Tuesday, May 1, 2007 | It’s simple to track price movements in most financial markets.
One share of Coca-Cola stock, to cite one of endless examples, is priced identically to any other share of stock in that company. An ounce of pure gold from one mine is worth the same as an ounce extracted from a mine on the other side of the world.
In contrast, every home is different. Even two houses with identical floor plans in the same neighborhood may have different views, lot sizes, fixtures, traffic noise, landscaping, Indian burial ground proximity … the list goes on.
This lack of fungiblility, as it is known, makes it difficult to determine the changes to a home’s marketable price. If someone sells a share of Coke stock for $50, that means that for the time being, your Coke shares (and everyone else’s, for that matter) are all worth $50. But if someone sells a house down the street for $500,000, that gives you a far more limited idea either about how much people would pay for your house or about what’s happening with home prices in general.
In short, it’s tough to measure changes in home prices with any accuracy. But that doesn’t stop people from trying.
The most commonly cited home price metric is the median price, which is calculated by putting all the individual home sale prices in order from lowest to highest and choosing the price that’s exactly halfway down the list. As popular as the median is, however, it’s not a very good way to gauge home pricing power.
The median does a fantastic job of telling us how much the typical buyer paid to buy a house. But by and large, people are a lot less interested in transaction size than they are in what it all means to home pricing power at large. This is where the median can fall prey to any of these confounding factors:
Changes in who’s doing the buying
Because the median price is simply the middle value in an ordered list of prices, a change in the number of buyers at either end can move the median even if market prices haven’t changed at all.
Let’s run through a quick example. Suppose that five homes sold for the following prices: $300,000, $400,000, $500,000, $600,000, and $700,000. (Residents of this neighborhood really like round numbers — maybe I should consider moving there). The median sale price of these homes was the middle one: $500,000.
Now let’s suppose that a year goes by. Prices haven’t budged, but this year, only three of the five houses sell: the $500,000, $600,000, and $700,000 ones. This year’s median is $600,000. That 20 percent increase over last year’s median might make it seem like prices have risen, but they actually haven’t budged. The rise in the median price was due not to changes in pricing power but to the fact that fewer low-end homes sold than in the prior year.
That was obviously an extremely simplified and exaggerated example, but the fact is that this phenomenon is occurring — in a subtler fashion, obviously — at this very moment. The recent tightening in the subprime lending market has had a bigger effect on buyers of lower-priced properties than on those of higher-priced properties, and the volume of low-end homes sold has dropped as a result. Thus, the subprime tightening is actually driving the median price higher than it would be otherwise, even though its real effect on home pricing power is surely a negative one.
Changes in what buyers are getting for the money
The median price measures how much the typical buyer paid for a home, but it doesn’t adjust for how nice a home that buyer got in return. So to the extent that buyers are getting more (or less) bang for the buck than they used to, the median price will overstate (or understate) actual home price increases.
This “simplistic and exaggerated example” motif is working for me, so let’s keep at it. This next scenario involves two brothers, Festus and Hrothgar. Each has decided that he is going to spend $500,000 on a house. Festus pulls the trigger right away, but Hrothgar thinks home prices might decline so he decides to wait a year. Hrothgar turns out to be right, and the next year he is able to buy a house that’s bigger and nicer than his brother’s. Market prices went down in this particular example, but you’ll notice that the transaction price didn’t budge. If we were to aggregate the behavior of Festus and Hrothgar across the entire city (which we can do, because we’re still in the hypothetical example), the median price would have stayed constant even though the price of a given home declined.
The above scenario describes a situation in which the median has overstated pricing power, but this phenomenon — like all of the potential distortions discussed here — can cut both ways.
During the boom, as buyers reached the upper limit of what they could spend, they compensated for the lack of affordability by lowering their standards and buying less desirable homes. So for a couple of years, there, changes to the median price actually understated the extent to which individual home prices were increasing. Since the boom ended, the opposite has happened. Now, the extent to which buyers have been able to get more and more bang for their homebuying buck has not been entirely reflected in changes to the median purchase price.
Another unlikely, but hopefully instructive, hypothetical: suppose every homeowner in San Diego put in $50,000 worth of home improvements. All things being equal, the median sale price would probably rise by somewhere around $50,000. This increase in the median would only reflect improvements to the properties, however, not a change in actual pricing power for an unmodified home.
I don’t think this is such a big factor any more, but I can’t say the same for the period in 2004 and 2005 during which huge swaths of the population appeared to be covering every horizontal surface they could find with hardwood and granite.
