The iBuyers: High(er)-Frequency Trading Comes to Home Buying

Byrne Hobart
May 31 · 20 min read

Stop me if you’ve heard this one before: companies have found a way to grind out steady above-market returns by leveraging superior risk management. The only catch is that they’re making a levered bet on residential real-estate appreciation.

That is, of course, the story of AIG, Bear Stearns High-Grade Structured Credit, various German banks, and a host of smaller but equally unfortunate participants in the great housing bubble. It’s also a capsule summary of the emerging “iBuyer” business. When I first heard about that business — the business of making instant offers to buy homes, and flipping them to other buyers as quickly as possible — I was convinced it was a great way to tie up tons of capital in a structurally bad business.

I was totally wrong: iBuying is the most exciting thing happening in fintech right now. As about a million clever people have noticed, the twelve most dangerous words in technology are “The four most dangerous words in investing are ‘this time is different.’” This time is different, for many reasons: while the macro situation in residential real estate is still fraught — see this writeup for an overview — the microstructure has improved in favorable ways. The top of the demand funnel has moved online, but the bottom of the funnel is still a paper-based business. And there’s a long history of technologically-savvy liquidity providers producing superior returns in all sorts of markets. From high-frequency trading in equities to building undersea telegraph cables to report exchange rates, we have many case studies of effective risk-taking leveraging new technology.

The iBuyer Model

The real estate market is obviously broken, but exactly how it’s broken is non-obvious. Commissions have plunged from 6% to “roughly 6% unless you’re buying a mansion,” and while the top of the funnel has moved online, the actual purchase process still involves lots of physical paperwork.

iBuyers aim to simplify that. When you want to sell a house, you go to their site, plug in your address, and get an offer. It’s not a great offer; it’s a fair offer, minus a commission, which is usually around 6%. You’re paying for convenience, and 6% of a $200,000+ transaction sounds like a whole lot of convenience.

As it turns out, it is.

Buying a house is not a two-party transaction. In general, you sell a house to move into another house, and most buyers and sellers have a mortgage. So any one transaction is actually a complex multi-party negotiation: you, your lender, the seller, their lender. And since the seller is generally going to move somewhere else, there’s actually a daisy-chain of adjacent transactions. It’s like a line of hermit crabs, except that every crab also has to check in with a loan officer.

From that perspective, it’s amazing anyone ever manages to sell a house.

What the iBuyers do, then, is offer convenience to a deeply inconvenienced constituency that also — by virtue of their plans to buy an asset that banks are willing to lend against — has access to capital.

A 6% fee sounds expensive, but as it turns out, home sellers are not that price-sensitive. We have evidence of this from Redfin, whose model is to charge a lower commission and make it up on volume. The business chugs along, but it doesn’t achieve high scale: as it turns out, real estate agents vary in their skill level, and the ones who can close deals well prefer to earn higher commissions. So the business constantly suffers from the worst kind of churn.

As for why 6% is the number, that’s less clear. Probably, the bundle of services that agents offer has converged on about 6%-of-the-median-transaction’s worth. Since competing on price by a little makes you look desperate, and competing on price by a lot gives up crucial economics, the 6% fee persists. (This situation — sticky fees that drift downward, but more slowly than anyone expects — also shows up in hedge fund compensation and IPO underwriting fees, but not in trading commissions and spreads. It seems endemic to any transaction with large information asymmetries and relatively few repeat buyers.)

At the same price point, an iBuyer offers a far better service than a traditional broker: they close quickly, with no need to find a counterparty, and they offer standardized terms. In Zillow’s case, the iBuyer is also connected to a mortgage provider — Zillow acquired a mortgage originator to go along with their instant-offer business. That takes one more troublesome third party out of the multiparty negotiation.

