The Case Shiller Home Price Indices are constant quality house price indices for the United States. A number of Case Shiller home price indices exist: A national home price index, a 20-city composite index, a 10-city composite index, and twenty individual metro area indices. These indices were developed by economists Karl Case and Robert Shiller. (wiki)
The Case Shiller Index has been one of the most keenly watched indices over the past few months. It has been argued (and perhaps understandibly so) that the fortunes of the global economy are closely linked to the behavior of this particular index over the next few months (and years). And given the importance attached to measuring house prices at the present juncture, every infinitesimal movement either way is subject to critical examination by analysts, media and everybody else who might have an interest.
However in their haste to present the latest findings on state of US real estate, chartmakers sometime tend to overlook the importance of presenting the data in the right context. This one was featured in a recent NY Times article which talked about the house prices in many US cities showing signs of stability after been in a decline for the past two years.
Contrary to what you many think at first glance, this is not a graph of the US Home Prices as measured by the Case Shiller Index – it is actually a graph showing the rate of change of US Home Prices as per that Index. While the former is benchmarked to a base of 100 starting in the first quarter of 2000, the later is its first order differential, meaning that it does not show the actual metric but the rate of change of the metric over a period of time. It is easy to miss this point – the graph is titled “Home Prices in Selected Cities, Through May 2009”. I picked up the chart for Phoenix. The first thing that you notice is the big hump in the middle. One can be forgiven for jumping to the conclusion that the peak of that hump is the point where housing prices were at their highest.
If you leave aside incorrect labeling, this chart highlights what is commonly called the base-effect. The base effect manifests itself when you compare each month’s price to the price one year back. It is even more acute on either sides of a turnaround (whether from normal-peak-normal or normal-bottom-normal) where it tends to obfuscate the actual absolute impact of the change.
An example would bring this out:
If you open up this NY Times chart, and concentrate on the April-2006 bar for Phoenix, it shows an increase of +30% from a year earlier. The corresponding figure for previous (Mar-2006) is +34.7%. You would tend to believe that the rate of appreciation has been lower in April than March. However that’s not correct. The actual appreciation in absolute terms in April-2006 over March-206 (month-over-month) was 1.8 times than what took place between March-2006 and Feb-2006.
What essentially happens is that the change in the base offsets the changes in the present figures.
House price Index in the month of Feb-2005: 160.00
House price Index in the month of Mar-2005: 165.96
House price Index in the month of Apr-2005: 173.08
House price Index in the month of Feb-2006: 222.65
House price Index in the month of Mar-2006: 223.23
House price Index in the month of Apr-2006: 225.12
So while in absolute terms, the month-on-month increase in Apr-2006 has been 1.59 points compared to 0.88 in the month of Mar-09 (roughly 1.81 times), at the year-on-year level, that same would mean that the prices in Apr-2006 were 130.07% of the prices an year earlier compared to 134.69% in the month of Mar-09. The problem is that if the chart does not provide a means of ascertaining what those levels were in absolute terms an year back, the reader can begin to make incorrect assumptions. Its not what the chart shows but in what it does not (the absolute value of house price an year ago), that give rise to this problem.
Here is the same graph with three different bases. The first one is house prices compared on a five year-on-year basis followed by the 3-year and 1-year ones.
If you look at the charts (the first three are with the same scale), the shapes are markedly different. especially the point of cross-over. If you look at the month-over-month chart, you can see dips and sudden hikes that are not visible in any of the previous charts.
The last chart which has the actual index numbers, shows that the peak occurred in the month of June-2006. Infact, of all the charts shown above, I would place the greatest emphasis on this one – largely because the reader can ascertain the changes clearly and unambuigously – the numbers and their rate of change (slope) are both clearly visible. If one really had to use anything else, it would be the month-over-month chart to check if there have been any unexpected hikes or drops. The 1 year, 3 year and 5 year charts all suffer from the base effect and until unless you have a compelling reason to use them, would be better left untouched.
You can download the charts for the Phoenix Case Shiller Index here or click on the button below: