iBuying Homes and the Market for Lemons

Zillow’s decision to shut down its iBuying division made waves in the online home-buying market. Most notably, Zillow lost over $500 million in the process, and as it winds down its division is selling thousands of houses at a loss.

Why did this happen? After all, Zillow has a wealth of data on the entire housing market — it knows what homes people are looking at and where, it accesses and amasses public records, and it has a large engineering team to help predict home values as close as possible.

Part of the answer lies in a fascinating piece of economics called “adverse selection”. It’s a fairly simple idea; so simple, in fact, that George Akerlof’s related work illustrating this in the “market for lemons” won him and his colleagues the Nobel Price in Economics in 2001.

What’s adverse selection, and what does it have to do with lemons? In the famous study, “lemons” refer to bad or defective cars, ones worth a lot less than a good car. The issue is that buyers may not know whether a car is a “lemon” or not at first glance, while the seller knows for sure which is the case.

  • If the buyer offers a “high” price, the seller will sell the car regardless of quality
  • If the buyer offers a “low” price, the seller will only sell if the car is a lemon (the price is too low for the owner of a good car to part ways)
  • Since the buyer in this example can’t tell the difference between the good or bad car, they hedge their bets and offer a price in between the fair value of both (let’s say halfway between the “lemon” price and the good car price)
  • This price isn’t high enough for the good car owners, and they leave the market, which only leaves lemons to choose from

There are ways to mitigate this, of course, but consider how the general model applies to online home buying (where transactions close fast, and possibly sight-unseen):

  • Zillow offers $X for the house, so the seller will sell if the actual value of the house is $X or less
  • Zillow uses “algorithms” to determine what they should offer the seller, based on available data

Zillow (and other online home buyers) certainly have some asymmetric info benefitting their side, such as knowing how the overall market looks, and perhaps upcoming market trends. However, the home owner has a lot of asymmetric information as well: knowledge about the foundation, plumbing issues, noisy (and/or nosy) neighbors, etc. In the end, the homeowner has to agree to sell; given how much more they can know about the house compared to Zillow, we risk ending up in a “market for lemons” scenario where Zillow buys houses that are worth less than they paid for (and less than they anticipate, even after using some more asymmetric information to clean up and try to resell the houses).

Normally, someone buying a house gets an independent appraiser and walks through the house/broader area themselves before closing a deal. They might even meet the homeowners themselves and chat. This helps correct asymmetric information in a way that algorithmic decision making often fails to take into account. Although Zillow could have done all of this, it doesn’t scale well in the way that SaaS startups are used to, and someone along the way made the decision about what would be “good enough” for prediction — falling into the adverse selection trap.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s