Then, the only question is: how does the iBuyer sell, and how much do they make? In Zillow’s case, the answer is that they sell through their existing agent network, and they make a margin of 0.2%, before interest costs, or -1%, after interest. At scale, they want to make 4–5% on the transaction, and lose 1–2% to corporate overhead. This math depends a lot on scale: right now, interest alone is about 10% of their costs, but halving the time they hold homes means halving the interest cost per home, and there are likely other economies of scale as well. If they have homes that fit the needs of some buyers, that’s good; if they have a home for every buyer on their site, turnover soars, interest costs per transaction approach zero, and the business starts to look pretty decent.

And that’s ignoring the mortgage side. Lead-generation for mortgages is a skewed business. If it’s lead-gen in the sense of an email address from someone vaguely interested in comparing mortgages, that’s worth tens of dollars, but if it’s the name and phone number of someone who is about to sell one house and buy another, the value of that lead is in the thousands. It’s a large financial transaction with someone who needs money and has collateral, and that’s worth a lot.

But it’s also worth a lot because actually originating a mortgage is hard. One industry participant I talked to claimed that it’s the most regulated part of the economy outside of healthcare. Brokers basically have a fiduciary duty; if they get paid to pitch a mortgage that doesn’t also offer the best rate, they’re in big trouble.

Actually scaling a mortgage business is hard, too. Approval happens county-by-county, so if you want to tie mortgages to a real estate brokerage business, your options are either a) somehow get regulatory approval everywhere, and also constantly explain to bureaucrats at the CFPB why your model isn’t against the rules, or b) work closely with a large bank, which has effectively unlimited capital and can originate a mortgage everywhere, but who will also have bureaucrats demanding that you explain yourself. In consumer finance, rapid growth tends to accompany some combination of rule-bending and credit quality deterioration, so any iBuyer whose mortgage arm achieves rocketship/unicorn growth rates should expect to spend a lot of time and energy on lobbying.

So let’s set aside the mortgage part. It’s a piece of the story, but more a piece of the 2029 story than the 2019 story. Instead, we can focus on the economics of buying homes and selling them.

From Zillow’s Q1 ’19 earnings, here’s how they break out the economics:

They cite some levers for raising their margins — note that their long-term margins do not assume any revenue from the “adjacencies” they mention:

These adjacencies vary in value, though. Moving leads, for example, seem to cost under $100. At the $730/home pre-interest margin, that’s material. But in a steady state, it’s not. Remodeling leads are in the $200–300 range. Home seller leads are big — Zillow is obviously a bit opaque on exactly who pays how much for a lead, although they are quite clear on the general argument that Zillow leads are very cheap.

Overall, it’s not out of the question that iBuyers can drive the kind of long-term economics Zillow talks about, merely by scaling up some of their repeat costs. Lead-gen on top of that would be gravy — a few drops for moving and remodeling, big but possibly unattainable dollops for mortgages.

The Supply Side: Taking Advantage of Bubbles and Legibility

Zillow is buying homes in markets like Phoenix, Las Vegas, and Miami. Opendoor is in similar markets. Both companies skew towards the “sand states” that were ground zero for the real estate bubble. This is not a coincidence: the bubble increased housing supply in these markets, and in order to most efficiently increase supply, homebuilders tended to standardize homes. This reduces the number of variables necessary to model the market price of a home — a model with fewer features is a model that’s less likely to overfit.

I actually lived in Phoenix for a year during the height of the bubble (2005–6), and it was creepy how identical houses were. There were a handful of templates copied over and over again. The iBuyer model is not going to come to Brooklyn for quite a while; walkups, old heating and plumbing, and mixed-use neighborhoods all make it hard for your algorithm to spit out a consistent price. How is Opendoor going to evaluate my old apartment located above a Thai restaurant, that smelled strongly of Moo Tod and Pad See Ew from about 11am to 10pm, with a living room whose floor sloped noticeably enough that pens and pencils all accumulated against one wall? The confidence interval for the price of an old building is a mile wide; one resident’s “charming” is another one’s Lovecraftian.

