Alternative Land‐Price Indexes for Commercial Properties in Tokyo

• Author: Erwin Diewert,Chihiro Shimizu
• DOI: http://doi.org/10.1111/roiw.12443
• Published date: 01 December 2020
• Date: 01 December 2020

© 2019 International Association for Research in Income and Wealth

784

ALTERNATIVE LAND-PRICE INDEXES FOR COMMERCIAL

PROPERTIES IN TOKYO

by Erwin DiEwErt

The University of British Columbia and The University of New South Wales

AND

Chihiro Shimizu*

University of Tokyo and Nihon University

The System of National Accounts (SNA) requires separate estimates for the land and structure com-

ponents of a commercial property. Using transactions data for the sales of office buildings in Tokyo,

a hedonic regression model (the “builder’s model”) was estimated and this model generated an overall

property price index as well as subindexes for the land and structure components of the office buildings.

The builder’s model was also estimated using appraisal data on office building real estate investment

trusts (REITs) for Tokyo. These hedonic regression models also generated estimates for net depreciation

rates, which can be compared. Finally, the Japanese government constructs annual official land prices

for commercial properties based on appraised values. The paper compares these official land prices with

the land prices generated by the hedonic regression models based on transactions data and on REIT

data. The results reveal that commercial property indexes based on appraisal and assessment prices lag

behind the indexes based on transaction prices.

JEL Codes: C2, C23, C43, D12, E31, R21

Keywords: commercial property price indexes, transaction-based indexes, appraisal prices, assessment

prices, land- and structure-price indexes, hedonic regressions

1. introDuCtion

When estimating commercial property price indexes, we are confronted with

the following two problems: how to incorporate quality adjustments in the estima-

tion method and which data source to use in the estimation procedure.

Research studies on commercial property price indexes have emphasized the

problem of data selection when formulating indexes. Traditionally, transaction

prices (also called “market prices” in the literature) have usually been used to esti-

mate price indexes. However, the number of commercial property market trans-

actions is extremely small. Furthermore, even if a sizable number of transaction

Note: The authors thank David Geltner for helpful discussions. The authors gratefully acknowl-

edges the financial support of the SSHRC of Canada and the Nomura Foundation of Japan.

*Correspondence to: Chihiro Shimizu, The University of Tokyo and Nihon University, Kashiwa,

Chiba, 277-8568, Japan (cshimizu@csis.u-tokyo.ac.jp).

Review of Income and Wealth

Series 66, Number 4, December 2020

DOI: 10.1111/j.1475-4991.2019.12443.x

bs_bs_banner

Review of Income and Wealth, Series 66, Number 4, December 2020

785

© 2019 International Association for Research in Income and Wealth

prices can be obtained, the heterogeneity of the properties is so pronounced that

it is difficult to compare like with like, and thus the construction of reliable con-

stant-quality price indexes becomes very difficult.

Under such circumstances, many commercial property price indexes have

been constructed using either appraisal prices from the real estate investment

market or by using assessment prices for property tax purposes. The rationale for

these price indexes is that, since appraisal prices and assessment prices for prop-

erty tax purposes are regularly surveyed for the same commercial property,

indexes based on these surveys hold most characteristics of the property con-

stant,1 thus greatly reducing the heterogeneity problem as well as generating a

wealth of data.

However, while appraisal prices look attractive for the construction of price

indexes, they are somewhat subjective; that is, exactly how are these appraisal

prices constructed? Thus these prices lack the objectivity of market selling prices.

Such considerations have led to the development of various arguments concern-

ing the precision and accuracy of appraisal and assessment prices when used in

measuring price indexes (on these issues, see Shimizu and Nishimura, 2006). In

particular, the literature on this issue has pointed out that an appraisal-based

index will typically lag actual turning points in the real estate market.2 Geltner

etal. (1994) clarified the structure of bias in the National Council of Real Estate

Investment Fiduciaries (NCREIF) Property Index, a representative U.S. index

based on appraisal prices. In a later study, Geltner and Goetzmann (2000) esti-

mated an index using commercial property transaction prices and demonstrated

the magnitude of errors and the degree of smoothing in the NCREIF Property

Index. These problems plague not only the NCREIF Property Index, but all

indexes based on appraisal prices, including the MSCI–Investment Property

Databank (IPD) Index.

