Pub. 4 2016 Issue 3

www.uba.org 14 even tell you how many sold in your state. Keep in mind, they can’t tell which direction the data moved that month 75% of the time on the national level either, much less locally. But that’s not the case for all Census data. The permits we discussed above don’t provide a huge amount of detail, but they are much more reliable because more per- mit-issuing places provide the data each month. The margins of error on permits are comparatively tiny. Often times people in our industry like the idea of housing starts instead of housing permits because it indicates actual construction activity, but be careful, because if you are using Census figures you’re dealing with a much higher percent- age error with housing starts than permits (Metrostudy’s housing starts data are a different story, they are more reliable than both, we’ll explain later). #6 Highs and Lows So apart from being able to tell whether we can compare a data set to last month or need to look at a year ago, or whether the change was statistically significant, or whether it’s preliminary and likely to be revised, or whether it contains the types of structures we think it does, or whether the sample size is good enough to be taken seriously, we come to one of the topics that has some appreciable whimsy to it, and that’s the occasionally flamboyant language used to describe a move in data that is relatively insignificant relative to the peaks and troughs of the dataset. Many of us in the industry are guilty of this from time to time as it is in our interest to draw attention to positive moves. But that doesn’t mean we don’t need to take a more measured approach before we act on the data. Take our previous example of February new home sales. US News, via AP, declared that, “US Home Sales Surge in February.” We’ve already established in Section #2 that this statement is not entirely accurate because the Census doesn’t actually know if they increased or not. But further, the 15-year peak new home sales was over 1.2 million and the 15-year average for new home sales is 725,000, so the 536,000 (preliminary) new home sales reported in February are 26% below average and a mere 45% of peak, and this is after a couple years of housing recovery. Besides, in most parts of the country there isn’t going to be a “surge” in activity in the middle of the winter, so applying these adjectives to seasonally adjusted figures is dubious already. Metrostudy’s Brad Hunter pro- vides a more measured commentary, noting that while the February figures are the highest since February 2008, the outlook is for gradual improvement, no surges, spikes, or rocket ships! Closing and Takeaways So you might wonder what is the best way to avoid so many of these pitfalls in the mysterious world of housing data? As I mentioned before, looking for data sets like the Census Permits that have a very large sample size is a good start. But many times to get good information you need to collect it yourself or get it from someone that directly collected it. And that usually isn’t free. Ask any real estate consultant if they trust information from someone who hasn’t physically visited comparable developments before they provide their recom- mendations. At Metrostudy, we are fortunate in this regard because we do primary data collection on a massive number of subdivisions nationwide. And while we often refer to it as our survey, it’s really more of a census, as we physically visit every development with ongoing activity each quarter. This may sound like a commercial, but it’s difficult to characterize it any other way, as it’s an unparalleled research effort in the residential construction data space. We hope you find this document helpful, and please contact us with questions, comments, corrections, or great examples you come across of some of the things we discussed here; contact information follows our Takeaways below. As we noted in the introduction, this is far from an exhaustive list of data challeng- es, but these are among the most common we encounter. Don’t forget to check out the Key Housing Indicator table, it describes which indicators to keep an eye out for, how frequent- ly they are released, and key thresholds to remember. It’s a great reference so that you don’t have to remember the nuance of each data release. #1 Takeaway - Seasonally adjusted data: can compare month to month, represents an annual number, but is a statistical creation. Non-seasonally adjusted data: data is more raw, but can’t really compare month to month, need to compare to the same month a year (or more) ago. #2 Takeaway - Margins of Error and Confidence Intervals: just because a change is reported, doesn’t mean it really happened. Check to see if the change is statisti- cally significant before relying on that data for business decisions. #3 Takeaway - Be Wary of Preliminary Data: Always remember the data reported today will probably change, potentially by a lot, so don’t bet the farm on early estimates. #4 Takeaway - Single Family is a big tent: The U.S. government considers many structures that we think of as “attached” like townhomes and side-by-side condos to be single family homes. #5 Takeaway - Sample Size is Important: Not all data has the same sample size, be careful of drawing stark conclusions based upon data that only looks at a tiny slice of activity and extrapolates it to a national number. It may be a rough bellwether over time, but month to month it can be misleading. #6 Takeaway - Keep It In Context: Watch out for gratuitous language describing a single period of data without putting it in the context of historical peaks and troughs. Such language is especially suspect when describing seasonally adjusted figures from traditionally slow times during the year. n For Questions or Comments, Please Contact: Jonathan Dienhart Director, Custom Services and Published Research Metrostudy, a Hanley Wood Company jdienhart@metrostudy.com  Housing Data Mistakes — continued from page 13