The front page of the business section in the StarTribune for Sunday, December 5, 2021, prominently features an article headlined “Rental firms snap up metro houses,” or in the web version, “National investors are snapping up Twin Cities area houses to rent.” This topic relates to the streets.mn core values: the mix between owner-occupied and rental properties affects our communities’ ability to empower and include the vulnerable and marginalized. The type of entities that own rental properties also has an impact, whether they be local individuals or national corporations. Alas, the newspaper’s highly visible article is not a reliable guide to this topic.
This will be a long explanation of some of what’s wrong with the article focusing on what I see as the core of the piece, which is its presentation of the Minneapolis Fed’s data about property ownership. I’ll ignore the more context-setting stuff regarding the difficulties buyers are facing.
The property ownership data passed through three consecutive stages. First, county assessors’ offices in each of the seven counties accumulated it for reasons unrelated to this analysis. The fact that they had other purposes in mind and that the seven were operating independently and not always consistently introduces the first opportunity for problems.
Second, the Minneapolis Fed took the county data set and did their best to clean it up into a consistent, usable form, as described on their web page. I respect the work they did, but this is inherently difficult enough that it wouldn’t surprise me if some problems cropped up at this stage; some properties may be misclassified, for example. For larger units of analysis, these probably don’t matter much. But problems can arise once you start zooming in on a particular category of property in a particular census tract. If a number looks out of whack, it’s worth digging in to see what’s going on.
And third, we get the Strib’s summary of the Fed’s analysis. That seems to be a big area where problems originated. They made choices that amplified some of what was problematic at the first two stages, setting some of their parameters of analysis to maximize the anomalies in individual small areas and then not critically examining those anomalies but rather trumpeting them.
Further, they introduced some major new problems just by getting the vocabulary confused. They incorrectly included condos in “single-family houses.” They attributed to the entire city of Hopkins data that is just from a single census tract within that city; likewise for Fridley. And then there’s the question of what differentiates “investor-owner” from a “corporation” or a “national buyer,” terms which the Strib seems to use interchangeably.
Finally, in addition to amplifying anomalies and botching some of the vocabulary, the Strib paid too little attention to what the data show about the evolution of the situation over time. One would never know from the article, for example, that some of the most heavily investor-owned census tracts have been coming down in investor ownership in recent years. Instead, we get the current status of those tracts (still heavily investor-owned) juxtaposed with metro-wide data on the trajectory (steadily more investor ownership).
The Fed’s homeownership data tool has four primary variables (or filters) you can set to find the number you’re looking for, in addition to a few others that just control the visualization. Those four variables you can control are 1) the cutoff number of properties to define investor ownership, 2) the type of properties to include (single-family homes, townhomes, condos), 3) whether to include properties owned by banks and public institutions, and 4) the granularity of the geographic units.
By default, the Fed has these filters set to 10 properties; all of single-family homes, townhouses, and condos; excluding bank-owned and public/community-owned properties; and using individual census tracts.
In order to reproduce the data used in the Strib’s article and charts, I found it necessary to change the investor-owner cutoff to two, leaving the other settings at the Fed’s defaults. As I’ll discuss below, some of those default settings are themselves relevant to understanding what the Strib got wrong.
First, however, let’s consider the impact of changing the definition of “investor-owned” from “an owner with 10 properties or more” to “any owner who owns at least two properties.”
The Fed’s published data set does not distinguish ownership by a single person (or married couple) from ownership by a corporation. It also does not distinguish local ownership from far-off ownership. Instead, the only way to get a handle on what kind of owner is in question is by looking at whether the owner lives in the property via the non-homestead tax status and how many properties the owner owns.
The big national corporations that are the article’s ostensible focus would be included using either the Fed’s cut-off of 10 properties owned or the Strib’s two. But lowering the number all the way down to two also inadvertently sweeps in a bunch of other types of owners. This includes small-time local investors, who likely behave differently than big national firms (Do they pose the same issues, for example, of being neglectful landlords?). It would also incorporate wealthy individuals who buy vacation homes or have the luxury of not immediately selling their old home when they purchase a new one.