When you hear about a company with a novel plan for speculating in real estate, and hear about locations like Phoenix and Vegas, alarm bells should go off, but these markets make perfect sense for a model that requires evaluating real estate at scale. They’re essentially following a path forged by the last bubble — when the market had high demand for generic residential real estate, the result was some extremely generic residential real estate.

This is not an uncommon pattern. The political scientist James C. Scott talks about “legibility,” or the ability of the state to keep track of individual people for the purpose of conscripting and taxing them. Bubbles systematically impose legibility on the real world, because every bubble starts with an abstract vision of how the world ought to work, and then deploys capital to make it so. You can see countless examples of this in history:

  • The rise of global trade incentivized the development of longitude, literally adding an axis to the graph of where everything is in the world.
  • The railway bubble added the dimension of time, by giving travel within countries a fixed timetable.
  • Electrification in the 1920s standardized factory layouts and appliances; power was a sort of “API” built out in the first half of the twentieth century, which later technologies like microwaves, TVs, and computers could directly tap into.
  • The Internet bubble pulled a lot of information online, jammed it all into HTML, and then made that HTML the endpoint of search engines. If you didn’t format your content in a way search engines liked, it basically didn’t exist (sorry, flash-based websites!), so everyone stuck to the common standard.

As a consequence, the story of iBuyer growth will be a story of real estate bubbles run in reverse; once we’ve finished with the Sand States, we’ll hit the Sun Belt, after which every S&L default hot-spot will be an iBuying hot spot. The last places to get iBuyer offers will be the first places in the US where people built homes, or possibly the places where crazy artists decided to live.

I’m not sure if this weird-looking home, where I spent a couple years of my childhood, will be literally the last building Opendoor is able to make an offer for, but it’ll be close.

Market Forces: The Balance Sheet

While there are changes in technology that make iBuying more feasible, there’s also a crucial change in market microstructure that makes residential real estate a less risky asset class. Historically, residential real estate has had low volatility, but strong momentum. This is because the biggest source of funding for real estate purchases is real estate sales — since most people don’t live in the same place their entire lives, and homes are a big investment, new home purchases get funded by selling the previous home.

This, combined with downwardly sticky price expectations, means that home price declines are serially correlated, and are accompanied by liquidity drying up. When prices are down, one form of folk risk management is refusing to answer the phone.[1] If you close on a house, but your deal to sell your current home falls through, the other deal falls through, too; you and the seller will both re-list the property, but you’ll tend to anchor to whatever price you sold at. If you keep having trouble selling, you’ll eventually mark it down, but it’s psychologically difficult to sell an asset for less than you just recently agreed to sell it for.

This leads to a “bullwhip” effect: if you agreed to sell something at a certain price, and find you can’t sell it at a price somewhat below that, the natural reaction is denial followed by panic. This leads to distressed selling, especially when there’s leverage involved. (In A Man for All Markets, Ed Thorp tells the story of begging his trader to buy cheap futures the day of the 1987 crash; prices had overshot so much that there was an easy arbitrage from buying the futures and selling stock. Sometimes, panicking less than somebody else is a source of alpha.)

That was the pre-bubble status quo, and that was one of the reasons housing prices fell so fast. From 1987 to now, the Case-Shiller index has compounded at about 3.7% per year. In the early 2000s, growth accelerated, and at the peak in Spring ’06 prices were about 40% above the long-term trend. Prices declined steadily from there, finally bottoming in early 2012.

Normally, when an asset gets cheap, financial buyers get interested. There are natural buyers for stocks, bonds, hard assets — anything you can buy millions of dollars of at a time. For real estate, though, the natural buyers were existing landlords, who were generally financially stressed. (Because it’s a levered and cyclical asset, real estate’s investor population tends to be dominated by over-indebted optimists at the peak of the cycle. This happens in other markets, too, but since real estate is illiquid, levered players can handle distress for longer without defaulting. The longer any real estate cycle lasts, the more it’s dominated by people whose ego and promotionality match that of, say, the President.)

