2.3 Main Real-Estate Indices Worldwide
2.3.3 MOODY’S/RCA COMMERCIAL PROPERTY PRICE INDEX
The Moody’s/RCA Commercial Property Price Indices (CPPI) are transaction- based indices that measure property prices at a national level. Indices cover apartment, retail, office, and industrial properties. The RCA family of indices is produced by Real Capital Analytics (RCA), a data vendor tracking commercial real-estate transaction activities and prices in partnership with the MIT Center for Real Estate (MIT/CRE) and the firm Real Estate Analytics LLC (REAL).
The series of price indices and capital returns have been published monthly on a national level since 2000. In addition, there are quarterly indices for the main property types and annual indices for select metropolitan statistical areas (MSAs).7The indices are transaction-based, constructed using a repeat-sales methodology similar to the one employed to produce the Case-Shiller/S&P housing prices indices. The majority of properties included in these indices are properties sold for more than $2.5million. Geltner (2007) describes how property derivatives contingent on the Moodys/RCA Index could be used to hedge CMBS risk.
Figure 2.4 shows that the CPPI exhibits a similar evolution for the level series and for the return series as does the IPD and NCREIF indices. However, the autocorrelation plot shows that the positive short-term serial correlation is followed by some weaker negative correlation on medium-term periods and an even stronger negative correlation for longer-term periods. There is no surprise then that the Ljung-Box Q-tests strongly reject the hypothesis of no autocorrelation at various lags such as 6, 12, 30, 48, and 96 monthly lags.
2.3.4 S&P CASE-SHILLER INDEX
The S&P/Case-Shiller family of Home Price Indices are the product of a part- nership that includes S&P Dow Jones Indices LLC, Core Logic, and Macro- Markets LLC. They were developed using the research of Karl Case and Robert Shiller. This family of home price indices is the most representative for United States, covering 20 major MSAs and widely used as a benchmark for house prices there. The indices are based on the repeat sales pricing method, which
7MSAs are geographical entities that the US Office of Management and Budget specifies and is used for the collection, tabulation, and publication of Federal statistical agencies.
90 100 110 120 130 140 150 160 170 180 190
Moodys/RCA Dec 00 Jun 02 Dec 03 Jun 05 Dec 06 Jun 08 Dec 09 Jun 11 Dec 12 Jun 14
−0.04
−0.03
−0.02
−0.01 0 0.01 0.02 0.03
Returns Dec 00 Jun 02 Dec 03 Jun 05 Dec 06 Jun 08 Dec 09 Jun 11 Dec 12 Jun 14
0 20 40 60
−0.5 0 0.5 1
lag
Sample Autocorrelation
0 20 40 60
−1
−0.5 0 0.5 1
lag
Sample Partial Autocorrelation
Figure 2.4.The Moodys/RCA Index: December 2000 to November 2014.
Notes: The upper graphs show the index levels and percentage returns for monthly series. The lower graphs show the sample autocorrelation and partial autocorrelation functions plots over the same time period.
has been recognized as a reliable methodology to account for housing price movements by the Office of Federal Housing Enterprise Oversight (OFHEO).
A full description of the methodology behind this family of real-estate housing indices is described in S&P Dow Jones Indices (2014).8
The S&P/Case-Shiller Home Price Indices are calculated monthly and the indices for different MSAs are aggregated to form a 10 areas composite
8An historical account of the introduction of this index is given in Shiller (2008a).
0 50 100 150 200 250
Jan-87 Sep-87 May-88 Jan-89 Sep-89 May-90 Jan-91 Sep-91 May-92 Jan-93 Sep-93 May-94 Jan-95 Sep-95 May-96 Jan-97 Sep-97 May-98 Jan-99 Sep-99 May-00 Jan-01 Sep-01 May-02 Jan-03 Sep-03 May-04 Jan-05 Sep-05 May-06 Jan-07 Sep-07 May-08 Jan-09 Sep-09 May-10 Jan-11 Sep-11 May-12 Jan-13 Sep-13 May-14
Composite-10 CSXR Composite-20 SPCS20R National-US CSUS
Figure 2.5. S&P/Case-Shiller Indices monthly between January 1987 and October 2014.
Notes: Historical evolution of S&P/Case-Shiller Indices: Composite 10, Composite 20, and National-US, monthly between January 1987 and October 2014.
index (Composite-10 CSXR) and a 20 areas composite index (Composite- 20 SPCS20R). The S&P/Case-Shiller US National Home Price Index (the US national index National-US CSUS) represents the price of single-family homes in the United States. The historical data goes back to January 1987 for CSXR and National-US CSUS and to January 2000 for the SPCS20R. The three series are illustrated in Figure 2.5. The peak of all three indices occurred during the months of August and September in 2006. Some market observers have taken the moment of downturn in the S&P/Case-Shiller indices as an essential signal to a possible property market downturn.
