The Rise of the ADU
Should YOU build an ADU in your backyard?
On my ongoing job search, I have come across many different start-ups with wide ranging ideas. One of the the ideas that really piqued my interest was the business model put forward by “Rent the Backyard”. They, and other companies like them, propose to build an accessory dwelling unit (ADU) in a homeowner’s backyard and then split the rent with them until the unit is paid off. I’m intrigued by this idea because of the power it has for social change. ADUs (sometimes called granny flats) can be both a smart investment and are a potentially key part of the solution to the lack of affordable housing seen in most large American cities.
But how would a company like this, or even a motivated individual homeowner, decide that building an ADU in their backyard is a good option? Intuitively we might expect that a yes or no recommendation would be based on data like home value, mortgage value and equity, rental price, and rental market saturation in that particular area, etc. In this, the first article on the topic, we’ll be taking a look at some of these (publicly available) factors to see if we can come up with some recommendations for enterprising homeowners looking for prime ADU building areas. We’ll be using data from the city of Portland, Oregon in making our recommendations for reasons described hereafter.
Portland, Oregon, and ADUs
Portland is currently in a self-described State of Emergency on Housing and Homelessness. This gives the city leeway to expedite construction of affordable housing and change zoning codes that stand in the way of progress, among other measures. In this case, “other measures” includes incentivizing ADU construction. Portland has the least stringent ADU standards in the country and this has led to an explosion in the number of permitted units in the city. Still, the percentage of lots that currently host an ADU is less than 5% of lots that are “ADU friendly.”
Clearly we haven’t missed the boat here. There is still plenty of incentive to jump into the game in Portland. So let’s start by looking at a map of Portland. Here we can get a sense of its size and the various areas (for those unfamiliar with its layout). Portland is divided nicely into various neighborhoods. There is a river that runs directly through the middle of the city, which, while tough to see on this map, divides Portland into the West-side and the East-side. The downtown area is located on the West-side and includes neighborhoods like the Northwest District, the Pearl, and of course, Downtown. Northwest Portland is very hilly and is home to one of the largest urban forest reserves in the United States, Forest Park. Most of the neighborhoods that will be ideal for building an ADU will be on the East-side and in the Southwest.
The first thing we’ll want to look at in determining the best places to build is the current geographical distribution of ADUs in the city. To accomplish this we’ll have to download and work with GeoJSON files that encode geographical features. We will use these features to visualize all our neighborhoods and points of interest. The Python libraries most helpful here, and the ones I used, are GeoPandas and Shapely. Shapely gives us the geometries that will allow us to visually explore our data, here points and polygons. GeoPandas extends the data types allowed by Pandas to include geo-spatial data. This simplifies plotting immensely. We won’t be diving into the widespread functionality of these libraries here, but if you’re curious check out this article by Tom MacWright which details most of what you’d need to know. If you are interested in the code that makes the following plots happen, please follow along in my GitHub.
Of interest here is the functionality of polygons and points. Polygons, when graphed using the Shapely library will give us a visual representation of all the Portland neighborhoods, which we can then plot points over. Let’s do this now. Utilizing PortlandMaps Open Data for neighborhood boundaries we can plot an spatially empty map of Portland neighborhoods.
Onto this blank canvas we can now plot the location of permitted ADUs in the city. This information is also on PortlandMaps Open Data, contained in the residential building permits dataset. This data includes all residential permits issued in Portland beginning in 1994 and continuing through June 2019 (this source is continually updated). There are lots of columns, but of interest specifically to us is the column, “IS_ADU”, of type boolean, and the two columns that denote latitude and longitude. By pulling this GeoJSON file into a geopandas dataframe we can use standard pandas functionality to filter all the permits and keep only those for ADUs.
This leaves us with 3297 permits for ADUs in Portland proper over the last 25 years. To graph these points we’ll write a function that takes the latitude and longitude of each permit and returns a Shapely point. These points can then get plotted over the axis that contains our empty Portland neighborhoods map. The result is below. We can see with this map that ADUs are concentrated in the more affluent neighborhoods of inner Northeast and Southeast Portland. Given the potential costs of building an ADU, this is not surprising.
Another way to view the above data is by color coding the neighborhoods by the number of ADUs. This may be a more intuitive view but expresses the same thing. See below. The bright yellow neighborhood is Concordia. Apparently they have lots of ADUs! The other neighborhoods in order of count are Richmond, Sellwood-Moreland Improvement League (what a name), King and University Park.
AirBnB Supply?
This is a good start, but doesn’t fully capture the total number of competing units an enterprising homeowner might encounter if they were to build an ADU. The explosion of AirBnB has added another layer of complexity to this problem. Depending on the area of Portland, and their willingness to be more hands on, a homeowner could look to short term rentals as a significant income source.