After the Financial Crisis, though, institutions got interested. There’s a point at which an asset is so incredibly cheap that even big private equity funds are willing to do high-touch transactions to invest in it $100k at a time.

Today, there are companies like Invitation Homes that own vast portfolios of residential real estate. They have professional management teams, dry powder, and access to effectively limitless capital. When they change their view on real estate, they can pull back fast, and when prices drop, they can lever up and buy more.

This has meaningfully changed the dynamics of real estate: it’s no longer a momentum asset. Before the crisis, sequential changes in the Case-Shiller index predicted the next month’s change with an r-squared of .92. Since 2006, it’s been .82, and since 2012 (when institutional money got big and prices started to rise) it’s been .43. Looking over the entire history of Case-Shiller, sequential correlations have varied, but have generally been higher.

This means that long-term historical data on real estate prices actually overstates risk. The market players are different today.

When market movements are hard to predict, market-makers win because they face less adverse selection and less risk. Just as equity market-makers prefer markets that don’t trend (payment for order flow is basically paying to only see the 100-share orders that aren’t part of a 100,000-share order), housing market-makers benefit when the market is a random walk.

Industry Precedents: Broker Model to Dealer Model

Some markets emerge as a dealer market: A wants to sell, B makes an offer and hopes to sell to C at a markup. Other markets evolve to a broker model: A wants to sell, B tracks down C, and A sells to C while somebody pays a commission to B.

The key driver of the broker vs dealer model is information asymmetry. Specifically, if there’s asymmetry with respect to assets, you expect a dealer model; if there’s asymmetry with respect to buyer and seller identities, you expect brokers. In markets where the homes are all different, we’d expect dealers to eventually win. This is because the dealers can mitigate adverse selection through better investigation of individual assets.

Equities are an interesting case study here: market-making as a business has fluctuated depending on which party had better information. In the 60s, the equity market makers who thrived were the ones who had the capital and risk appetite necessary for large block trades; in the 90s, SOES bandits took advantage of loopholes in NASDAQ’s rules for small market-makers to get better data on order books, which they used to quickly snipe trades.

In fixed-income, interchangeable assets like treasuries are easy to trade; you don’t have to call up a broker and ask if they have any ten-years lying around. But corporate bonds and mortgage-backed securities trade more rarely; if you’re one of five people who might want to own an obscure company’s bonds, it pays to sell to a middleman rather than hunting down one of the other handful of natural buyers.

Every market evolves in a path-dependent way, so it’s hard to generalize, but the overall trend I see is that market-making sounds like a business where there should be just one big player, but that almost never happens for very long.

  • Equities, equity derivatives, treasuries, futures — these are all highly standardized products, and there are lots of market-makers. You’re seen as a huge player if you have more than 1% market share in a business like this.
  • Small-cap equities were a cartel until they got caught and settled with the SEC. I chalk this up to insufficient automation. The two constraints on a human market-maker are capital and attention. One person can keep an eye on 50 stock prices, but you have to be pretty unique to handle 500. So the business was naturally fragmented, but since the best source of liquidity is another market-maker, and quotes were displayed publicly, it was a business socially optimized to maintain cartels. Every market-maker could see if somebody else was refusing to quote only odd-eighths, and could punitively turn down their business.
  • Bespoke derivatives: on an individual basis, customized derivative contracts are a series of small monopolies, but since exotic options can be approximately hedged with vanilla options, it’s hard to get a durable competitive advantage. For some values of “a lot,” there are “a lot” of people who will write you some elaborate bet on the correlation between soybean volatility and Finland’s term spread, if that’s your thing.
  • In high-yield bonds, market-making was Milken’s monopoly for a long time, although even before Drexel got shut down competitors were making inroads. This seems to be an example of the one-giant-brain theory of scalable market-making: Milken had an eidetic memory for who owned what, and he worked hundred-hour weeks. Even so, Drexel’s share slipped over time.
  • Enron also had something close to a monopoly in energy trading, which collapsed with the rest of the company. Interestingly, both Enron and Drexel’s collapses were overdetermined: the market-making business got taken down by a funding crisis when a different arm of the business ran into legal trouble, but those market-making businesses both depended on a limitless risk appetite. Risk tolerance plus leverage means disaster eventually, although the exact nature of the disaster is not obvious in advance.

