“So what are we, chopped liver?” a disgruntled friend asked in reaction to a report indicating that while home prices had risen 17.2 percent over July 2012 for the Twin Cities metro area, they actually dropped for St Anthony Park, where we live.
The answer is no. This apparent discrepancy probably reflects vagaries of economic data, particularly that collected by industry associations, rather than some fundamental trend disadvantaging our neighborhood. As with economic indicators from any source, people should not read more into the report than is justified.
Before getting into that, it is important to note a couple of general points.
First, while higher prices are good for the balance sheets of current owners, no one should expect a return to the go-go days of 2000-2006. Indeed, it would be alarming if this happened, just as loose talk about “Dow 20,000” should provoke wariness rather than anticipation.
Second, for any transaction there is a buyer as well as a seller. Higher home prices are good for owners but bad for those buying their first home. The price bubble early in the century made a lot of us feel rich, but it was hard on young families who saw one prospective home after another pushed beyond their reach.
Third, from a national macro point of view, recovering home prices are positive in that they restore economic confidence. There is a “wealth effect’ from the perception of higher net worth, even if only from a change in the market value of assets. For the large majority of households, a home is many times more important than stocks or bonds. So rising house prices have a much broader effect on confidence than booming Wall Street indexes.
Finally, many asset markets have a tendency to overshoot, to move beyond prices justified by fundamental factors, on the upside as well as the downside. Prices went to unjustified highs eight years ago, and in many areas they dropped to unjustified lows a few years later. A recovery represents a return to pricing based on more sensible and reliable information.
THE PITFALLS
Now think about pitfalls in drawing conclusions from reports of economic indicators. The data in this case came from local associations of real estate agents, based on sales completed in the month of July. The changes reported are in comparison to July a year ago.
I expect these tabulations are accurate in terms of averages for these monthly sales. The danger comes in concluding that a year-over-year change in the average price of two particular sets of sales over any pair of months indicates a fundamental change in value of all the houses in the area in question.
This problem, reaching a conclusion about a “population” as a whole, an entire group of some kind, from measurements on only a part of this group is what statistics is all about.
At the most basic level, early statisticians learned that to reach a reliable conclusion about an entire group from measurements of a subset of that group, two conditions had to be true.
First, the large group or “population” had to be “normally distributed.” In other words, the particular things being measured, whether the height of Austrian soldiers, bushels of wheat per acre or home prices in the Twin Cities, had to fall into a pattern most people know as a “bell-shaped” curve, with a symmetrical distribution on both sides of some central average. If this is not true, drawing reliable inferences is much more dicey and, when possible at all, requires more sophisticated math.
Second, the subset or sample must be randomly drawn from the population. If the cases measured result from self-selection or are driven by some nonrandom factor, one cannot reach trustworthy conclusions about a whole group from measuring a subgroup.
UNRELIABLE SURVEYS
That is why sex surveys done by men’s or women’s magazines from voluntary responses by readers may make for interesting reading but are unreliable in describing how people behave. Ditto for the Iowa political straw poll or the results of party caucuses in predicting general election results.
In the case of home prices, there are questions about whether they actually are normally distributed. More importantly, the sales that actually take place in any given month clearly are not a random sample of all houses in the specified area.
A year ago, many sales still were “distressed,” driven by actual or impending foreclosures.
In such situations, sellers, whether financially pressed families or mortgage lenders, often are willing to accept a lower price than would be accepted by those not under financial pressure.
Moreover, the proportion of sales driven by financial pressures is different for a neighborhood with many blue-collar lower-income households, such as Thomas-Dale, compared with tonier ones like Highland Park or St. Anthony Park, inhabited by doctors, lawyers and professors.
These difficulties in achieving scientifically reliable measurements of home prices have always existed.
They are hard to tackle because homes are not a uniform or “homogeneous” item like the cans of tuna fish or children’s underpants measured monthly in tabulating the Consumer Price Index.
The Case-Shiller index initiated by two academic economists and now carried out by Standard & Poor’s is a good attempt at correcting this. But it measures only 20 metro areas and does not include information on sub-areas within these metropolises.
SAMPLE SIZE
One final aspect of a valid statistical survey is that its reliability varies with the size of the sample relative to the total size of the population.
The greater this fraction, the more sure you can be that the average for the sample is close to the fundamental average for the whole group. But this fraction becomes very small for any given neighborhood, and it may vary greatly from one neighborhood to another.
There simply are not many sales in a neighborhood in one month, and so when one or two high- or low-price sales take place, they can throw off the average.
So the info Realtors tabulate is useful. But don’t read too much into it. And don’t commit the “fallacy of false precision” by assuming decimal points matter.
A change of 17.2 percent for a small group sold in one month probably means the real average for all houses changed by something between 12 percent and 22 percent. But it is impossible to be any more sure than that.