Fast and Accurate Valuations Are Key for Real Estate Investors
New and Improving Tools Are Arriving Just in Time
by Katie Brewer
In a buyer’s market, the advantage goes to the purchaser who can make a wise decision faster than other potential buyers. So, for investors in today’s market, time has become the crucial obstacle to success. Some have called it a modern-day Gold Rush, and just as in olden days, prospectors who can sift through the national inventory of properties faster and spot the hidden nuggets more clearly are the ones who will prosper.
The combination of historically low interest rates and shifting demographics has triggered a buying spree in single family homes that has accelerated during the coronavirus pandemic. As a result, inventories are low across the country. According to the Radian Home Price Index, there are currently 40 percent fewer homes on the market, on average, than any time over the past decade. High demand and low supply have inevitably driven prices skyward. During the month of April, the Radian Home Price Index rose 10.4 percent, with the median home in the U.S. now costing $277,365.
But perhaps the most illustrative measure of the real estate market’s velocity is the time between the listing of a property and a pending purchase offer. According to Zillow, in April, nearly half (47 percent) the homes for sale in the United States were on the market for less than a week before going under contract and more than three quarters (76 percent) were on the market less than a month.
Evaluating any property in days—or even hours—adds a level of risk that many would be unwilling to take in a normal market. Doing it at scale locally, regionally, or nationally requires either a prospector’s instinct or a technological edge. Most choose the technology.
AVMs, AI, Machine Learning
One solution is to use an automated valuation model (AVM) that applies mathematical models to a database of real estate information such as current and historical prices, number of bedrooms, and property improvements. With an AVM you can run estimated values on properties in a matter of moments. When a new property hits the market, an investor can quickly identify its essential characteristics and make a judgment about its suitability for their real estate portfolio.
But AVMs are only as good as their data inputs. While they have the potential to deliver objective valuations without human bias, most models analyze data that has been collected in-person, and that can result in a distorted valuation. This is most consequential when the owner of the property is a member of a historically marginalized group.
According to the Urban Institute, a Washington, D.C.-based think tank focused on social and economic policy research, AVMs have produced larger pricing errors in majority-Black neighborhoods than those in majority-white neighborhoods, “potentially contributing to the wide housing wealth gap between Black and white homeowners.” To correct for bias, AVM operators often look for ways to expand or augment the data sets their model is analyzing. The more data that is available, the more accurate the final analysis will be.
But a new approach, using artificial intelligence (AI), promises the fastest and most accurate analysis and valuations short of an in-person inspection. Modern AI can analyze an enormous amount of structured data such as price histories, property attributes and surrounding market data, but also unstructured data like photographs, floor plan sketches and text or speech. Analysis of images in particular can help to establish the condition of a property and objects that differentiate value using consistent and repeatable tools without the in-person challenges. These tools can also evaluate newly built properties that have no price histories or recent comparables.
One of the most intriguing benefits of AI and Machine Learning (ML) is a model’s ability to refine and improve the process of identifying the best sales comparables to help establish a current price. Reinforcement learning allows models to improve accuracy simply through use and more data. And, unlike the human-based valuations, models can be independently tested in very large samples to ensure bias and other unintended consequences are limited or mitigated. Considering how many billions of photographs of properties there are on MLS listings nationally alone, the possibility of “better than in-person” accuracy through AI is a very real possibility.
The real estate Gold Rush shows no signs of diminishing. The Commerce Department recently reported that the median price of a new home sold in April of this year was 20.1 percent higher than a year earlier, the strongest annual gain since 1988. At the same time, the majority (54.6 percent) of homes in the middle price tier and almost half (48.0 percent) of homes in the top price tier went under contract in a week or less, up 60 percent and 71 percent year-over-year, respectively.
Just like prospectors for gold, Investors looking for fleeting opportunities in this scorching hot market need faster and more accurate ways to judge the true value of their finds. Fortunately, new and improving tools are arriving just in time.
