Discussion paper No. 45
Perspectives on the High Price of
Japanese Land
by
Peter Boone
June, 1989
Economic Research Institute
Economic Planning Agency
Tokyo, Japan
This
research
authors
were
Research
Institute,
working
ing Agency in 1988.
here
are
was
authors'
the
done
for
while
Economic
Economic
Plann-
The views expressed
and
do
not
represent
those of the Economic Planning Agency.
Perspectives on the High Price of Japanese Land
by
Peter Boone
Harvard University
June 1989
This paper is part of ongoing research into the Japanese land problem
and represents a revised version of an earlier paper titled "Perspectives on
the Japanese Land Problem". I am grateful to Masaru Yoshitomi for inviting me
to the EPA this past summer in order to gather information on this issue.
And I wish to thank Mr. Umetani, Mr. Ando, Mr. Furukawa, Mr. Hanazaki, Mr.
Harada and the staff of the Economic Research Institue for their helpful
discussions and hospitality.
At Harvard I have benefitted from conversations
with John Kain, Lawrence Summers, Changyong Rhee and especially Jeffrey
Sachs. Any remaining errors are my own.
Abstract
In this paper we argue that the cause of high land prices in Japan can be
factored into two main sources. The first is a high intensity of land use which,
under a simple rule of thumb, raises the value of land by one percent for each
one percent increase in GNP per unit of arable land. And second, by a low rate
of capitalization on rents. This, we argue is due to expectations of an expected
faster growth of the Japanese economy in comparison to other industrial
countries, and to differences in tax policies.
We argue that many of the popular explanations for the high price of
Japanese land are misdirected.
In particular the observations of excess
concentration in the Tokyo region, inefficiencies in Tokyo land-use, and other
particular aspects of zoning regulations and urban planning are not at the heart
of the problem. This is not to say that these problems do not reduce the quality
of life for those living in the Tokyo region, but rather that in order to explain
high land values and prices throughout Japan we must look to other explanations.
Introduction
The value of land in Japan is more than twice the value of all the land
in the U.S.
problem".
This single fact illustrates what is known in Japan as the "land
With an area which is one twenty-fifth that of the U.S., the average
price of a Japanese unit of land is roughly seventy times the price of American
land.
Economists and the popular press have related the so-called "land problem"
to the most topical policy debates.
The high cost of land is deemed to raise
the Japanese savings rate through the requirement to save for a down payment on
a home.
to
argue
This has led the U.S. Advisory Committee for Trade Negotiations (ACTN)
that
a
solution
to
the
U.S.-
Japanese
trade
imbalance
requires
macroeconomic adjustments within Japan which will reduce the value of land.1
It is also argued that the high price of land has distorted the price earnings
ratios on Japanese stocks and can explain, at least partially, the differences
in price/earnings ratios across countries.
And the high cost of land makes both
housing and food expensive, so that even while Japan has the highest income in
the world when measured in dollars, measures of income which are adjusted for
purchasing power place Japan at the same level of income as the U.K..
The causes of the land problem have been related to the protectionist
policies towards agriculture, with the argument being that the problem of an
1 The
role of ACTN is to provide advice to the President on all aspects of
U.S. trade policy. In their report "Analysis of the U.S.-Japan Trade Problem"
(Feb. 1989) they discuss the role of agricultural policy, through raising rural
land prices, and the high urban land prices in explaining why the Japanese
savings rate is so high (pp.39-40). They suggest that Japan should reform their
"tax and housing systems" (p 42) and implement the structural reforms recommended
by the Maekawa Commission including a change in agricultural policy (p xv).
－1－
already scarce land supply is exacerbated by restrictions on food imports.
And
differences in tax policies are also claimed to be important sources of the
problem.
The list of relevant topics has led many to argue that the land problem
is one of the most important distortions in the Japanese economy today.
Table 1 shows a comparison of the two basic measures of the differences
in land prices and values which we focus on.
The average price of land, which
is the value of land divided by the area of arable land in the country, is shown
for major industrial countries along with the ratio of land value relative to
GNP.
In both cases Japan is an outlier.
In this paper we try to explain these observations.
