By Chris Fraley, Chief Investment Officer, RealtyMogul
Approximately ten years ago I was conducting final round interviews for a very senior Asset Management position. Somehow the conversation turned to tough decisions that we have had to make as Investment Committee members. The interviewee proudly recounted, with a tinge of arrogance, a story about a deal she killed while an Investment Committee member at a top real estate private equity fund due to her own personal experience. It was a retail center she had worked at as a teenager and recalled a shooting that had occurred at the center and rejected the asset as a potential investment.
Huh? Was she really admitting to me that she allowed her own personal bias from an experience decades ago taint her view of the opportunity? Shouldn’t we look at every investment opportunity with fresh eyes, understanding that neighborhoods may change and evolve? Needless to say, she was not hired.
In his book “Animal Spirits”, Robert Shiller discusses the need to understand the importance of human emotion in economic decision making. Fear can cause paralysis during the best acquisition periods while over-confidence can cause investors to strike at the moments of greatest economic risk. Technology probably will not completely replace human decision making, but it may become an even more critical and essential tool for decision makers.
The technology exists today to generate an entire pro forma for nearly every apartment building in the country simply by inputting an address. Using machine learning tools and artificial intelligence, this technology will become more precise over time as more data is inputted. There are also tools to analyze the competitive set of an asset and track its performance relative to the competitive set. This is rapidly becoming a game changer for our industry. Below are seven ways that artificial intelligence, machine learning, and big data could disrupt our industry.
I have never been a huge fan of appraisals. Appraisers generally rely on cap rates over the trailing 6-12 months. Unfortunately, the actual pricing is often set several months prior to the closing when the bids come in and the buyer is selected. The best indicator of market is upon selection of the buyer. Investment firms that are actively bidding on and selling properties, and have the technology to track and analyze the sales, could have an advantage in the future by seeing movements in markets possibly 18 months prior to the appraisal industry, which typically relies on older data.
The cap rate can also be unreliable because it does not tell whether the metrics are being applied to forward or trailing numbers, if there is an adjustment for any increase in real estate taxes upon closing, or adjustments to any capital requirements, levels of reserves, etc. These factors can easily cause 100 basis point swings in cap rates. Technology, however, can take the comparable sale and rebuild the financial pro forma to truly understand the metrics of the sale.
Finally, Appraisers are human and can often lack complete objectivity. In the run up to our last economic crash in 2008, appraisers were often criticized for being too heavily influenced by their clients for fear of economic retribution. Technology can provide a more objective valuation for real estate investors through utilizing machine learning and big data, which really should be the goal of any appraisal.
2. Research Functions within Large Institutions
In 2010, when I was working at a large private equity firm in New York, I was tasked with building out an internal research function. I chose as my first task to meet as many heads of research of the big institutional real estate players as I could and find out how they structured their group. These groups are mostly used as an internal check and balance to ensure aggressive acquisitions officers don’t rely on overly optimistic assumptions in their numbers. In many of the large institutions, the rental growth rate assumptions are dictated by the research departments, who usually rely on some combination of output that the three or four market prognosticators generate through very complex financial models.
Unfortunately, the models sometimes do not accurately capture the subtle nuances in submarkets that can create value. For example, the experts might say that NYC office rents will grow by 3%, 3%, and 2% over the next three years, respectively. That may not capture market moving events on a micro level, like the explosion of high margin private equity firms in the Plaza District in the 2003 timeframe, or Google forever changing the Meat Packing District with the acquisition of 111 8th Ave.
I witnessed the limits of the current real estate research model first hand in 2005. We were selling an office building in Orange County, CA and the bids came in far higher than expected. In the interview with the selected buyer, we tried to understand the motivation behind the aggressive offer. We were told by the selected buyer that they were limited in projecting rent growth assumptions by their Research Department, who were in turn limited by the economic forecasting services to which they subscribed. These services had identified Orange County and West Palm Beach as the areas of greatest potential for rent growth. At the time, we had recently purchased one of the top office buildings in West Palm Beach, FL. We also put that asset on the market hoping to take advantage of the same market enthusiasm.
