Does OpenDoor have an adverse selection problem?

Interesting post from our friends at Marginal Revolution on OpenDoor Labs, a company the provides liquidity (for an average price of 7-12% to homeowners) to housing markets by buying and re-selling houses.

An analysis of property records prepared by Michael Orr, a real-estate expert at Arizona State University, shows that, through mid-December, OpenDoor had bought and sold just over 200 homes. It paid an average $230,000, reselling them within 90 days for an average of $245,000.

However, they do hold an unsold inventory of houses:

But the records show OpenDoor also owned about 30 homes as of mid-December that it had failed to resell for at least six months

If the pricing model is estimated with data from ordinary transactions, and if homeowners posses information not available to the model, like the “feel” of the house, then we should expect that owners who think the model predicts a price too high will be more likely to sell. This might explain the unsold inventory.

If this kind of adverse selection is a big enough problem, OpenDoor may be forced to charge higher fees for its service, and that will discourage all but home owners with hard-to-sell houses from using the service, which is a kind of Adverse Selection Death Spiral (links here)