Sales data: What's in the basket?
Here's a bit of a conundrum for the data wonks among you: I wrote earlier this week about the Case-Shiller index, which shows that Boston-area prices declined by 4.6 percent in February. This morning I got a new report from Radar Logic, a real estate analytics firm, showing that Boston-area prices declined by 11.3 percent in February.
Boston-area homes were selling for about $200 per square foot at the end of February, according to Radar Logic. That means a 2,000-square-foot home was selling for about $400,000. Prices per square foot last reached these levels in 2003.
That's the news. Now comes something a bit technical, but hopefully worth the trip.
Both Case-Shiller and Radar Logic use complicated formulas to get around a basic dilemma: The mix of homes sold in a given month can create the illusion of a change in home values. If more larger homes sell in a given month, average prices may appear to rise, even if the price of every home has stayed the same. If more smaller homes are sold, prices may appear to fall. Radar Logic divides the house price by the number of square feet to create an even comparison of homes of all sizes. Case-Shiller derives trends by comparing multiple sales of the same homes.
My first thought was that Radar Logic's data may still be skewed by changes in the mix of homes sold, because smaller and less expensive homes generally have higher values per square foot than larger and more expensive homes. If that's true, it should follow that Radar Logic will report a relatively healthier market than Case-Shiller in months when a relatively larger number of small homes are sold.
It has been my assumption that many of the people forced to sell in this market were the owners of smaller and less expensive homes. Among other things, this explained why Case-Shiller has been reporting smaller price declines than Warren Group, whose raw data on median sales prices is skewed downward by more small-dollar sales.
The problem, of course, is that Radar Logic reports a much bleaker situation than Case-Shiller.
Is it possible that the sales mix actually is tilted heavily toward more expensive homes? One indication the answer is yes comes from the Case-Shiller data, which breaks down sales trends into three bands based on home prices. In February, Case-Shiller reported that Boston homes in the least expensive band lost 14 percent of their value while homes in the top band lost 1 percent of their value. Without diving too deeply into math, the fact that the overall Boston index fell 4.6 percent implies that the mix of sales is indeed tilted toward more expensive homes.
I'd be interested to hear your thoughts on this, and I'd be very interested to see any data bearing on the question.
Enjoyed this post? Get blog updates delivered to your reader. Click here.
I think you may be right about the case shiller mix.
The truth is any data at this level is really too fuzzy to get anything but a general trend indication out of. People love to read too much into small numbers, but they aren't as useful.
In any given month it doesn't take that much to skew numbers over small data sets, but that should be smoothed over time.
And I'd call even a fairly "big" market like south end condo sales a small data set.
The S&P/Case-Shiller, Radar Logic and OFHEO data essentially tell the same story... prices are heading south.
Radar Logic data strongly captures the seasonality of the housing market, particularly for Boston, but normalized, Case-Shiller and the Radar Logic data (and even the OFHEO data) are strongly correlated.
Here's a more thorough comparison of the data
And additional analysis...
Here is a link to the latest CSI for Boston normalized to the latest Radar Logic data for Boston...
This brings up another interesting point though... It appears that the latest Radar Logic data may be indicating that Boston is now experiencing the price "free-fall" phase of the housing bust where prices fall even through the spring and summer months where strong seasonal demand usually drives prices higher.
This occurred at about this same point in the 90s bust... sales volume collapsed and prices fell for over 20 consecutive months.
You say "smaller and less expensive homes generally have higher values per square foot than larger and more expensive homes."
But in certain places (say Beacon Hill) I've observed the opposite to be true at times. For example I will see a 1000 ft condo go for $700k ($700/sq ft) while a 3000 sq ft brownstone mansion will go for $3M ($1000/ sq ft). Presumably this is due to the higher quality and luxury features of the mansion.
I'm sure you know more about the market as a whole than I do, but perhaps your assumption of the price per square foot trend is not holding true for the mix of homes being sold these days?
Your commentary illustrates beautifully why tracking short term residential real estate pricing trends is a fools errand, especially in the Boston Metropolitan area.
In the Boston area we are not dealing with acres of similar tract houses, but a myriad of small neighborhoods with incredibly diverse residences constructed over decades.
Attempting to commoditze Boston real estate in order to force the data is a game for academics with surplus cycles and a surfeit of capital.
Prices are headed south. Thats all you need to know.
Radar Logic data is interesting, but the data and method are fundamentally flawed if the goal is to accurately measure price change - a bunch of mixed fruit (ever-changing and unknown mix of property types, etc.). Note that square footage data is unlikely to be consistently accurate, and in any event, is a very crude hedonic adjustment. Any way you slice it, it's a noise-making median. Case-Shiller has its own shortcomings, but plainly more robust method are more trustworthy for gauging home price trends over time. @250% difference in YoY Boston change figure strikes me as material.
Shall we cheer.
The problem I see with Case Shiller is that smaller homes that have more turns per period is probably more accurate, but larger homes turn less frequently so the data set is going to be smaller and a spike or dramatic drop in one community or even one home can affect the outcome.
This is all very fascinating to me as a real estate agent and a former sales forecaster for a Fortune 500. Basically, you cannot look at the time frames as closely as most forecasters are and you really need to categorize and study each market. Remember markets can be geographic, price driven, category driven or even seasonal. To look at it as a whole and then to make generalizations about each market is a fool's errand.
I think it is safe to say there is downward pressure on all housing prices whether that means prices are down or the market is neutral depends on what you are looking at.
I think the Radar Logic might work well for more commoditized homes such as the traditional Post-war Cape or a 2 bedroom condo in Dorchester with large inventories and sales movements, but larger homes can vary widely in price because they are more subject to trends, style and the whims of buyers because they are in some form a fashion statement, not a necessity.
Great comment John Galt.
I see no such "conundrum"
The rich are getting richer (and consequently there is less slide in price at the top), while the poor are getting poorer. There is actually plenty of data out there to back this up - since the current bunch of thieves in the White House came to power real incomes in the top 1% have risen many many times more than they have for the rest of us.
This blogger might want to review your comment before posting it.
Recent Posts
browse this blog
by categoryINside Boston.com