Again, the offers for our West Palm Beach property far exceeded our expectations and there was an outright feeding frenzy among the top institutions in the US. The same broker that sold us the building in 2003 told us that we were the only major institution in the bidding process at that time. The economic prediction services, upon which many of the major institutions had relied, had turned a sleepy secondary submarket into a hotbed of economic growth and value creation.
Shortly thereafter, I was invited by one of these real estate forecasting services to attend their annual client conference. There was a presentation from one of their lead economists, who was a former rocket scientist with no apparent real estate experience. He flashed the “secret sauce” of their economic prediction up on an enormous screen which included about 30 co-efficients. The result was a prediction that the best risk-adjusted return for office investments in the US was West Palm Beach. I did not take the time to fully understand the equation, but as someone who had sat through countless leasing meetings with our asset manager for our property in West Palm Beach, I found the predictive service to be more dangerous than helpful and canceled our subscription.
I believe Big Data and machine learning will allow us to analyze factors on a more micro level in the near future, helping us to avoid market wide generalizations and include factors which impact real estate investment more precisely.
3. Indexing – Tracking Performance Relative to Competitive Set
As mentioned above, the technology currently exists to create an entire financial pro forma for almost any apartment community in the country simply by inputting a street address. Revenues can be pulled from the various market data services on a near-real-time basis or pulled directly from individual community websites using data scraping technology. The expenses can be estimated using a combination of inputs from existing portfolios and various market data services. Machine learning algorithms can also quickly identify the competitive set for an individual apartment community.
In the near future, as technology in this space further evolves, we should be able to track the performance of apartment communities on a real-time basis relative to a competitive set in the same way the hotel industry uses STAR reports. This will be a powerful asset management tool, by identifying if an asset is underperforming or outperforming its competitive set in real time.
4. Apartment REIT Valuations
One of the challenges of investing in publicly traded REITs is that they often lose the benefit of diversification into real estate and are often more closely correlated with equities rather than real estate assets. Using real-time valuation technology powered by Big Data and machine learning, which is available today for the multifamily sector, analysts are better able to track the value of the underlying assets relative to the stock price. Instead of taking days to value a massive portfolio, it can take minutes and can inform the analyst whether a public REIT is a buy or a sell.
I predict that commercial real estate underwriting will change from a zero-based underwriting approach to an approach where we use the outputs from the machine learning tools as the base case and study where the technology may be incorrect and why.
Today at RealtyMogul, we compare three things when analyzing a potential real estate investment: the pro forma from our operating partner, our own internal underwriting, and the output from our machine learning tools.
From what I have seen over two decades of experience in real estate finance, most financial pro formas are created by junior analysts or associates who do not have the benefit of investing through cycles. I used to say that an analyst or associate in 2007 had never seen a market downturn and in 2011 had never seen a market upturn. We know that altering a few key variables in a model can move the IRR a couple of hundred basis points in either direction. Technology can provide crucial checks and balances to make sure that the underwriting is not too influenced by “Animal Spirits.”
Using the indexing strategy above, one can scan a market using advanced technology and determine which residential communities are underperforming the market and are possibly an opportunity to acquire and create value. I am convinced that the investment world of the future will be divided by those who can access this this type of technology and those who do not.
7. Removal of Bias
Too often, real estate investment professionals are impacted by human bias, which can be a dangerous factor in the process. I have always been a big fan of select submarkets in the Prince Georges County, MD apartment market. For example, the class B apartment market in Laurel, Maryland saw rental increases during every year of the worst economic downturn in recent history (2008-2013)1. Despite this strong historical performance, cap rates tend to trade at least 100 basis points higher than prime DC Metro markets. I have, in the past, raised capital for value-add opportunities in that market and I have found it pretty shocking to hear the reaction of some institutional real estate investors. I have heard terms like “demographically challenged” referring to the large percentage of African American population. In this case, human bias may be preventing investors from identifying a great opportunity.
I do not anticipate that technology will completely replace human decision making in the real estate investment process any time soon. However, I do believe that technology will play an increasingly critical role in the acquisition and asset management functions going forward.
- Per a 2016 suburban Maryland metro analysis prepared by REIS.