With specific reference to Japan’s real estate bubble period, Nishimura and

Shimizu (2003), Shimizu and Nishimura (2006), and Shimizu etal. (2012) esti-

mated hedonic price indexes based on commercial property and indexes based

on residential housing transaction prices, contrasted them with indexes based on

appraisal prices, and statistically laid out their differences. An examination of the

estimated results revealed that during the bubble period, when prices climbed dra-

matically, indexes based on appraisal prices did not catch up with transaction price

increases. Similarly, during the period of falling prices, appraisal-based indexes did

not keep pace with the decline in prices.

Furthermore, in the case of appraisal prices for investment properties, a

systemic factor of appraiser incentives emerges as an additional problem. This

problem differs intrinsically from the lagging and smoothing problems that arise

in appraisal-based methods. Specifically, the incentive problem involves inducing

1Two important characteristics that are not held constant are the age of the structure and the

amount of capital expenditure on the property between the survey dates. Changes in these characteris-

tics are an important determinant of the property price.

2Another problem with appraisal based indexes is that they tend to be smoother than indexes that

are based on market transactions. This can be a problem for real estate investors, since the smoothing

effect will mask the short-term riskiness of real estate investments. However, for statistical agencies,

smoothing short-term fluctuations will probably not be problematic.

Review of Income and Wealth, Series 66, Number 4, December 2020

786

© 2019 International Association for Research in Income and Wealth

higher valuations from appraisers in order to bolster investment performance (on

this point, see Crosby etal. 2010).

In this connection, Bokhari and Geltner (2012) and Geltner and Bokhari

(2018) estimated quality-adjusted price indexes by running a time dummy hedonic

regression using transaction price data. Geltner (1997) also used real estate prices

determined by the stock market in order to examine the smoothing effects of

the use of appraisal prices. Finally, Geltner etal. (2010), Shimizu etal. (2015),

Shimizu (2016), and Diewert and Shimizu (2017) proposed various estimation

methods for commercial property price indexes using real estate investment trust

(REIT) data.

In this paper, we will examine the three alternative data sources suggested in

the literature that enable us to construct land-price indexes for commercial prop-

erties: (i) sales transactions data; (ii) appraisal data for REITs; and (iii) assessed

values of land for property taxation purposes. We will utilize these three sources

of data for commercial properties in Tokyo over 44 quarters covering the period

Q1:2005 to Q4:2015 and compare the resulting land prices.

Section 2 explains our data sources. Sections 3 and 4 use sales transactions data

and a hedonic regression model that allows us to decompose sale prices into land

and structure components. The model of structure depreciation used in Section 3

is a single geometric rate and Section 4 generalizes this model to allow for multiple

geometric rates. Section 5 implements the same hedonic regression model using the

same transactions dataset, but we switch to a piecewise linear depreciation model.

Section 6 compares the alternative depreciation schedules.

It will turn out that the land-price series that are generated using quarterly

transactions data are very volatile and thus they may not be suitable for statistical

agency use. Thus, in Section 7, we look at some alternative methods for smoothing

the raw land-price indexes.

Section 8 estimates a hedonic regression model using quarterly appraisal val-

ues for 41 Tokyo office buildings over the sample period. Since we have panel data

for this application, our hedonic regression model is somewhat different from our

earlier models.

Section 9 estimates quality-adjusted land prices for commercial properties

using tax assessment data. Section 10 compares our land price indexes from the

three sources of data. Section 11 constructs overall property price indexes for Tokyo

commercial properties using the models estimated in the previous sections; that is,

we combine the land-price indexes with a structure-price index to obtain overall

property price indexes. We also estimate a traditional log price time dummy hedonic

regression model and compare the resulting index with our overall indexes. Section

12 concludes.

In summary: there are two main purposes for our paper: (i) using hedonic

regression techniques, we show how overall property price indexes as well as land-

price indexes for commercial office buildings in Tokyo can be constructed using

information on property sales and property appraisal information for REITs, and

we compare the resulting land-price indexes with a land-price index based on prop-

erty tax assessment data; and (ii) we show how the hedonic regressions can be used

to estimate commercial property depreciation rates. Our focus is on decomposing