As there are clearly
a multiplicity of factors which contribute to cross-country differences in land
values, our approach is to try to isolate specific factors which are capable of
explaining these sharp international differences in land prices.2
There has been surprisingly little theoretical work done on this issue.
Iwata and Yoshida(1988) have examined the implications of tax policies upon land
prices in theoretical models.
within and outside Japan.
A number of empirical studies have been conducted
Numerous white papers and reports have been issued
by Japanese government agencies and the OECD discussing the key issues of the
problem but in general these empirical studies have not been done within a
2 The
relationship between high average land prices and the "Japanese land
problem" is more subtle than this table suggests.
For example, the rent on
housing land depends both on expenditures on housing land, and on the
distribution of land between housing and non-housing sectors. A country could
have both a high price of average land, and yet very inexpensive housing land
if the land actually used by the housing sector was sufficiently large.
In
general though, Japanese land prices in all sectors are high relative to other
countries, so our focus upon aggregate land values and average land prices in
this paper simplifies the discussion. As we discuss in our conclusions, it also
provides a useful frame of reference within which to analyze the Japanese land
problem.
－2－
theoretical framework.
Sachs and Boone (1988) pointed out the potential link
between the high price of land and the various issues of market regulations, tax
policy and agricultural trade policy.
simulation
potential
models,
for
a
and
strong
in
an
causal
Here we showed in both multisectoral
illustrative
relationship
theoretical
between
model
the
that
restriction
there
on
is
food
imports, the differential tax treatment of agricultural land versus urban land,
and various causes of market inefficiencies, with a higher level of land prices
across the country.
Ito (1988) extended these theoretical results and showed
in an OLG model with a growing supply of land that the effects of various
policies on the current account depend upon factor intensities in production
sectors.
His paper also includes an interesting discussion of empirical data
showing the negative impact of high land costs on the stock of housing. 3
The nature of our approach is determined by the availability of data and
some theoretical issues.
Ideally we would like to specify a model of land price
determination based upon supply and demand for land services, and then use data
from rents on land, the price of land, and determinants of demand and supply
across countries to econometrically estimate the importance of various potential
explanatory variables.
But such a theoretical model would require a detailed
specification of the determinants of urban spatial structure, and certainly there
is no consensus of opinion in the economic literature on how to model these.
Further, data measuring rents on land, and even the price of land is in general
very sparse.
We were able to obtain some cross-country surveys of rents on
office buildings, and rents on agricultural land.
3A
We also have good measures
high price of land does not neccesarily impact upon welfare. But if the
high price is a reflection of high rents on land, or if there are credit market
imperfections which distort mortgage markets, then welfare would certainly be
affected.
－3－
Table 1
Land Value over GNP, Average Land Prices and GNP per Unit of
Arable Land for Selected Industrialized Countries
(1977)
Country
Australia
Belgium
Canada
Denmark
France
Germany
Italy
Japan (1985)
Norway
South Africa
Sweden
Switzerland
United Kingdom
United States
Land Value
GNP
0.56
0.80
1.02
0.60
0.93
1.08
0.70
2.88
0.78
0.80
0.53
0.88
0.88
0.98
Land Price GNP/Arable Land
(normalized, US=1)
0.5
0.5
1.9
6.7
2.1
70.0
2.0
6.2
3.0
30.0
2.6
1.0
3.0
1.0
Source: Goldsmith (1985), World Development Report (1987), and
New Book of World Rankings (1988). Arable land from OECD(198?)
where available.
Note: Average land price is calculated by multiplying the land
value to GNP ratio by GNP in 1986 and then dividing by
total arable land (as measured in New Book of World Rankings).
－4－
of the price of various types of land across the U.S. and Japan.
very good data on regional land values within Japan.
And we have
But there is not
sufficient, reliable data to carry out a good cross country study of land price
determinants.
Given these difficulties, we have chosen to take a less ambitious approach.
On the theoretical side we separate the issues of the role of urban spatial
structure, from the role of macroeconomic factors such as tax policies and
production structure.
This is outlined in section one, where we isolate a list
of factors which theoretically determine the value of land and we separate them
between those which are "macro-phenomena" and those which relate to urban
factors.
Potential reasons why urban spatial structure and macro factors are
different across countries are outlined in section two, where we discuss a list
of "popular" explanations for the Japanese land problem.