I don’t need a bizarre imaginary example here because I have one from my own, boring, non-bizarre life. I sold an out-of-state property last year, and as part of the deal, I had to pay about $3,000 to have the pool replastered after the sale closed. A comparison of the recorded sale price with the amount that the buyer actually paid after all was said and done shows that the sale price overstated the actual cost of the home by that $3,000. Such concessions happen all the time these days, and often for quite a bit more than $3,000. Last year, many homebuilders were famously giving cars away to buyers who’d take new condo conversions off their hands. Other builder incentives include things like free upgrades or the waiving of homeowners association, or HOA, fees for the first year, all of which reduce the effective purchase price in a manner that doesn’t show up in the median. Concessions take place in resale transactions as well. Sellers often pay for closing costs, interest rate buydowns, repairs, improvements, or plain old “cash bonuses” to buyers.
Now, think back to the time, circa late-2003, when buyers were lined up around the block to outbid each other for homes. Many buyers even took to writing the sellers sycophantic letters about how they would take the best possible care of the house and would never even consider removing that delightfully vast collection of pan-ethnic garden gnomes in the back yard.
A buyer with the temerity to ask that the seller credit back closing costs or fix a cracked foundation would have been sent packing in favor of a less demanding purchaser. Comparing the concession-laden closing prices prevalent today with the “as-is” prices of the late-stage boom overstates the increase (or understates the decrease) in what buyers are now actually paying for their homes.
As you can see, the median price has some serious shortcomings as a measurement of actual home price changes — not that you’d know it from median’s overwhelming popularity as an analytical tool. Fortunately, there are a couple of alternatives.
One such measure is the median price per square foot. I report here on this metric each month, often referring to it as the “size-adjusted median” (a label that’s only slightly less unwieldy). This figure is calculated by dividing each home’s closing price by its square footage and then taking the median value from the resulting list of price per square foot figures.
The size-adjusted median attempts to reduce the severity of the first two distortions listed above. If fewer low-budget buyers are active, it’s likely that fewer small homes will be sold. And buyers who are getting more for their money are probably getting some of that benefit in the form of increased square footage. Measuring how much the typical buyer is paying for a square foot of home, rather than for the home itself, adjusts for both these phenomena.
The problem is that home size is only a very rough proxy for home quality. It takes no account of location, for starters: a square foot in La Jolla is not the same as a square foot in La Mesa. Even similarly situated homes could be of very different quality and desirability. And adjusting for the size of a sold home doesn’t even begin to address concessions or home improvements (with the exception of improvements that lead to a recorded increase in home size).
Nonetheless, adjusting a home’s sale price by its square footage is a big step in the right direction for measuring not just what was paid, but what was received for that payment.
Yale professor Robert “Irrational Exuberance” Shiller and his colleague, Wellesley professor Karl Case, have taken an even bigger step in the right direction with a house price measurement they recently developed. The Case-Shiller Home Price Index (HPI) measures market price changes based on repeat sales of individual homes. A given home sale price, in other words, is only compared with the price at which that very same home last sold. By gathering enough of these same-home comparisons, the Case-Shiller HPI can model the price movements for an overall market without being affected either by changes in who’s doing the buying or changes in the quality of homes they are getting for the money.
The Case-Shiller HPI even attempts to account for home improvements. If a certain home has changed in price more than other comparable homes, Case and Shiller assume that some of the price change was due to a property-specific factor, not market conditions, and that particular home is given a lower weight in their calculations. The problem with this approach is that if a whole lot of homeowners are making improvements (see late 2004/early 2005), then that will seem like the norm, and owners who aren’t making improvements will be given a lower weighting. The HPI calculations in this case would mistake widespread home improvements for an increase in market-wide pricing power.
The HPI has similar problems with the effect of concessions and the resulting inaccuracy in recorded purchase prices. A home sale with a concession that is way out of line with what’s happening in the market will be given a lower weight. But if a majority of sellers are granting concessions — and this could very well be the case right now — then home sales without concessions will be given a lower weight instead. Like the other home price measures, the HPI probably has a tendency to overstate home prices during a buyers’ market (like today, when concessions are frequent) and to understate home prices during a sellers’ market (like the panic-buying days of 2003-04, when concession demanders were shown the door).
The treatment of improvements and concessions may be problematic, but the HPI’s focus on same-home sales solves some of the problems afflicting the size-adjusted median and even more suffered by the plain-vanilla median. It only measures movements in single-family homes, not condos, but that’s a tolerable tradeoff to get a more accurate price indicator. Unfortunately, the HPI numbers for a given month don’t become available until almost two months later, which is why I’ve stuck to the more timely size-adjusted median for the monthly updates. We need all the analytical help we can get, however, so from here on out I will add the Case-Shiller HPI to the list of items we follow at the Nerd’s Eye View.