One place this shows up in the Strib’s data is near the end of their table of “top 20 Minn. census tracts with high investor property ownership” and the accompanying map. (Aside: the abbreviation “Minn.” in this caption seems inappropriate. The dataset covers the seven-county metro area, not the state.) At position 17 out of 20 is a census tract described as in “Oak Park Heights-Stillwater,” in Washington County. On the map, that’s the light tan blob adjacent to the St. Croix river. It’s the only census tract on the list outside the two core cities of Minneapolis and St. Paul, with the important exceptions of Fridley and Hopkins. (I’ll get into those two later.)
The table has a column labeled “share of single-family rentals that are investor-owned,” which for this particular census tract has a value of 14.4%. That matches what the Fed’s tool shows with the cutoff changed from 10 to two, but everything else left at the defaults. Aside: Note the column heading’s odd phrase “single-family rentals.” This is just plain wrong: we’re not looking at 14.4% of the residential units that are rented out, but rather 14.4% of all single-family homes, townhouses, and condos, including those that are owner-occupied. (This also excludes many rental units—apartments and duplexes—as both the Strib and Fed explain.)
But anyhow, 14.4% of a certain group of properties is “investor-owned.” How critically does this number depend on the definition of “investor-owed” as being “properties owned by entities that own two or more properties?”
Let’s look at what the Fed’s team did. Changing the filter in the data tool back to its default of 10 properties owned drops the percentage from 14.4% down to 1.3%. On the map on the Fed’s site, you can see that the census tract goes from standing out in dark color to being just one of many in the middle of the range.
So what’s happing here? Remember, this is a census tract that includes properties by the St. Croix River. Perhaps wealthy people have a second or third home there. Maybe a small-time local investor from elsewhere in the metro area has a property there they sometimes use themselves but often put on the short-term rental market via Airbnb. Or maybe a small-time investor from within that very census tract chose to buy the unit next to their own and rent it out on Airbnb. The reality is no doubt a mix of these scenarios and others. It’s complicated, but it surely has something to do with the particular circumstances of this area. It doesn’t really belong in an article about national corporate investors swooping in and gobbling up metro properties. The Fed’s threshold of “10 properties owned” for investor buyers would have been a better guide.
OK, OK, I hear you saying, what about Fridley and Hopkins? Aren’t those cases what make the article so shocking? Again, Strib’s dramatic findings don’t hold up to closer scrutiny. Their decision to move the investor-owner threshold from 10 to two is still part of the story, but the bigger problems the piece runs into here have to do with the geographic granularity and the types of residences included.
First, to see what we’re fact-checking, it may help to quote what the Stib says about these two suburban cities. They first show up in the article’s lead: “Now, data show that some neighborhoods in Minneapolis and the entire suburbs of Fridley and Hopkins are being transformed by corporate ownership, a phenomenon that shows no signs of letting up.” These same two cities must also be the suburbs mentioned in “And in a half dozen neighborhoods and suburbs, they own more than 20% of the single-family houses.”
That “half dozen neighborhoods” refers to six rows in the table of census tracts that are above 20%, two of which are labeled Fridley and Hopkins, at 27.1% and 24.8%, respectively. And at a later point in the article, we get another variation of the theme: “And in Fridley and Hopkins, more than 25% of the single-family houses are owned by an investor who doesn’t live there.” (Aside: 24.8% is not really more than 25%, but if the number were rounded to an integer percentage, one might think so. That’s a minor issue.)
So, let’s start taking this apart piece by piece. The first thing that jumps out at me is that the quoted text explicitly refers to “the entire suburbs of Fridley and Hopkins,” yet the data table is at the finer granularity of census tracts. If we were to look at the entire cities, would the percentage still be “more than 20%”? How about “more than 25%”?
Conveniently, one of the selectable filters on the Fed’s site does precisely this—one can switch from census tracts to cities/townships. If we do that, leaving everything else as the Strib had it (including a cutoff of two properties), we find that the rate of investor-owned properties in Hopkins drops from 27.1% to 14.4% while Fridley’s rate goes from 24.8% to 11.1%. That’s still a substantial percentage of investor-owned properties in each city, but we are left with a picture that is much less dramatic than what the Strib achieved with the simple slip of using a particularly extreme census tract to represent the entire city.
OK, but what is going on in those two specific census tracts that makes them stand out as so extreme? And is that all that’s pulling the city-wide numbers up into the teens? Our old friend the cutoff of two vs. 10 may play a role, but that’s a squishy judgment call, so first, let’s dig into another outright misstatement. The two quotes about “more than 20%” and “more than 25%” both use the phrase “of the single-family houses.” And that’s not what the numbers are—they include townhouses and condos. Thankfully, the Fed has checkboxes we can uncheck, so let’s do that.