We then proceed by
examining the actual role of these factors by asking whether there exist crosscountry differences in each which might be able to explain the sharp differences
in land prices and values.
In section 3 we provide background information on
the historical pattern of changes in land prices and land values, and on regional
variations of land values within Japan.
In section 4 we begin our examination of the importance of the various
factors outlined in sections one and two which might explain differences in land
values.
Here we compare urban spatial structure in the U.S. and Japan, and we
argue that the general pattern of price dispersion across cities is similar in
the two countries.
We also show that the pattern of urban concentration in
Japan, as measured by broad aggregates of population concentration, is similar
to that in other countries.
In section 5 we examine the "macro-phenomena" which
may be able to explain the differences in land values, and we show that two key
－5－
factors may be differences in expected growth rates and intensity of land usage.
Sections 4 and 5 give a profile of the similarites and differences in the
key determinants of land values across countries, but they do not provide any
evidence concerning the relative empirical importance of these factors.
In
section six we carry out an econometric investigation of the variation in land
prices and values across prefectures in Japan.
estimate
the
empirical
importance
of
This investigation allows us to
differences
expected growth rates and urban spatial factors.
in
production
structure,
And given our previous
discussion of the differences in these variables across countries, we are able
to infer the relative importance of each in explaining differences in land values
across countries using the estimated coefficients from our regressions.
Our
conclusion
is
that,
holding
equal
differences
in
the
rate
of
capitalization on rents, differences in average land prices between Japan and
the industrial countries are due to differences in the relative intensity of land
usage.
We suggest a simple rule of thumb whereby a one percent increase in
density (as measured by GNP per unit of land) results in a one percent increase
in average land prices.
And we argue that the remaining difference in average
land prices and land values relative to GNP may be attributed to differences in
capitalization rates, such that we predict the rate of capitalization applied
to land in Japan is somewhere around one quarter to one half that in other
industrial countries.
The differences in capitalization rates are attributed
to differences in expected rates of growth across countries and to differences
in tax policies.
The view that these two factors explain the differences across countries
observed in Table 1 is supported by the econometric investigation in section 6,
and the finding that a number of puzzling observations outlined in the previous
－6－
sections, such as the observation that rents on buildings across countries vary
much less than land prices while at the same time the cost share of land in the
construction
costs
of
new
buildings
is
extremely
consistent with the arguments which we present.
differences
in
expected
growth
rates
and
high
in
Japan,
are
all
Further, we show that plausible
tax
policies
could
easily
explain
observed variations in capitalization rates, and given this we suggest that it
is more appealing to focus on fundamentals rather than "bubbles", or chronically
inefficient markets to explain the long-run pattern of land values in Japan.
Throughout the paper we deal only with the long-run nature of the Japanese
land problem, and we do not discuss the short-run fluctuations in land prices,
such as the recent surge in Tokyo land prices.
We believe more time is needed
before we could sort out the true cause of this phenomena.
policy issues in this paper.
Neither do we examine
These require a careful study in their own right,
and any policy discussion must be prefaced by an understanding of the causes of
the land problem.
1 The Determinants of Land Prices
In this section we present a model of land price determination which is
used as a frame of reference throughout the text.
As discussed in the
introduction, we focus upon two key facts concerning the relative value of
Japanese land.
First, land value measured relative to GNP in Japan is roughly
three times higher than that in other nations, and second that the average price
of land is many times (in the range of seventy compared to the U.S.) higher than
that in other nations.
We begin by modelling the determinants of land value in
a country with Cobb-Douglas production technologies and then consider how land
－7－
value relative to GNP, and land prices are determined and related.
Finally, we
consider the effect of relaxing various assumptions in the model. 4
Suppose a country produces a number of different goods, say n, using a
Cobb-Douglas technology with land, labor and capital as inputs.
Let F denote
production of the ith good, then output of sector i, Y will be determined as
follows:
Y ＝ F ( K,L,Z) ＝ K Z L
(1)
where L and Z are labor and land inputs respectively.
If p is the price of this
good then a well-known property of Cobb-Douglas production functions is that the
share of the expenditures on land by sector i will be simply the share parameter
for land,
, multiplied by the value of output for this sector.
