Zillow’s now-shuttered house-flipping unit vastly overpaid for properties in some cities, leaving the company dangerously exposed to even a mild cooling in the real estate market.
That’s the conclusion of a Bank of America research report that examined a selection of more than 300 properties that the real estate firm’s Zillow Offers unit had purchased in a range of markets and currently has listed for sale on the site.
For instance, in Austin, where the real estate market has been particularly hot during the pandemic, homes have appreciated by an average of 6% per year over the past five years, according to the widely used Case-Shiller real estate price index. But Bank of America found that on average the amount Zillow had paid for its Austin houses represented a 27% annual price increase since they were last sold, looking back over a four-year period.
As the bank’s analysts conclude, “This could suggest that Zillow may be purchasing speculative houses that could be at risk for price deflation in a downturn.”
Not surprisingly, Zillow is now having trouble selling these houses for a profit. Bank of America found that on average it is currently listing its Austin properties for 7% less than what it paid for them.
And it wasn’t just Austin. In Sacramento, Portland, Ore., Cincinnati, Tampa, Phoenix, and Denver, the bank’s analysis showed that Zillow was paying prices that represented a run-up of more than 20% per year in the period prior to the company buying them.
Past performance does not guarantee future returns
Paying top dollar for houses that have already run up at historic rates is not a recipe for house-flipping success—as any fan of the television show Flip This House can tell you. Bank of America found plenty of evidence that Zillow had bought many of its houses at the top of the market. For instance, it found that Zillow tried to initially list its houses at an average 5% markup to the amount it paid. But in many cases, the researchers concluded, it had been unable to sell the properties for this price and had to offer multiple discounts, averaging about 6% to the initial listing price, in an attempt to move the properties. On average, across its sample, the bank concluded that Zillow would take a 2% loss if it managed to sell these houses for the current listing prices.
It is still unclear what led Zillow to make such bad decisions, but it may have been lulled into thinking that the surge in real estate demand, which spiked during the pandemic, would continue. So-called ibuying services such as Zillow’s Instant Offers work by using algorithms to try to predict what a house is worth currently and what it may be worth in the future. Those algorithms are often trained on historical data that may not provide good insight into future price movements.
For one thing, Zillow appears to have gone heavy into homebuying just as the number of sales peaked. For instance, for homes priced between $250,000 and $500,000, which seems to be the price range where Zillow was most active in homebuying, year-over-year growth in the number of homes sold peaked at 48% in May 2021, according to Bank of America. That rate has since declined to about 3%. And for homes priced below $250,000, the slowdown has been even steeper, going from a flat year-over-year sales market in May 2021 to one that was –23% in September.
Rich Barton, Zillow’s chief executive officer, cited the inability of the company’s technology to accurately forecast real estate prices as the reason for shuttering the house-flipping unit. “We’ve determined the unpredictability in forecasting home prices far exceeds what we anticipated,” he said, adding that continuing Zillow Offers might put the entire company at risk. The company had cited labor shortages and a lack of builders to fix up the properties it bought as the rationale for pausing its home purchases. But two weeks later it announced it was exiting the business.
The company currently has an inventory of 9,800 houses across the U.S. that it is trying to sell to investors. It has another 8,200 houses that it has under contract to buy. The company said it expected to take a loss of between 5% and 7% across this portfolio. It also said it would take a $540 million charge in the third quarter to account for the write-down of the business.
Investors have punished the company, driving its stock down by more than a third in the past week.
More must-read business news and analysis from Fortune:
This story was originally featured on Fortune.com