Follow Insights Blog


CoreLogic Econ


CoreLogic Case-Shiller Indexes: Out With The Old and In With The New

Fewer Foreclosure Resales the Key Difference in New Case-Shiller Indexes

David Stiff    |    Housing Trends, Property Valuation

How much did home prices fall during the housing market crash? According to the new monthly CoreLogic Case-Shiller National Home Price Index, U.S. prices fell by 27 percent. But if measured by the old, quarterly Case-Shiller index, the peak-to-trough decline was 35 percent. So why do two home price indexes, both calculated using the same Case-Shiller methodology, differ so noticeably on the magnitude of the home price crash? The simple answer is because they are not tracking identical populations of housing transactions.

Changes in Single-Family Home Price Indexes, Most Recent Price Crash and Recovery
  Peak-to-Trough Trough-to-March 2014
Old CoreLogic Case-Shiller U.S. Index -34.6% 21.4%
New CoreLogic Case-Shiller U.S. Index -27.4% 20.3%
CoreLogic Combined U.S. HPI -32.4% 24.4%
CoreLogic Combined, Distressed Excluded U.S. HPI -26.5% 20.2%

The new CoreLogic Case-Shiller indexes are estimated using CoreLogic public-deed record data, while the old indexes used data from other vendors. In general, the newly incorporated CoreLogic data provides more detailed and comprehensive information about individual property transactions and broader geographic coverage than the data sources previously had. These differences in turn generate different populations of repeat-sales pairs – the observations that are used to estimate our repeat-sales indexes.

A repeat-sale pair is created by finding two or more arms-length transactions for the same property. Within a particular market, this process is repeated for every property with two or more transactions to generate a set of pairs that populate the index. Non-arms-length transactions, like the transfer of properties between family members or a bank repossession of a foreclosed property, are excluded from the pairs because the recorded price for these transactions often falls below the market value of the property. A property is only included if it has been involved in two or more recorded non-arms transactions, and for most markets, the repeat-sale pair population will include thousands of properties.

The CoreLogic Case-Shiller pairing process does not allow two transactions to be paired if there was an unknown or a non-arms length transaction between the first and second transaction. In general, preventing the spanning of unknown or non-arms length transactions in the CoreLogic Case-Shiller methodology increases the accuracy of an index because pairs that span these transactions provide less-reliable information about property-level price changes. The detail and depth available with the public-deed records that were used to calculate the old quarterly Case-Shiller indexes were, however, often insufficient to prevent pairs from spanning unknown or non-arms length transactions. Now, the more detailed and comprehensive CoreLogic public-record data makes the CoreLogic Case-Shiller “no spanning” rule much more effective so that, in most cases, very few repeat-sale pairs span unknown or non-arms length transactions.

This tightening of the “no spanning” rule now possible with CoreLogic data is the key to the biggest difference between the two Case-Shiller home price indexes. During the housing market crash, thousands of properties were repossessed by mortgage lenders after foreclosure. Because bank repossessions are not arms-length transactions, they are not included in the CoreLogic Case-Shiller pairing process. Furthermore, since they are not arms-length, bank repossessions that occur between a regular sale to a buyer who lost his or her house to foreclosure and a subsequent resale of a bank-owned property will prevent the pairing of the regular sale and the bank sale. Most resales of REO properties are therefore excluded from the new CoreLogic Case-Shiller indexes, whereas most of these transactions were included in the old Case-Shiller pairing process. This means that for the most part, the new CoreLogic Case-Shiller indexes exclude distressed sales. (It’s important to note, however, that both the old and new indexes do include short sales since these are still arms-length transactions.)

CoreLogic public-record data also provides improved coverage for smaller housing markets, most of which had much less foreclosure activity than in larger, more expensive markets and, as a consequence, had smaller peak-to-trough price declines during the housing market crash. We can calculate weighted averages of county-level CoreLogic Case-Shiller indexes to roughly approximate the effect that expanded geographic coverage in the repeat pairs data has on the estimated peak-to-trough decline in the new CoreLogic Case-Shiller national index. In the chart, the line labelled “Existing Counties” is a weighted average of the new CoreLogic Case-Shiller county indexes for counties covered by both the old and new Case-Shiller indexes. The line labelled “New Counties” is a weighted average of counties covered by the new indexes, but not by the old. The weights are Census 2000 county-level estimates of the aggregate value of single-family housing stock.

Case-Shiller County Indexes

The peak-to-trough decline in the new counties weighted average index is 20 percent compared to a 30-percent decline in the existing counties index. This implies that the more moderate decline in the new CoreLogic Case-Shiller U.S. index is partly a consequence of the broader geographic coverage of the CoreLogic public-record data. But this effect is probably relatively small when compared to the differences between the old and new indexes caused by the inclusion/exclusion of REO sales. The aggregate value of housing in the new counties is only one-tenth of the value in the existing counties. This means that differences in the geographic coverage of the repeat pair populations used to estimate the old and new CoreLogic Case-Shiller indexes probably explain about 0.7 to 1.0 percentage points of the 7-percentage point difference in their peak-to-trough changes.

So, which national CoreLogic Case-Shiller index is most accurate? They are both accurately tracking price changes, but for different populations of property transactions – the new CoreLogic Case-Shiller index simply tracks price changes for a slightly larger geographic population of transactions that exclude most REO resales. The magnitude of the peak-to-trough change in the new index is very close to the change in the national CoreLogic single-family combined, distressed-excluded Home Price Index (HPI), which excludes both REO and short sales. Furthermore, the recoveries since the trough in both of these indexes are nearly identical. The old Case-Shiller index tracks price changes for a slightly smaller geographic population of transactions which includes most REO resales. It is most similar to the national CoreLogic single-family combined, distressed-included HPI, which estimates that home prices fell by 32 percent during the housing market crash.

© 2014 CoreLogic, Inc. All rights reserved