Predicting Commercial Real Estate Values? Use House Prices!
On September 7, 2017, Amazon announced that it started its search for a second headquarters, comparable in size to its very significant Seattle presence. The rest of that parable is history, at least for the casual consumer of daily news.
For investors in real estate, however, the rollercoaster of the Amazon search, selection, and, eventual, deselection of an additional location for its headquarters had and continues to have material implications. After all, just the prospect of landing the headquarters of a company planning to employ 40,000 highly paid workers, let alone the creation of thousands of indirect jobs, makes mouths water. Think of the apartments and homes needed. Think of the office space required. The retail that all these hipster millennials need. Just think of the yoga studios alone. Many markets on the shortlist announced January 18, 2018, saw heightened investor interest in anticipation of Amazon’s final choice. Of course, when the final word came with not one but two additional future headquarters — Long Island City (NY) and Arlington County (VA) — capital markets really started to rumble. Then when Long Island subsequently got nixed, earmarked capital immediately began to dissipate.
Amazon HQ2: Putting a value on a (great) prospect
There is a lot of speculation about the impact of corporate moves such as the Amazon example (but also: Google’s increasing presence in Manhattan, Facebook’s move to San Francisco, etc.). Understanding the impact of the looming arrival of the world’s largest company (by market cap) on the value of commercial real estate assets presents a significant challenge. After all, commercial real estate assets don’t trade frequently, and certainly not with sufficient numbers to measure market changes in pricing at the local level. (I’ll leave the technical details of index construction out, but ideally, you measure market trends by analyzing repeat sales, or multiple sales of the same building — RCA, for example, is using that technology for their indices.) To assess the appreciation or depreciation of commercial real estate, the industry, therefore, relies on manual appraisals, typically done once every 1-3 years for a given building. But, how does an appraiser incorporate the prospect of a company establishing presence in an area?
Intuition can certainly give hints. Conservative estimates show that 40,000 people need 4 million square feet of office space, 30-35,000 apartments (even on an Amazon HQ wage, you can’t always afford a studio or home for yourself), and the average U.S. consumer has 23.5 sq.ft. of retail space at its disposal (although I suspect that Amazon employees live their shopping life predominantly online). That’s a lot of additional demand for space, and some of it is already materializing! Additional investor capital will flow into the market, lowering the return expectations (the “cap rate,” in real estate jargon) and increasing prices. But it’s easy to see that translating intuition into the valuation of real estate is hard!
Using REITs to gauge market sentiment
One way to gauge the impact of Amazon on the local commercial real estate market is to look at the price of publicly traded REITs with exposure to either one or both Amazon locations. For example, JBG Smith, an office REIT, has about a third of its portfolio located in Arlington County (other REITs owning assets at that location include Simon Property Group, Equity Residential, Ashford Hospitality Trust, Vornado Realty Trust, and a few more). The announcement by Amazon wasn’t necessarily the cleanest ‘event,’ as markets had some anticipation of location choice (a CEO that has a home in the area seems to be a reliable predictor of company HQ choice). But still, the stock price of JBG Smith outperformed the broader REIT index around the announcement of Amazon, and its stock rose further when Amazon decided not to go to NY after all (Schadenfreude, as the Germans call it).
REIT-owned assets in Arlington County (highlighted) and beyond.
REIT-owned assets in Long Island City, NY (highlighted) and beyond.
The housing market as a leading indicator for CRE values
The change in REIT pricing reflects investor perception of the change in portfolio asset value. However, it doesn’t tell us exactly or directly how the valuation of individual assets changed. Another (indirect) way to look at the impact of the Amazon HQ location choice on the commercial real estate market is to analyze housing prices. Compared to the commercial real estate market, the housing market is liquid and deep — an average of 14,000 homes and apartments sell every day! The leading homes sales and rental platform, Zillow, provides a detailed set of monthly home price and rent indices, available at the ZIP code level. After the announcement of Amazon to split the “new” HQ between Arlington and Long Island City, Zillow analyzed predicted price and rent growth. Now a few months after the announcement and with hindsight knowledge of the retraction out of Long Island City, we can look at actual data.
The figure below plots the development of the Zillow Home Value Index (based on Zillow’s Zestimate, standardized to 100 on August 1, 2018). Interestingly, housing prices in Long Island City do rise in anticipation of the Amazon announcement, but they quickly decline after the Amazon announcement to locate in Long Island, perhaps in anticipation of what was to come? For rents, we see a reverse effect. Asking rents declined until just before Amazon made its announcement to move to Long Island City. Unfortunately, that growth began to decrease right after the pivotal moment of the NYC pullout.
Using automated valuation models (AVMs) for CRE
While it may seem far fetched to relate housing market price and rent developments to changes in commercial real estate markets, that’s precisely what a modern data-driven valuation does. It relates the price of commercial real estate assets to quantitative, observable fundamentals in the local market and in the macroeconomy. Of course, this plethora of information cannot filter into property valuation manually, it requires machine learning models that already power so many products and industries. Automated valuation models (AVMs) for commercial real estate are able to incorporate hundreds of variables (technically, the number of variables could be unlimited, but some adult supervision is helpful), using sources updated as frequently as daily. Think about the movement in the stock market and in interest rates. Both input into cost of capital and thus the value of commercial real estate. Think about jobs, income levels, and changes. Consider ‘hyperlocal’ metrics such as crime, events, amenities, and access to public transit.
AVMs, housing prices, and the effect of Amazon
The GeoPhy AVM includes a number of features that reflect levels of and changes in house prices — from local house price indices to permit data, and from local housing market rents to market velocity. The picture below shows how GeoPhy valuations of commercial real estate change based on a percentage change in house price-related features (note this is ceteris paribus, so everything else in the models stays constant). Now, what would the impact of Amazon on Long Island City’s commercial real estate market have been? There are many factors that would have changed in the process, but looking at changes in housing prices alone, we estimate that commercial real estate values increased by approximately one percent in anticipation of, and subsequent to Amazon’s decision to locate in NY. We’re still in the process of bubble deflation, but we expect this effect to disappear rather quickly. For buy-and-hold investors, similar to those following the Amazon saga with nothing but bemusement, the Amazon flip-flop was thus merely a blip, and they likely haven’t even noticed the change in value of their assets.
But. . . our AVM indicates that real value was created in the Amazon location decision, and real value was eventually destroyed!