Limiting to just single-family homes but leaving everything else alone (in particular, cutoff two), the city of Hopkins drops to 2.4% investor ownership, less, for example, than neighboring St. Louis Park, which is 2.6%. Likewise, Fridley is down to 3.5%, less than neighboring Columbia Heights’s 4.2%. Changing the Fed’s tool back to the census tract level, and limiting our search to just single-family homes drops the Hopkins stand-out tract from 24.% to 2.1%; the tract in Fridley drops from 27.1% to 4.6%.
In both cases, the seemingly huge numbers are being driven not only by the focus on extreme census tracts but also by the inclusion of a large number properties that are classified as investor-owned condos. I’m really not sure what’s going on with those. If someone cared, it would make sense to dig into the specifics of those investor-owned condos. But that’s a side issue compared to the fact that the article’s claim about these cities just isn’t true. National corporations are not buying up a huge fraction of single-family homes in Fridley or Hopkins. In fact, if we not only focus on single-family homes but also follow the Fed’s lead and turn up the cut-off to10 to better approximate those national corporations, we find that the city of Fridley has investor ownership of 1.1%, only slightly over the metro-wide percentage.
And although this number has been rising since 2006, the trajectory has closely followed the metro-wide one; there’s no sign that some huge gobbling-up is taking place there. Instead, we’re back to the article’s opening, which was establishing the contrast point for their dramatic lead: “Twin Cities residential real estate has been slowly reshaped over the past few years by national buyers who turn single-family houses into rentals, but the numbers appeared small against the overall market, and the effects marginal.” Having debunked the dramatic contrast with that, we’re left with a slow reshaping that is small and marginal.
The data for Hopkins are even further from the article’s narrative. With the entire city included, a limitation to single-family homes, and the investor-owner cutoff set to 10, the investor ownership is only 0.3%, well below the metro-wide 1.0%. And unlike Fridley and the metro as a whole, Hopkins’s percentage has not been rising. It was highest in the early years of the dataset (2006–2008), fell off from there, partially recovered, and has been quite stable since 2014. So very definitely there is no sign that Hopkins single-family houses are actively and recently being bought up by national corporations.
Having broached the topic of how investor ownership has changed over time, let’s dig into that more. This isn’t something the article really gets into much, aside from showing the metro-wide rise from 1.8% to 4.1% (using the cutoff of two and including townhouses and condos). But you’d think it would be an important topic if this is a story about what’s been happening lately to transform the situation rather than just about the longstanding reality. Recall again the article’s lead, “Now, data show that some neighborhoods in Minneapolis and the entire suburbs of Fridley and Hopkins are being transformed by corporate ownership, a phenomenon that shows no signs of letting up.”
So let’s take a look at one of the North Minneapolis census tracts that genuinely does have high investor ownership of single-family homes. To make it personally relevant, I picked a tract in the Folwell and McKinley neighborhoods where one of the owner-occupants is a family member of mine. Even with the limitation to single-family homes and cutoff of 10, that tract clocks in at 11.5% investor ownership (141 houses out of 1,231). Throwing townhouses and condos into the mix doesn’t change that by much, but if we also crank the ownership threshold down to two, we get the Strib’s number of 22.3%, the fifth-highest on their list of top 20. But the question is, how has this number changed over time?
Using the Strib’s methodology, the percentage of investor-owned homes in the census tract started at 12.6% in 2006, rose to a peak of 27.3% in 2017, and over the past four years has fallen off to the current number of 22.3%. If we use the more narrow criteria of cutoff 10 and only single-family houses, the numbers are lower, but the pattern is the same: starts at 1.4%, rises to a peak of 14.4% in 2017, and now has fallen back to 11.5%.
What can we take away from this? There are some specific areas, this tract included, that have long faced challenges to owner occupancy and even in 2006 had substantial numbers of single-family houses rented out by large-scale landlords. As the bubble in subprime mortgages and home prices burst, this investor ownership became more pronounced, continuing to grow for a decade. However, the recent trend has been for increasing owner occupancy in these hard-hit areas. Is that the story the Strib’s article tells? No. But streets.mn readers should recognize it as the story the data tell.
All images are from the Minneapolis Federal Reserve’s data tool.