To illustrate competition here we can turn to the data collected by the folks over at Inside AirBnB. They maintain a phenomenal source of scraped listings data for the purpose of examining how AirBnB has affected various cities. Check this website out if you haven’t already. Utilizing the same process as above to plot points, but this time filtering for the listings that offer up the “whole place” we’re left with 4,116 listings in the Portland metro, wow! All of these listings are plotted in the same style as the two figures shown above. We notice a similar supply pattern, but it does seem that AirBnBs are more localized around what we might call the center of the city, especially in downtown.
Okay, so we have a sense of supply of competing places in Portland proper. Given this narrow subset of data we could make a recommendation that you should avoid building an ADU in the neighborhoods that ring the ‘center’ of the East-side. But basing a recommendation on this alone would be silly! Despite increased competition, these places are probably more desirable and have higher density. We might also want to look at home value, expected rental income, neighborhood size or housing inflation year over year to inform our recommendation. We’ll use Zillow data to inform our recommendation further in the next section.
Zillow’s Home Value Index and High/Low Risk Construction
Building an ADU is expensive! There are lots of different quotes online, but given the range of values I’ve encountered, it seems that a true detached ADU will cost around $150,000 to build. Feel free to research this yourself if you’d like.
We’ll use this number to assess the risk associated with building an ADU relative to the equity contained in your home. Zillow maintains a continually updated median home value index that is methodologically rigorous and available at several different geographic levels, including neighborhood. Follow the link if you’re curious about their methodology. Downloading this data for Portland’s neighborhoods and doing some string matching on the neighborhood names using the difflib library allows us to examine, graphically again, how home value is distributed in Portland. Unsurprisingly we see that as we move away from the river to the East, home value goes down.
This information on home price, though interesting, would be more useful if we could turn it into a categorical high/low risk map that denotes ADU cost ($150,000) relative to home equity.
Let’s imagine a homeowner, perhaps the average homeowner, in every neighborhood that has paid off 50% of their mortgage. This leaves them with 50% of their home’s value(here, the ZHVI of that particular neighborhood), in equity, that they can borrow against to finance an ADU. Before the housing crisis and subsequent recession of 2008, homeowners could take out a second mortgage on up to (or even over) 100% of the equity stored in their home! Nowadays, the maximum equity a homeowner can borrow against is 75%, give or take.
Given this information, we’ll term building an ADU as high risk if the total cost of the ADU is 75% or more of the equity stored in a home owned by our typical homeowner described above. As an inequality, high risk is (ZHVI * 0.5) * 0.75 ≤ $150,000. Areas where this is false will be termed low risk. Let’s look at map of Portland color coded accordingly. Using this particular inequality, which we could adjust to different levels of equity and ADU cost, will term neighborhoods with ZHVI > $400,000 as low risk and the reverse for high risk. As such, we do see the same sort of pattern associated with the price map, wherein higher value neighborhoods are a less risky option for building an ADU.
Combining this risk map above with a plot of all the ADUs and AirBnBs we can use both insights gleaned thus far to identify neighborhoods with low competition AND low risk based on home value. This plot is below. While this is quite a busy plot, we can identify some neighborhoods with low competition and low risk ADU construction.
Without yet taking into account property type, rental prices (which me might expect to scale linearly with ZHVI), or housing inflation year over year, let’s use the information above to generate some recommendations!
Ignoring all the neighborhoods termed as high risk, turning our Shapely geometries into area projections (more complicated than it sounds), and then sorting neighborhoods by number of combined ADUs and AirBnBs per sq/km gives the following plot of neighborhoods that we might recommend with this straightforward analysis. Of the 70 neighborhoods termed low risk, these 35 neighborhoods have the lowest number of competing ADUs and AirBnBs per sq/km. As a simplistic recommendation, this is a pretty good start. If you’re curious what the names of these neighborhoods are, feel free to check out the code!
Conclusions
Using data on Portland permits issued for ADUs and scraped AirBnB listings data we first looked at a map to illustrate where competition might exist for homeowners wanting to build an ADU. Then, using Zillow’s Home Valuation Index we looked at neighborhoods that might be termed high and low risk investment areas based on some logical leaps about home equity and financing. What we’re left with is a list of theoretically suitable neighborhoods for ADU construction, as detailed in the map above. For a homeowner looking to build an ADU in Portland, Oregon, this is good conceptual starting block for thinking about the suitability of your property. Of course, your mileage may vary based on your specific home’s value and stored equity.
However, let’s refocus on the first half of the question posed at the start and restated here. How might a COMPANY (not an individual), who proposes to build an ADU with zero homeowner cost, identify neighborhoods that make for good investment opportunities? In this case, the terms are different. They are quite likely to define high and low risk neighborhoods quite differently, as they won’t be pulling a second mortgage out of the home’s stored equity. A company with this business plan should be more focused on average household income, property type and size, expected rental income, neighborhood home value increase year over year, and other more specific features like street parking availability, public transit access, and apartment zoning areas. We’ll take a look at how some of these variables might influence a company’s ideal investment neighborhoods in a later article.
If you have any questions or comments feel free to drop a line below and I’ll be sure to respond. If you’re curious about the code I used to make these insights, you can find that at my GitHub. Till next time and, as always, thanks for reading!
