We get asked a lot about our “secret sauce” for multifamily forecasting. How did we get good at it – or at least a lot BETTER at it? No forecast is ever perfect, but we’ll put our track record up there with anyone.
RealPage correctly predicted the Sun Belt’s sustained outperformance (even pre-COVID) in the early 2010s. RealPage also made correct calls on suburbs, on Class B and on the minimal impact of escalating new supply levels – all counter to Wall Street narratives at the time.
And when COVID-19 hit, RealPage correctly made two calls that appeared bold at the time: There would be no big spike in either vacancy OR evictions.
Of course, we’ve had some misses, too. (For example: While we were bullish on the Sun Belt, we incorrectly pinned Tampa as a laggard.) But we’re proud to get it right more often than not. So, what’s the secret?
Getting it right comes down to 10 things NOT to do. These are very common mistakes of multifamily forecasters – including some things we learned along the way, too.
1. Don’t build a one-size-fits-all model. Everyone knows New York City and Oklahoma City are very different markets. So you HAVE to account for market-specific variables/drivers. That should be a no-brainer, but very few actually do it because it’s time-consuming and complicated.
2, Don’t overweight the impact of supply. This is a big one, and why Wall Street perpetually gets the Sun Belt wrong. Too many models judge supply’s impact by looking at prior peaks in supply – which were 2001-02 and 2008-09… when supply was definitely not the root problem.
3. Don’t overweight anything happening in the for-sale housing market. Conventional wisdom is wrong here. There is zero evidence that a strong for-sale housing market hurts apartment demand. In fact, they’re generally correlated. A rising tide boosts all ships.
4. Don’t use household income stats on broader population as a variable in rent forecasts. Apples and oranges. You have to look at incomes among those who actually live in the segment of the rental market you’re forecasting.
5. Don’t assume something is right “because the model says so.” Models are only as objective as the variables and methods you subjectively choose to include or exclude. You have to build in structural flexibility — the “art” of real estate. If your model produces outcomes that are tough to justify or explain, that’s a problem. Good forecasts can be explained with logic – not just by econometric speak.
6. Don’t assume the world never changes. Most real estate models are not dynamic enough to detect changing relationships. Too many forecasters mess up on ranking individual markets by reverting to the old orders of yesteryear.
7. Don’t ever use a “rule of thumb” like “jobs per unit of demand.” That’s a junk metric. Ratios are fun but grossly overrated. And remember for rentals: Demand is absorption. You can only absorb what’s available — no matter how many jobs created.
8. Don’t use false “reversion to means.” When you average out the best of times and the worst of times, you get basically flat numbers. Pick a side.
9. Don’t limit (as best you can) possibilities based on time or assumptions. Modern data science tools can analyze much more data much faster – and provide far more potential scenarios – than traditional econometric modeling software.
10. Don’t forget to learn from your mistakes. If you’re always too bullish or too negative on certain metros or certain market segments, make the adjustments. Understand your biases – which are often shaped by where/how you live.