pricing discrimination.
Here is the paper's abstract:
Price Discrimination in the Housing Market
NBER Working Paper No. 18069
Issued in May 2012
NBER Program(s): PE
This paper sets out a new research design to test for price discrimination by sellers in the housing market. The design controls carefully for unobserved differences in the quality of neighborhoods and homes purchased by buyers of each race, using novel panel data from over two million repeat-sales housing transactions in four metropolitan areas. The results indicate that black and Hispanic homebuyers pay premiums of around 3 percent on average across the four cities – differences that are not explained by variation in buyer income, wealth or access to credit. The estimated premiums do not vary significantly with the racial composition of the neighborhood or, most strikingly, the race of the seller. This latter result rules out racial prejudice or animosity on the part of sellers as the primary explanation for the estimated premiums.