Investment forecasting: Art & science

Market participants agree that it’s a strange and challenging time to be forecasting valuations, as we are met with a barrage of daily news on everything from new economic indicators and diplomacy to global warming and AI development. Compounding the task is a dearth of transaction data and reliable comps. “Every day there is new information or a new event that needs to be factored into valuations, yet it feels like every day we have no data to actually work off in order to create a good forecast,” says Jorge Paredes, senior director, Advisory at Altus Group.

Within this landscape, investors are learning more about the “art” of forecasting. At the same time, the adaptation of technology is also moving at an increasingly blistering pace. What resources are investors using to help pinpoint cap rates in the current volatile market? And what does the future hold for more automated forecasting? A panel of industry experts shared their strategies and perspectives in a session aptly titled Investment Forecasting: Art and Science at the recent Altus Connect 2023.

The biggest challenge across all asset classes is the absence of a functioning debt capital market, which has undeniably constrained buying and selling. The office sector faces heightened uncertainty due to secular changes impacting the demand for space. Historically, determining valuations has been a relatively simple exercise, in the sense that there were always ample sales comps. Investors have been able to see where the trends were moving, whether yields were coming down as interest rates went down, and where rents were moving. “Those data points are far and few between today, and that really is impacting both setting marks on existing assets for the open-end funds, and separately, underwriting anything new,” says Kevin Kujawski, partner, president and COO at Menlo Equities, a private equity investment firm.

With credit to new technology products and an influx of data sources available to help develop projections, forecasting methods have been getting better and better. They were never entirely exact, but the tendency was usually right, notes Tal Peri, senior vice president and head of US East Coast & Latam, Union Investment Real Estate, Germany’s largest open-ended fund with about $60 billion globally in AUM. However, forecasting couldn’t predict events, such as COVID, or the fact that a metro would land a major employer like Tesla or Amazon. To a degree, there are always unforeseen events occurring outside of the data that can shift forecasting. However, the pandemic really magnified those events. For example, demand for e-commerce exploded during the pandemic. “So, forecasting does become more of an art, but the speed of all of these changes is so extreme,” adds Peri.

Art and science play a role

Some argue that science becomes even more important in the absence of a volume of reliable comps. “In times like these, we need to rely more on the science and less on the art,” says Heidi Learner, head of innovation, Altus Group. “We all have fewer comps to rely on, but maybe through our historical work we can figure out the comp that we do have,” she says. For example, maybe a comp isn't very good because the comp is in a neighborhood 10 miles away or it’s 20 years older than the asset being valued. “We can actually model the impact of those features and come to some resolution in the absence of a whole lot of transaction data,” she says. “It’s not perfect by any stretch, but the reliance on models becomes more important.”

Another example is being able to aggregate some of the data points that go into your ARGUS models or your DCF models and relying on that science to gather the insights to measure what is driving the gap between a transaction price and a valuation. Investors can look to other capital markets variables to explain some of that pricing gap, such as daily views on what's happening to credit and volatility and interest rate expectations. “By leveraging some of that data and turning a little bit more to the science, it can help us out in illiquid times like these,” says Learner.

Private investors are looking for valuation cues in what’s happening in the public markets and the sharp drop in valuations that has occurred within many REITs. However, appraisers are still saying that values are up marginally because NOI has floated up and cap rates are relatively the same. “When you look at some of the other benchmarks out there, whether it's the ODCE or performance of the non-traded REITs, they're not taking a lot of adjustment downward right now. So, there are some big disconnects in the market right now", Kujawski adds.

At the end of last year, one of the things Menlo Equities looked at to better understand portfolio values in its open-end fund, was where values and spreads were at pre-pandemic when the overnight rate was at 2.5 to 2.75% (before the Fed started cutting rates). That prompted some marginal adjustments to asset valuations. Another challenge to forecasting is what’s going to happen with rent growth. Given the shadow vacancy in even robust markets like Silicon Valley and Austin, there is going to be some downward pressure on rents, notes Kujawski. “The old method of underwriting to 3% rent growth probably doesn't hold. So, I think it's a really challenging time to just pull data, particularly historical data and put that through a model and arrive at a valuation. I think there’s a lot of nuanced art,” he says.

It’s also important when you look at comp sheets to know what’s behind it, adds Peri. For example, the purchase of a building might come down to two final bidders. If the winning bid is 0.2% above the second bid, that’s a very efficient market. If the bid is 10% over, then that is a different story. It’s important in this market to look at the story behind the comp, says Peri. What is the capital source and the financing that was used in the transaction? Investors don’t want to be misled just by believing in a number and paying that same number. “Data can help, and technology can help,” says Peri. However, sometimes it comes down to calling people to find out a little bit of that story to make sense of the comp and what contributed to that price, he adds.

Can ChatGPT automate valuation?

Investors are finding different ways to bridge data gaps. It’s important to keep in mind that there is a global capital market environment, adds Learner. For example, there may be only a few data points for Miami multi-family, but the investment set isn't limited just to Miami. By looking at sales in other markets that take into account the supply, investors can aggregate and also standardize what might be a limited data set, says Learner. For example, you can look at sales comps in relation to where the financing rate was at the time of a transaction. “So, there's still an ability to be able to connect the dots. It's just a matter of using a wider pool of dots than we had previously,” says Learner.

In a market where AI is fueling significant advances in data and analytics, will automate valuation modeling become a reality in the not-too-distant future? There was a time when people just did “back of the envelope” calculations, notes Peri. Then came Excel and ARGUS and people got a lot smarter. The reality is that investors are usually acquiring assets in a competitive bid situation. If two investors are bidding on the same deal and each has the same automated valuation, at some point, one will have to decide who outbids the other. “As an investment person, you still have to know how to digest it and how to read it and figure out where you stop on the bid for an asset,” he says.

Some of the applications for automation are very realistic, including using technology to automate some of the more manual processes, such as gathering comps. One of the challenges for real estate is that there’s still a lot of information that is closely held and, as a result, not made public and easily accessible. “If we think about the information that we have at Altus, being able to harvest our own internal data sets and scrape that data can make our process a lot smarter in terms of using our own data in a better way and supplementing it with some of the data that's publicly available,” says Learner.

The industry does have more science to help with valuation forecasting, and more advances that are moving towards a future where automated valuation is possible. “Maybe it's not an automated valuation model, maybe it's an augmented valuation model that's augmented both with technology and with human intervention,” adds Learner. “Because we're not nearly at the point where we can just be hands off, much the same way we're not quite at the point where we can be hands off with driverless automobiles.”