How Zillow's homebuying scheme lost $881 million

They failed in the same way as my bad fantasy football team.
How Zillow's homebuying scheme lost $881 million
Image by paulbr75

Zillow made news last week as it reported a loss of $881 million on its house-buying business last year⁠. That’s especially remarkable because house prices rose dramatically for much of the year.

Naturally, this got me thinking about fantasy football. I’ll get to that later. But first, let’s back up and talk about what the business idea⁠—often called iBuying⁠—is actually about.

The business of market-making

Sellers want their money quickly, often because they need to put it towards the next house. But it takes a typical buyer weeks to secure financing and do other due diligence—and sometimes deals fall through.

Zillow was hoping to fill a gap between buyers and sellers. They would hunt for bargains with (what they thought to be) relatively low offers for the homes. They would make all-cash offers, hoping that certainty and speed would make up for a low offer price. Then they would turn around and sell the houses to buyers at a higher price on a more relaxed schedule.

This is not intended to be a pure “investment” strategy; it’s more of a service, where, from the seller’s perspective, you expedite the house-selling process in exchange for a fee. It just so happens that the way you expedite it is by buying and briefly holding the house yourself.

In the real estate industry, this service is called iBuying, but in more general terms it’s often called market making. It is a perfectly legitimate service performed in a variety of markets. For example, when you buy a share of stock, a market-maker firm might sell you one of theirs, and then replenish their inventory the next time a seller shows up.

“Our aim was to become a market maker, not a market risk taker,” Zillow founder Rich Barton said in a letter to shareholders, addressing the failed initiative.

But the job of a market maker is much harder in housing than in stocks. There are a couple of reasons why.

The fantasy draft problem

Zillow’s plight makes me think of one of my poor fantasy football teams. (Don’t worry, knowing the sport is unimportant for the analogy.) In fantasy football, you choose a roster of real-life players, and the performance of your players in real life determines your score in the fantasy league. The rosters and scores are usually tracked online. (We used Yahoo!)

The most important day in fantasy football is draft day, when you take turns claiming the players you’ll start with. Later you can add unchosen players, or trade with other participants, but the draft remains key to success.

One time, I missed draft day. This was a disaster. In my absence, Yahoo! picked the best available players⁠—according to a list compiled by its staff⁠—on my behalf. This is sometimes called autodraft. Meanwhile, my friends who showed up got to select players of their choice.

At first glance, autodraft sounds not so bad, just like Zillow’s strategy doesn’t sound so bad. After all, Yahoo’s fantasy staff are fairly knowledgeable about football. But the list has some flaws⁠—for example, it might only be compiled three days a week, and miss some late-breaking news until the next update. Or it might not be well-tailored to the specific settings of your league. All in all, it is going to overrate some players and underrate others.

Here’s the problem: my autodraft team ended up with a roster almost exclusively stocked with the players idiosyncratically overrated by Yahoo!. For example, an older player whose younger backup had just earned praise from the coach, suggesting that he might be about to lose his job. Yahoo! would eventually get to downgrading that player in a future update, but only after he already ended up on my roster. If memory serves, I also got a few recently-injured players, and a few that Yahoo!’s experts just happened to love much more than the overall consensus of football fans.

My team never recovered from my horrid draft, and I had to wait until next year to field a competitive team again.

Zillow’s iBuying wing suffered a similar fate. Zillow’s Zestimate®, like Yahoo!’s fantasy football experts, is pretty smart, and pretty good at valuing things. But it’s not omniscient, and it certainly doesn’t know as much as the sellers. It values home prices fairly most of the time, but occasionally it overrates or underrates a home.

The problem is that⁠—much like my fantasy football team⁠—Zillow’s formulaic purchasing strategy virtually guaranteed that it would get the slice of the inventory it had most overrated. Any time there’s a problem with the home not captured in the Zestimate, (but perhaps visible to a slower, more careful buyer), Zillow’s iBuying program was likely to overpay. And the people getting a great deal from Zillow would be those most likely to take the deal.

This is a mix of problems that economists call “adverse selection” and “imperfect information,” most famously studied by George Akerlof in the 1970s. (His example was used cars.) Any time that you’re working with imperfect knowledge and trying to operate a business on a large scale, you’re likely to run into this kind of trouble.

Do market makers in the stock market run into this problem? Absolutely, but it’s not nearly as severe. They are only trading a limited number of public stocks, and they can more quickly detect patterns that suggest mispricing. For example, if they find that a huge number of people are clamoring to sell a particular stock, a market-making algorithm might surmise that there’s an actual problem with the stock, and limit its exposure until trading stabilizes again.

It is harder to do this with houses, since every house is different. Zillow may not be able to systematically detect the patterns in the houses it overrates, or adjust its algorithm to avoid stocking up on overrated homes.

The holding period problem

Barton’s quote⁠—“our aim was to become a market maker, not a market risk taker”—hints at another problem with iBuying. Market makers don’t want to “invest” in their market, in the sense that Warren Buffett might. They want to trade with impatient buyers and sellers, and earn a sort of “patience premium” rather than develop a deep thesis about the future of the assets they’re holding. They want to keep minimal inventory.

This is not too hard with stocks. With stocks, holding the inventory is relatively costless, in a physical sense. You take on some risk that the stock market generally will go up or down, but that’s not too bad. The market goes up more often than it goes down. You can probably even offload that risk to a long-run-style investor who is happy to be exposed to it. In addition, the relatively commoditized nature of stocks, and their fast transaction speed, make it easy to unload. A market maker might be able to complete its whole task within a few seconds for a well-traded stock.

With housing, the inventory is costly to hold physically. You have to secure and keep up the home. And housing is very likely to be a bad investment while it’s vacant and waiting for sale, since it generates no rent. Each house is different and buyers are slow, so an aspiring market maker might get stuck holding it for a while.

Furthermore, housing transactions take much longer than stock sales. Effectively, if you hold a large portfolio of homes for five months, you end up being a speculator on housing prices, whether you want to or not. So even if Zillow aspired to be a short-run market maker, it ended up being a market risk taker, implicitly betting on the longer-run prices of the housing it acquired.

This became a real problem last year, as it failed to manage a volatile market. With fast-rising real estate prices early in the year, Zillow’s offers weren’t competitive enough. They changed the algorithm to bid more aggressively, and ended up with too many aggressive winning bids just as the market began to soften. Ironically, while Zillow's business model was premised on being patient, the company itself showed remarkable impatience.

iBuying is not dead, just sickly

It’s easy to write a business model off when it fails, or when a problem with a market is described. Above I outline some serious issues with the iBuying model. But there’s also a clear need for market-making in housing. Most buyers don’t have the flexibility to offer sellers the quick cash offers that many sellers prefer.

A problem like the fantasy draft problem will merely slow iBuying down and make it worse, not kill it. Zillow’s exit from the market strengthens the position of companies like Opendoor or Offerpad, which do something similar. A less competitive market is less dynamic and worse for sellers, but it is better for the iBuyers that remain. They can make less competitive offers, perhaps offers low enough that the iBuyers have margin for error, even knowing that they’re likely suffering from a winner’s curse and getting the houses they most overrated.

When economists first started examining the “market for lemons,” they didn’t conclude that the used car market would collapse completely because of people selling bad used cars. Instead, they concluded that the used car market would be less robust than it would if buyers could more easily determine car quality, and that buyers might have to do costly work to address their lack of information.

It is likely that the iBuying market will endure, but it will be smaller and less efficient than its proponents had hoped.

Comments

Only members can comment.
Please subscribe to a free or paid plan or sign in to join the conversation.