National Transaction-based Land Price Indices

Subsequent to completing this paper, we discovered that Nichols, Oliner, and Mulhall have a working paper that also examines land price indices in the U.S. While there are similarities between the papers there are some distinct differences as well. For instance, we analyze 20 years of data (1990–2009) while they analyze 15 (1995–2009). We estimate the indices on a quarterly basis while they estimate the indices on a semi-annual basis. We provide a separate index for industrial land use while they combine commercial and industrial land uses into one index. Both papers use CoStar data in the analysis; however, they employ more explanatory variables in the model which we elected not to employ due to the problematic nature of doing so. For instance, they use non arms-length transactions such as foreclosure and eminent domain sales and attempt to control for these unusual conditions of sale by including relevant explanatory variables; whereas, we eliminate these types of transactions from the analysis thus eliminating the need to include these control variables. They also employ variables relating to the condition of the property and its intended use; however, we found the incompleteness of these data across the 20 MSAs to be substantial and therefore did not attempt to control for these characteristics. Compared with our paper, they employ more complex spatial variables, which enhance the rigor of their paper; however, a similar attempt with 20 MSAs, as analyzed in our paper, would be problematic on a quarterly basis. Even though the pricing models in the two papers have distinct differences, the general pattern of the land indices are similar, which is very encouraging. Overall, the richness of the CoStar data allows for considerable investigation of land markets and related research questions.

CoStar Group, Inc. investigates and compiles real estate transaction data in most major markets of the United States. Summaries of the transactions are provided to interested parties on a subscription basis. The authors thank CoStar Group, Inc. for their generous assistance with the data.

After obtaining the land transaction data from CoStar Group, Inc., initial filters were employed to screen the data for possible input errors and to prepare the data for analysis. These filters include a sales price minimum, a land area minimum and maximum, the availability of latitude and longitude, the identification of the geographic area, the maximum distance from the CBD, and whether the land parcel is located in one of the twenty major markets as identified by the S&P/Case-Shiller national housing price index. In addition, only transactions that were considered “arms length” in nature, i.e., no unusual conditions of sale such as foreclosure or eminent domain transactions, were used in the analysis. After the initial filtering, the data were further screened by trimming the top and bottom one percent of all transactions in each of the twenty major markets as indicated by the price per square foot of land area. The above process resulted in 129,782 transactions.

A mean distance of 23.57 miles to the CBD may seem high, however a closer examination of the radius of some of the more growth oriented metro areas, i.e., Los Angeles, Phoenix, Dallas, and Atlanta finds that this distance is reasonable and well within the metro area boundaries. The land transaction data indicate, as expected over the last two decades, that most growth occurred outside the CBD. Because of the national scope of our analysis and considering some very large metro areas, we limited the data to transactions that were less than a reasonable distance from the CBD, to preserve as many transactions as possible. Overall, we find that the major results are robust with respect to choice of maximum distance.

There is a large body of literature on price index construction and the accompanying problems and challenges. See for example Bailey et al. (1963); Munneke and Slade (2000, 2001); Gatzlaff and Ling (1994); Knight et al. (1995); Case and Shiller (1989); Fisher et al. (1994); Fisher et al. (2003).

We explored the possibility of using repeat sales methodology; however, we encountered some unique problems in applying this methodology. For instance, we discovered that parcel characteristics frequently changed between the transactions, i.e., parcels were divided or assembled; therefore, the paucity of identical repeat sales prohibited the use of this index construction technique.

See for example: Colwell and Munneke (1997); Colwell and Sirmans (1978, 1980); Thorsnes and McMillen (1997); Colwell and Munneke (1999, 2009).

The natural logarithm of sales price, as the dependent variable, was also examined for robustness purposes. The index results are virtually identical compared with the natural logarithm of sales price per square foot; however, because the latter specification provided an overall superior fit it was shown in the final regression results.

Comparing our national land indices with those constructed by Nichols et al. (2010) we find some differences. These variations may result from differences in the data samples, in alternative estimation techniques, and in the variables used in the respective models. For instance, we use a more parsimonious model. The indices do however follow the same general pattern, but do peak in different periods. For example our “All Land Use Index” peaks at the end of 2005 while their composite index peaks in 2007. They argue that our conclusion seems less plausible given that many real estate markets were still booming in 2005. The reader is left to determine the accuracy of the respective indices; however, it is reasonable to assume that land prices would lead improved property prices because land is a fundamental or first ingredient in the production or development of improved properties; therefore, we believe that changes in land markets would naturally be the first signal of a pending change in the improved property markets. In all cases, our land indices lead the improved property indices.

The NCREIF TBI is found at http://​mit.​edu/​cre/​research/​credl/​tbi.​html. The website defines the transaction-based index as follows: “The MIT/CRE CREDL Initiative has developed a Transactions-Based Index (TBI) of Institutional Commercial Property Investment Performance. The purpose of this index is to measure market movements and returns on investment based on transaction prices of properties sold from the NCREIF Index database. This is a new type of index that offers advantages for some purposes over the median-price or appraisal-based indexes previously available for commercial real estate in the U.S. Median price indexes are not true price-change indexes because the properties that transact in one period are different from those that transacted in the previous period. Appraisal-based indexes are based on appraisal estimates rather than actual prices of actual transactions.”

As defined on the NCREIF website: www.​ncreif.​com “The NCREIF Property Index is a quarterly time series composite total rate of return measure of investment performance of a very large pool of individual commercial real estate properties acquired in the private market for investment purposes only. All properties in the NPI have been acquired, at least in part, on behalf of tax-exempt institutional investors—the great majority being pension funds. As such, all properties are held in a fiduciary environment.” The NCREIF appraisal-based appreciation index reports the intertemporal changes in values of the properties included in the index.