The S&P/Case-Shiller family of indices measure changes in housing market prices without taking into account changes in the type of houses or in their physical characteristics and leaving out newly constructed houses, condomin- iums, coops/apartments or multi-family dwellings. The indices are calculated monthly based on rolling three-month periods, using a sample of sale pairs, representing the change in price between two consecutive sales of the same single-family home. The quality and size of the house is ensured to be the same.
Sale pairs for the current month and the previous two months are employed in order to compensate for the delay in reporting house sales data. The sale pairs are weighted, by the repeat sales index model, to account for increasing varia- tion in price changes due to transactions occurring over greater time periods.
Case and Shiller (1989) find positive serial correlation as well as inertia in house prices and excess returns, concluding that in the United States the
market for single-family homes is inefficient. These characteristics make the real-estate market unique and the financial economics arguments invoked for product design, pricing, and hedging are bound to be different from other asset classes. The standard way to reduce or eliminate serial correlation effects is to work with the data on a returns scale. Even if serial correlation is not completely eliminated, the time series of returns is hopefully stationary whereas the price level series of real-estate is more trend-following.
As can be seen from Figure 2.6 the Case-Shiller Composite 10 index is highly autocorrelated, exhibiting a degree of predictability with a forecast R-squared at a one-year horizon of about 50%. This forecasting characteristic is attributed
60 80 100 120 140 160 180 200
Case−Shiller National US Index Feb 87 Feb 89 Feb 91 Feb 93 Feb 95 Feb 97 Feb 99 Feb 01 Feb 03 Feb 05 Feb 07 Feb 09 Feb 11 Feb 13
−0.025
−0.02
−0.015
−0.01
−0.005 0 0.005 0.01 0.015 0.02 0.025
Returns Feb 87 Feb 89 Feb 91 Feb 93 Feb 95 Feb 97 Feb 99 Feb 01 Feb 03 Feb 05 Feb 07 Feb 09 Feb 11 Feb 13
0 10 20 30 40
−0.5 0 0.5 1
lag
Sample Autocorrelation
0 10 20 30 40
−1
−0.5 0 0.5 1
lag
Sample Partial Autocorrelation
Figure 2.6.The S&P/Case-Shiller Composite 10 Index between 1987 and October 2014.
Notes: The upper graphs show the index levels and percentage returns for monthly series between 1987 and October 2014. The lower graphs show the sample autocorrelation and partial autocorrelation functions plots over the same time period.
to the profound illiquidity of the market. Furthermore, there seems to be a clear seasonality effect present in the returns series as well as in the autocorrelations, which are positive up to 2.5 years but turn negative afterwards. The visual inspection is confirmed by the Ljung-Box Q test that we conduct at 6, 12, 30, and 48 lags that correspond to the six months, one year, 2.5 years, and four years, which strongly reject the hypothesis of no autocorrelation.
A similar conclusion can be drawn for the S&P/Case-Shiller National US index over the same period. The only observable difference, see Figure 2.7, is that the decline in the National US index that seems to occur around the third quarter of 2007 is not so large as for the Composite 10 index. The S&P/Case- Shiller Composite 20 index has a shorter history starting in January 2000.
Nevertheless, this index has similar characteristics to the other two indices from the same family as can be seen in Figure 2.7. There is high positive auto- correlation on the short term and negative autocorrelation on the longer term.
The S&P/Case-Shiller US national home price index is calculated using the formula:
Ht=
i Hit HidVid
Divisord (2.3)
whereHtis the value of the US national index in periodt,Hitis the value of the home price index for the census divisioniin periodt,Hidis the value of the home price index for the reference perioddandVidis the aggregate value of single-family housing stock in divisioniin the reference periodd. TheDivisord is taken to make sure that the level of the composite index does not change due to changes in the reference period weights. The reference periods are: January 1990, January 2000 and January 2010. The national index is set equal to 100 at its base period in January 2000.
As with the S&P/Case-Shiller Composite-10 index, the Ljung-Box Q-test for the S&P/Case-Shiller Composite-20 index indicate that the hypothesis of no autocorrelation is strongly rejected at monthly 6, 12, 30, and 48 lags. There- fore, any model for the price dynamics of an index from the S&P/Case-Shiller family should be able to produce these empirical characteristics. For example, a geometric Brownian motion for the index levels would not be a suitable model because it does not hold the capacity to generate these kind of autocorrelations.