Weekend Reads | The Subtleties of Tracking Rent Increases

by Kevin Schofield

There's been a lot of attention on inflation as it soared during 2021 and 2022 and then slowly came back down to Earth in the past year. But what does inflation measure? As constructed by the federal government's Bureau of Labor Statistics (BLS), it's a complicated mixture of several weighted components. One of the largest of those is shelter costs — which, for many of us, means rent. Shelter costs make up 32% of the BLS's Consumer Price Index (CPI), the most commonly used index for broad-based inflation.

The BLS has its own way of collecting and analyzing data on residential rent prices, but it is hardly alone; several other organizations, both for-profit and nonprofit, publish their own statistics on rent. And none of them agree (if they did, most of them would probably have stopped publishing by now). But which should we trust?

This weekend's read is a 2022 paper by researchers at the BLS and the Federal Reserve Bank looking at several of the more commonly cited rent indices to understand why they don't agree, and proposing two new ones that the authors argue are more accurate, even than the BLS' own.

At first glance, it might seem it would be easy to construct a rent index. It turns out, though, that the devil is in the details. Here are some of the key questions one must answer:

• What kinds of rental units does it include data for? If you're getting data from the real estate industry's Multiple Listing Service (MLS) databases (as some rent indices do), then your sample will include almost exclusively single-family homes and not apartments in multi-family buildings.
• Is it gathering data from "asking" prices for rental units that are actively for rent, from new rental contracts signed, or from existing tenants? The vast majority of rental units are occupied at any given time, so the asking price for vacant ones isn't a good representative sample of what people are paying — though it is useful for knowing what a new tenant might expect to pay. Likewise, according to the BLS data in any given month, only 13%—25% of units have a new tenant, so even tracking new leases doesn't give anywhere close to the full picture of what people are paying for rent. Curiously, the BLS also says the average tenancy length is about three years: Together, this suggests there is a set of units that are almost constantly churning (college student housing perhaps making up a significant part of it?) and another set that is extremely stable with long-term tenants.
• Is the index measuring "repeat rent"? Is it collecting data on the same rental properties and measuring the changes in rent for those units, or is it a random sample of properties at any given time? Repeat rent is seen as a more accurate way of measuring rent inflation.
• How often are samples taken? Monthly? Every six months? Only when the tenant changes or a new lease is signed? And how good is the response rate for collecting data?
• How old is the data? The BLS tracks "repeat rent" on a stable set of properties, gathering updated data on each one at six-month intervals, staggered across the set so that in any given month only a subset is asked for updates. So its data is, on average, three months old.
• Is the set of data weighted to ensure it accurately represents the existing rental housing stock? For example, does the data set contain approximately the same proportion of studio apartments — and their age — as the real set of housing out there?
• Is rent data adjusted to create an "apples to apples" comparison even when some units include utilities and others don't?
• How do we handle some of the common events that happen to rental units: vacant periods, renovations, add-ons, condo conversions, and simply aging over time? The BLS rent index keeps a specific unit in its data set for six years to avoid issues with aging units, and it will adjust prices if it receives data that a unit has been updated or renovated. Other indices don't have access to the data they would need to make that kind of adjustment.

Of all these factors, it turns out the most important one is the scope of data collected: listings versus new tenants versus all tenants. According to the researchers, that factor explains the majority of the differences between the published indices. But they also point out that an index of new tenant rents isn't necessarily better or worse than an index of rents for all tenants; they're both good, but for different purposes. If you're looking for a new place to live, you care about new tenant prices, but if you're an economist or a policymaker trying to measure how much Americans are spending on rent, you probably want to look at all tenants.

That's not to say the indexes can't be improved. The researchers developed two new ones, one for new tenants and one for all tenants, that they propose are improvements on the BLS' official index (using the same underlying data). The big change they made is to try to address the staleness of the BLS data: Rather than assigning a new survey response to the day it was received (assuming it accurately reflects the price a tenant in that unit is paying at that time), they backdate it to either the move-in date if it's less than six months before, or to the most recent six-month anniversary of the move-in date, based on the assumption that most leases are based on increments of six months (so that's when the vast majority of rent changes would take effect). In addition, they also drop units that go through major renovations to ensure the consistency of the data set.

These changes mean they spend a lot of time revising the inflation measures for past periods of time — particularly in the past year — as they gather new data and backdate it. This kind of revision is, unfortunately, more common than economists would like to admit; for example, we constantly hear that national employment figures for past months have been revised up or down as better data trickles in.

With their new rent indices, the researchers look at how they relate to the BLS' original index, as well as to each other. First, they found that their new-tenant index is more volatile than their all-tenant index, which is not surprising: It's a smaller collection of rental units, focused on the ones that are changing hands. Their new-tenant index also leads the all-tenant index by about nine months: In other words, significant changes in rents for new tenants show up in the data for all tenants about nine months later (though usually not as extreme). And their all-tenant index, in turn, leads the BLS' index that is used to calculate CPI by about three months (because they backdate survey responses). What that means is that the changes we see today in the rental prices for new tenants will take about a year to show up in the official CPI inflation numbers — not because the CPI is wrong, but because CPI represents all tenants, not just the new ones, and it takes that long for enough existing leases to turn over to see the change.

So what should we conclude from this? None of the various rent indices is "right" or "wrong"; they are all just measuring different things. Which one you decide to use should depend on your goal. But the details matter — a lot — and when someone cites you a figure on how much rents have gone up (or down), you need to ask some hard questions to understand what that really means.

Disentangling Rent Index Differences: Data, Methods, and Scope

Featured image via sulit.photos/Shutterstock.com.