New twice a year public report back tests accuracy
Recently, industry blogger Rob Chrisman tallied up the number of home price indices (HPIs) that are published each month and said, by his count, there are now at least 14. “As the years have gone by, I am overwhelmed with them,” he wrote. “How are we to think about them all since, there are so many?” Interestingly, my team at CoreLogic had been grappling with the similar question: How could we demonstrate the accuracy of the monthly forecast, which projects our CoreLogic HPI 30 years into the future, to the numerous constituents who use the data?
Since we launched the improved version of the Forecast in June of 2016, we have been internally monitoring its accuracy. We decided to “open the kimono” and show the world how good or bad the Forecast is by comparing how much the forecast predicted homes prices would change over the last 12 months and how much they actually changed. The output of this exercise is the new CoreLogic HPI Forecast Validation Report, which we released last month, and going forward, we will share twice a year.
Specifically, our first report, published this past October, summarizes the back tested accuracy of our national and CBSA forecasts for the period from June 2016 to June 2017.
Here’s what the report showed. You be the judge on how well we did.
- Our 12-mont national forecast was within 0.7 percent of the actual HPI increase.
- The most accurate CBSA-level forecast was for the Phoenix-Mesa-Scottsdale, AZ area, which came within 0.4 percent of the actual HPI increase of 6.2 percent (Figure 1).
- The widest CBSA gap was in Seattle-Bellevue-Everett, WA with an 8.4 percent under-estimation of actual increase. Among other factors, the variance in this over-valued CBSA was due to unexpected acceleration in prices in early 2017 after a price deceleration earlier in 2016.
- The average absolute difference between the actual HPI 12-month increases and the forecasted 12-month increase for the 50 largest CBSAs was 2.5 percent.
- Among the 15 most accurately forecasted CBSAs, the average difference was 0.9 percent, and the range was between 0.4 percent and 1.5 percent.
Why are these points so significant? Well because products like the CoreLogic HPI and Home Price Forecast (HPF) are used to price portfolios and are incorporated into sophisticated risk models used for bank stress-testing required by regulators. And in the next few years, home price forecasts will become even important as financial institutions are required to predict future loan losses as part of the Current Expected Credit Loss (CECL) accounting standard. For our clients, our new report means that they can get specific insight into the degree of accuracy of our HPI forecasts at the national and CBSA levels, which gives them confidence in the accuracy of their own predictions. For CoreLogic, this translates into ongoing commitment to our subject matter and modeling experts to keep our HPIs in the market leadership position for accuracy, predictability, timeliness and stability. When you make a commitment in front of audience, you have to follow through!
© 2017 CoreLogic, Inc. All rights reserved