It's difficult for journalists to get statistics right: the real estate edition

A friend pointed me to a good article on the buoyant real-estate market in Boston (link). The journalist tells us the median home sale price in Suffolk County has gone up 31 percent in the five years since the 2010 crash.

The article makes some excellent points. When it comes to talking about median house prices, it fails basic statistics.

Here is a quote (maybe a misquote) from Barry Bluestone, whom the reporter describes as a "go-to mind" on Boston housing. He complains that "median home sale prices--the metric by which many people judge the market--can be misleading." How so?

He explains: "If a bunch of expensive homes sell all at once, it will appear like median prices have risen rapidly. And if a bunch of lower-priced homes have hit the market recently, the opposite will be true."

That is a strange thing to say. The whole point of using median prices, as opposed to average prices, is that the median--as it depends only on the ranking of prices, not the actual prices themselves--is not easily affected by a few extreme values.

Perhaps he is computing the median prices on a very small sample, such as the sales of homes in a local district within a short period of time. If that is the case, if he only has a handful of prices, then the problem is not which statistic but the sample size.

In addition, the median price is a one-number summary of a list of individual prices. So I must quibble with the use of the plural in this sentence: "median home sales prices in Suffolk County are up more than 31 percent..." If the reporter gets how the median is computed, he would have chosen the singular verb.

 

 

Comments

[jonathan]

Not disagreeing - there's obvious 6th grade level confusion about average versus median - but I think the point was lost on the reporter that you can see in Boston area stats how the mix of properties sold drives the median, especially if you're mentally referencing a submarket smaller than "Suffolk County". This isn't a big place. I live in Brookline, an affluent Boston submarket, and you can see how the mix of properties sold shifts from small condos to larger places and homes. Because the numbers of properties sold isn't that large, the median can move around, even going down as psf goes up, though obviously not as dramatically as average. This has become less true in the last year as prices have risen a lot (so the mix may understate (or overstate) the rise). I'm making an obvious point about the median, but I think the reporter mistranslated what his interviewee meant, particularly about the market and submarkets. PSF is a better gauge here, as in other high cost urban markets, because you can isolate that to location and condition, plus you can extract from it a profit measure, meaning the amount of profit available in buying, "renovating" and selling. I put "renovating" in quotes because it is currently cheaper here to gut than to actually renovate, in part because actual renovation skills are in shorter supply and thus are driven to higher cost psf projects. Even so-called "historic" rehabs, even in actual historic districts, are complete gut jobs.

[Ken]

He makes it clear in the next sentence what is meant "What’s required to get a clearer picture, he says, is something like the S&P/Case-Shiller Home Price Indices, a repeat-sales calculation of existing homes." So it is nothing to do with the behaviour of the median but rather comparing like to like.

[David Walker]

These days it is surely difficult to analyse statistic or any kind of real estate data even by experts because the market is fluctuating so fast especially in this sector.

Aspen Homes