Tuesday, April 10, 2012

Case Shiller and Teranet HPIs: A Comparison

To gain some perspective I decided to overlay the Case Shiller and Teranet house price indexes for direct comparison. The CS and Teranet HPIs use similar methodologies, tracking same-sales pairs to formulate an aggregate house price index, that is mostly independent of sales mix. The two measures are, for the most part, apples to apples and one of the best gauges available for house prices.

The Teranet and Case Shiller indexes are assigned the value of 100 at a particular date, however housing market cycles between regions and countries can be advanced or delayed based on local and macroeconomic factors. Here I propose rescaling the indexes based on their trough values and, for clarity of graphing, aligning the price peaks as shown below along with a nominal 2%, 3%, and 4% trendlines.

Why would we look at troughs? In this particular case most price troughs occurred in the 1990s, an era of positive real rates and, according to Professor Shiller's long-term analysis, a period where house prices are aligned with their century-long trend. Canada, arguably, due to similarities in its economy to the United States, would have had similar sentiment in its housing market under capital conditions at the time. 

The graph above indicates that speculative excesses can exceed by multiples the trough, therefore using a median or mode measure of the long-run average given there was obviously mis-pricing during that time, would tend to skew any data used as a baseline. A trough, on the other hand, I argue is governed more by rentier-seeking investors and relatively little in the way of speculative activity.

The above graph indicates that while Vancouver (and Montreal too!) is now above any current valuations of presented American markets, it has achieved nowhere near the excesses wrought by said markets. It could be argued that Vancouver's trough is lower than presented (its "trough" presented here occurred in 1998) but even so this would not significantly move its rank to the likes of San Diego, San Francisco, or Los Angeles.

Are Canadian and American markets directly comparable? I would argue, as presented, yes they are, insofar as the investment climate in these locales are roughly similar: if it's true that troughs are governed by baseline investor interest we would expect similar floors and the tax treatments and inherent risks of properties for investors are roughly equivalent. Las Vegas, Phoenix and Miami have oversupply problems and were hit particularly hard, to the point where they are approaching undervaluation; the markets of LA, San Diego, and Seattle -- coastal cities with growing metro populations -- have been slower to fall and would have appreciation higher than inland areas due to increasing land utility.

A note about comparing HPIs. Price levels do not tell us much without knowing their earnings potential. Vancouver has always had a high price-income ratio, in part due to future land utility improvements. Looking at aggregate market prices through HPIs is useful but won't directly tell us much about individual property profitability.

In summary, while Vancouver's appreciation from trough is significantly above inflation, it did not experience anywhere near the amount of price distortions in certain US markets. Vancouver and Montreal are currently the most highly priced metropolitan areas in North America based on that measure and while not achieving the insanity we saw Stateside, now find themselves the leaders.

9 comments:

vreaa said...

jesse -> This looks very interesting, but I'd appreciate some methodology clarification...
You've aligned the peaks, and all the troughs end up being precisely 'x' months prior? That isn't credible.

jesse said...

The troughs are not aligned, the graph does not extend to include the full dataset; what is most interesting to me is the relative valuation peak-to-trough.

jesse said...

If we assume the mid '90s are not the trough and Vancouver was still in a mild bout of speculation, we can add 20-30% to its peak. I would not eliminate that as a plausible scenario.

vreaa said...

You've calibrated each city's 'y' axis as if the last trough was 100, is that correct?

jesse said...

Yes, so a value of 200 is a doubling of prices since trough

jesse said...

Ah I see vreaa, yes the "nominal x%" graphs are not absolutely valid because they would need to be shifted with each individual market. That is Calgary (for example) had a much faster run up to its peak and we should start its "nominal x%" graphs at month -100 instead.

For most markets the starting point for the "nominal x%" is about right, plus or minus. To be more correct comparing nominals, given the assumptions in this post, alignment to troughs is more appropriate. I'll follow this up.

jesse said...

vreaa, see my follow up post http://housing-analysis.blogspot.ca/2012/04/case-shiller-and-teranet-hpis-follow-up.html

I often post interim here; certainly methods designed are always works in progress. That of course won't stop selective usage of graphs and data to suit one's own biases!

Shocked! said...

There is one problem using the CS Vs. the Teranet. They are not apples to apples. The CS uses the year 2000 as its base, however Canada uses 2005. If you standardize thse so both countries use the same baseline, lets say 2000, Canada's situation very similar of not worse in some markets. Download both excel sheets and do the math!

jesse said...

Shocked!, the indexes have been recalibrated to use the trough as the baseline. This is the best that can be done; actually even comparing Case-Shiller data from different cities doesn't tell us much either: if one city was in a trough when the series is marked to 100 while another is 20% through a massive run-up their relative scales won't match.

Aligning to troughs, as I have done here, is probably the best apples-to-apples comparison available.