Two weeks ago, when the Case Shiller index numbers for December were released, I mentioned that the WDC index came in dramatically lower than expected by the Feb ’12 contract prices. I noted then that the WDC index had been revised. (See table to the right.) As this revision impacted the settlement price of the Feb ’12 contract, and since this was the second “meaningful” revision of the WDC in the last 12 months, I thought that the process of an index revision merited its own blog.
Economic data is often revised. Many indices are based on sampling (which may not be precise) or on data collection (which may be revised if additional data is submitted at a later date). Housing indices -particularly repeat sales calculations that rely on recorded prices – fall into the second category.
While economic indices are often revised (e.g. GDP, CPI) those revisions are after the initial release. As such derivative trade (based on those economic indices) can still be settled using the initial index release. However the nature of the moving average in the Case Shiller index may give greater importance to historical index revisions.
(A note of caution here. I am using diagrams (below) merely to illustrate a concept. The underlying numbers, the weights to the indices, and actual index calculations may be very different. The numbers used here are hypothetical numbers used purely for illustration.)
This diagram (below) is an illustration of how a moving average calculation might work. Values (repeat-sales calculations) are complied across multiple time periods (in this case 3 months) and each period’s observations are weighted. Results are released with a two-month lag. Thus the index results for November are announced in January and cover the periods September, October and November. If there are no revisions, the values for the December index includes the same two observations for October and November, drops the September number and adds the results for December. As such, the December number is somewhat already ~2/3rd’s “baked” as it will contain the two of the same components (the Oct and Nov results) as in the November index. This makes predicting (or at least bounding) the December number somewhat easier.
The presumption in forecasting the December number is that the “fixed” components -the October and November results – are just that, fixed. But what happens if additional data comes in that changes the Oct and/or November components?
If we assume (solely for ease of calculation) that all time periods have the same weighting, then the November index (to be released in January) is the average of the three months Sept., Oct., and November or 184.
If one assumes that the -2 point/month trend will remain in place, and that the result for December is 180, then the expected result for the December index will be the average of Oct., Nov. and Dec or 182.
But what happens if the Oct and Nov. numbers are each revised down 3 points? They are revisions of historical results so they shouldn’t impact prior settlements, but they do figure into the results for December. In the example to the right, I’ve assumed that the individual December one-month prediction of 180 was still correct, but now the number reported for December is 180.
As a result of the historical revisions (to Oct and Nov) the December index result is 2 points lower than expected. I believe that this is a unique feature of the Case Shiller index using a moving average calculation.
So what, you might say. Again, historical revisions occur all the time.
My concern is that these “historical” revisions impact the first release of a economic result. If these revisions occur frequently it may impact the market’s ability to narrow bid/ask spreads as contracts near expiration. (That narrowing of the bid/ask spread allows traders with positions an alternative to holding to maturity which exposes them to these revisions.)
The possibility of index revisions may make calendar spread trading riskier (as a revision to the second leg of a trade, after the front leg is released, becomes more of a concern).
Unfortunately, I don’t see a trading solution to this challenge. Front month markets in contracts with frequent index revisions may have to be wider. (In longer term contracts (e.g. Nov ’16) an eventual 1% revision may not matter if a trader is trying to hedge against a possible 10% move.)
The key is to limit revisions by making sure that all of the underlying data is collected by, and reported in, a timely manner. This will minimize future revisions, thereby limiting the uncertainty of, and volatility of, prices in expiring contracts.
Again, this has been a hypothetical illustration of a situation, but of one with real-world implications. A 2-point “adjustment” to a spot index, that is, different that what may have been projected based on past index releases, is worth $500/contract.
As always, please feel free to contact me on the material discussed in this blog, or any other topic related to housing derivatives, at firstname.lastname@example.org.