(Bloomberg)—If you're looking for an apartment to rent in Atlanta, you might be dismayed to learn that local rents surged 14 percent during the year that ended in February—at least, if you go by the online listing service Zumper. According to Zillow, the increase was 4 percent. Data provided to Bloomberg by CoStar put the increase at 7 percent.
That wild range helps explain why the rents you're seeing out there might differ from what you've been reading—or why you’re not making money hand over fist on that Buckhead mid-rise you scooped up to lease out. It also presents a thorny problem for policymakers seeking to track prices at a time when rising rents are an acute social and political issue.
As the chart below shows, Atlanta isn’t an isolated case. The three data sources tell consistently different stories about the pace of rent growth in cities across the U.S.
There are a lot of sources on the rise in rents across the country, and they don't measure the same thing. Zumper's numbers, which reflect rents for two-bedroom units, are based on more than a million active listings from its website, a big number that nonetheless represents a fraction of all rental units. Zillow and CoStar take listing data as a starting point and then use statistical models to fill in the gaps.
The confusion starts with a growing cohort of online real estate services that have figured out they can get free publicity by publishing rent reports, wrote Daniel Hertz, who blogs about cities at City Observatory. Those reports are generally based on incomplete data, partly because rental markets are large and fragmented while the online services tend to focus on young, affluent audiences.
Hertz puts more faith in data compiled by the Census Bureau and the Department of Housing and Urban Development, but their figures tend to lag behind the real-time market by several years. That makes their data more useful to policymakers than to renters (or, say, journalists looking for a snappy headline).
It's harder to measure trends in rent than in home sales, which typically involve a mortgage and a publicly registered transaction. Apartments can be rented by word of mouth, and rents often go up when a tenant renews his or her lease, without the increase being reflected in a listing service. Apartments that do get advertised once aren't certain to be listed a second time, making it difficult to observe change over time.
A white paper presented by CoStar last month at a conference of the American Real Estate Society helps illustrate the challenge. After compiling information from its listing service Apartments.com, calling landlords, and scraping data from property management websites, CoStar's researchers wound up with at least one data point for units in about 110,000 rental buildings. In that pool of properties, however, there were only 2,200 buildings with consecutive monthly data for the last two years.
So CoStar created a model for estimating the monthly changes in rent for every apartment in the data set, basing its estimates on local market trends and the pricing relationship for apartments of different sizes. In all, the modeled data include more than 10 million observations, said John Affleck, a CoStar analyst. "We're using an order of magnitude more data than anyone else," he said.
Is it a better mousetrap? Apartment hunters, real estate investors, and housing policy wonks will probably have to settle for comparing the theories behind the various rent studies.
Zumper's chief executive, Anthemos Georgiades, said his company's reliance on real-time data, representing what he called the "vast majority" of vacant apartments currently listed for rent across the U.S., is a virtue.
"Our data explicitly says, right now, in terms of vacant inventory, this is the median price," he said.
"Simply taking the average listing price of what’s on the market now doesn’t provide a full and accurate picture of rents," said Svenja Gudell, chief economist at Zillow, which, like CoStar, uses a statistical model to estimate rents for vacant and occupied units and provide a view of rent growth across cities and neighborhoods.
Hertz, who didn't respond to a request for an interview, wrote in a March blog post that Zillow's rent data are useful for tracking rough trends in the relative affordability of a given city or neighborhood, while Zumper's data reflect only the units listed on Zumper's site.
"Unfortunately," Hertz wrote, referring to the field as a whole, "the vast majority of these rent estimates are completely made up."
© 2016 Bloomberg L.P