So looking at the iBuyer business, while there are economies of scale (it’s easier to renovate 100 homes at once than one at a time, and easier to sell a given buyer one home out of 10,000 than one home of a dozen), these scale economies tend to top out early. It’s rare for a market-making business to be winner-take-all, perhaps because the capital requirements scale up so fast. In finance, scaling up in a liquid market is glorious, but scaling up in an illiquid market means changing the nature of the market as you grow, which is trickier to model. Capital providers have learned through many painful cycles that it pays to be patient.

And the pure scale of the residential real estate market is astounding. Existing home sales are running at 5.2m/year, at a median price of around $267k. That’s about $1.4tr of value changing hands every year. 1% market share means $14bn in annual turnover, which, assuming a three-month holding period, requires $3.5bn in capital. Raising $3.5bn in capital is not easy. What you’d expect is for various iBuyers to incrementally raise hundreds of millions at a time.

And, in fact, that’s what you see. Zillow raised $650m in mid-2018 to fund their Offers business. Opendoor has raised $1.3bn in equity alone, most recently in $100m+ chunks that seem to include equity and debt. OfferPad is at $155m, Perch is at $250m, and Knock is just over $430m. In one sense, that’s a lot of money. In another sense: the iBuyers, combined, haven’t raised enough to reach even 1% market share, and they all seem to be in the market to keep raising more.

While there are network effects, they’re localized, which further contributes to national fragmentation. Assuming economies of scale, the best place for Opendoor to invest is wherever they have bigger share than Zillow, and vice-versa. That’s a recipe for a stable oligopoly.

The Social Impact: Lower Housing Exposure

I’ve written before that Americans have too much exposure to housing. We buy bigger houses than we need with more leverage than we should, and this massive housing exposure turns residential real estate into a crude macro-stabilizing tool: rate cuts get transmitted through the system fast (by prompting people to refinance mortgages), while rate hikes slow things down (by encouraging people to keep the low-rate mortgages they have). This is not ideal; it means there’s a tax on economic dynamism when the economy does well; you’re penalized if you move for a new job when the economy’s hot enough to justify higher rates, but you can move for a new job when rates are low because there aren’t any jobs.

iBuying doesn’t solve the entire problem, but it improves things on the margin by making it easier to sell a home. Since iBuyers negotiate with realtors at scale, there’s a good chance they will push down fees, too: while many homeowners aren’t sensitive to 1% changes in price driven by fees, corporations running a low-margin business are very sensitive to 100 basis points of incremental margin.

It’s a better bet than lawsuits, at least. Realtors have more concentrated interests than homebuyers; they’re fighting for their livelihood, while homebuyers are fighting to save a little money on one deal. But corporations have even more concentrated interests than realtors; in this fight, they’ll win.

Higher liquidity in housing should push more houses into the hands of institutional capital rather than individual investors. That’s socially optimal, too: since real estate correlates with local wages, buying a home near where you work is doubling down on your biggest un-hedgeable risk. Even if the total return of other assets is lower, your economic sharpe ratio is higher if somebody else owns your house and you just pay them rent.

Surviving a Collapse

The end of the real estate bubble is not the first thing iBuyers want to talk about, but it’s pretty much all they think about. Whoever you are, if you’re involved in residential real estate, you remember 2007 (and 2008, and 2009, and… ).

The iBuyer model gets unexceptional returns on assets even under the best circumstances, but it’s amenable to leverage; if a homebuyer can put down 10% equity, the owner of a diversified portfolio of homes ought to be able to run with higher leverage.

So the question is: can you prudently run a residential real estate market-making business with enough leverage to make the total return exciting, but not so much leverage that it becomes the wrong kind of exciting? Quite probably: the decreasing momentum in housing prices means that even if month-to-month volatility hasn’t declined much, expected volatility over the course of the transaction has declined.

Meanwhile, the history of market-making tells us something important: when the market turns, spreads widen, real-money traders pay more for liquidity, and market-makers clean up. They heydey of high-frequency trading was during the 2008 crisis; high-yield bond traders made their biggest fortunes during serious recessions. While a price dislocation is scary, it’s bound to be less scary to iBuyers than to other parties: there’s property-level data on all US houses going back to 2000, so you can literally stress-test your model against the biggest residential real estate crash since the Depression and see how it looks. That’s an aggressive stress-test given consumer develeraging and the rise of institutional capital, but it does give iBuyers a margin of safety.

If housing prices drop, I’d expect a lot of scary headlines about how iBuyers are instantly underwater. But if you’re charging a 6% commission on a three-month trade for an asset that has never dropped more than 4.5% in three months, you’re probably safe:

As this histogram shows, real estate is a boring asset most of the time.

So the headlines will be scary, but they’ll be wrong; iBuyers will charge higher commissions (or, equivalently, make lower offers). Their speed advantage will be bigger as traditional homebuyers struggle to finance purchases. They’ll lose money on some individual houses, but get insane deals on others.

But Does It Scale?

That’s the key question. We’re in an incredibly fortunate situation, though: there are private companies building iBuying businesses, but one of the biggest players is publicly-traded Zillow. They’ve recently bet the company (and switched CEOs). And they’re breaking out lots of operating detail around the offers business. Every quarter, we’ll be able to see how the margin profile of a scale operator changes. The big cost drivers — renovation and selling costs — are pretty commoditized, so if we know what Zillow’s margins look like at a given scale, we have a pretty good idea of what another operator’s are.

Where it gets really interesting is this: the pricing models will keep getting better. A good realtor learns all sorts of tricks for buying and selling homes. Weird stuff to double-check, which rooms to start in and end in, what features 95% of buyers ignore that 5% of buyers will reliably pay up for, etc. Software, of course, learns that much faster. It starts dumb, but it gets very smart. As software touches more parts of the home-buying transaction, the information asymmetry will grow.

As in other parts of quantitative finance, iBuyers won’t win from identifying one big signal that determines prices: they’ll win from finding hundreds of tiny signals that add up to differentiation. This actually creates competitive barriers to entry. If you’re trading, you want to live in a world where the obvious wins are arbed out, so nobody can compete.

I’ll be watching the progress of Zillow Offers quite closely, and comping it to other iBuyers when I can. Residential real estate is a very broken market, and this won’t fix it. But US residential real estate is worth over $30 trillion. Making that market just 1% better will make a lot of people rich.

Further Reading

BusinessWeek on iBuyers

NYT on iBuyers, and again.

This guy seems to know what he’s talking about, and he’s more optimistic than I am.

Kevin Kwok has a great writeup of the “Content Loops” model for understanding marketplace businesses, with an emphasis on Zillow Offers.


Don’t miss the next story. Sign up for my occasional email newsletter. Or check the About/FAQ page for more.


[1] You can materially improve your quality of life — particularly how much people trust you — if you force yourself to relay bad news faster than good news. There’s empirical evidence that everybody is bad at this. Among other things, stocks go up gradually and go down suddenly, implying that the bad news comes out all at once. On a micro scale, late news is bad news: an earnings release that happens hours later than usual rarely reports a beat. Good news is simple: our prediction came true. Bad news is complex: something went wrong, but if we knew what it was we would have avoided this problem.

The Startup

Medium's largest active publication, followed by +481K people. Follow to join our community.

Byrne Hobart

Written by

I write about technology (more logos than techne) and economics.

The Startup

Medium's largest active publication, followed by +481K people. Follow to join our community.