Indexing these
share parameters by each sector i we may write the aggregate value of rents in
the economy, R, as:
R ＝
(2)
P F
Σ
i
Further, if the pre-tax required return on land is r, then the value of land will
simply equal the present discounted value of future rents.
If we rule out
bubbles on land values then this can be written as:
∫
∞
LV ＝
(3)
t
{Σ
i
P F
4 One
}e
s
criticism sometimes made of this approach is that it does not
explicitly refer to demand variables and preferences. In fact this is not true.
The determination of the equilibrium production structure will depend upon the
nature of preferences. We refer below to urban spatial structure which would
introduce another potential channel for preferences to affect land values.
－8－
Gross national product equals the sum of factor payments to each input, which
in turn is just the sum of output in each sector:
GNP ＝ Σ P F
(4)
i
And thus if we divide (3) by (4) we get land value over GNP defined as:
∫
∞
LV/GN P ＝
(5)
P F / GN P
{Σ
i
t
}e
s
Assuming that each sector has growth in nominal output of rate g +
where g is
is price inflation
real growth in output when deflated by the GNP deflator , and
measured by the GNP deflator then we may substitute this into (5) and (3), and
solving the integral we may calculate land value, land value divided by GNP and
finally the average land price which is land value divided by land area, LD:
(6)
LV ＝ Σ
F /(r- g )
(7)
LV/GN P ＝ Σ
(8)
LV/L D ＝ Σ
i
s /(r- g )
i
i
d /(r- g )
where s
is the period t share of land rents in sector i in GNP, and d
"density"
of
sector
i
where
density
refers
to
the
production divided by total land area in the country.
average
land
prices
are
affected
by
density
of
real
value
of
is the
sectoral
The key point is that
production,
but
under
the
assumptions of Cobb-Douglas technologies, land values relative to GNP will be
independent of the density variable.
The difference in land values, and land values relative to GNP across
countries can be explained through variations in the numerator in (7) and (8)
－9－
or the denominator.
across countries.
The numerator is in both cases reflecting relative rents
In (7) this is the share of rents in GNP.
average rental rate on a unit of land.
In (8) this is the
Thus in order to explain why land values
are on average seventy times higher in Japan than in the U.S., and land values
relative to GNP are three times higher, using the numerator as the cause we would
predict that the share of rents in GNP in Japan is three times higher than in
the U.S., and that average rents are seventy times higher than in the U.S..
Clearly differing intensities of land usage will explain differences in rents
and average land prices, but the differences in rental shares will either be
due to Japan having a greater orientation of production towards sectors with high
land usage (for example the share of agriculture in GNP being relatively high),
or it will be due to the share of rents in the cost of production in identical
sectors being higher in Japan than elsewhere (i.e.
case may be a result of a non Cobb-Douglas technology.
being high).
This latter
In the CES case, when
the elasticity of substitution in the production function is less than one, the
share of land costs in production will tend to rise as the relative price of land
rises.
So countries with a high density will also tend to have higher rental
shares in total output and hence higher land values relative to GNP.
On the
other hand, when the elasticity of substitution is greater than one the opposite
will be true.
When countries have similar production technologies and similar sectoral
compositions of output then the numerator in (7) will be identical across
countries, and variations in land values relative to GNP must be explained by
the denominator.
Throughout the paper we will refer to the denominator, r-g,
as the capitalization rate.
price.
It converts average rents in (8) to the average land
There are a number of reasons why this could vary across countries.
－10－
Differing tax policies, discount rates, or risk premia on assets will affect the
required return on land: r, and varying expected growth rates across countries,
g , will also affect both average land prices and measures of land values
relative to GNP.
The major difference between these stories and the previous
is that rents on land are no longer at the heart of the explanation.
In order
for land value relative to GNP to be high in Japan, it need not be that the share
of rents in GNP be higher than in other countries, but only that required returns
on land are lower, or that there are expectations of greater growth rates in
Japan than elsewhere.
Likewise, the average price of land in Japan could be
higher than elsewhere even while rents are identical.
The model above serves as a useful framework for the macro analysis but
leaves aside the issue of urban structure in explaining the value of land.
Models from the urban economic literature use the same set of formulas to
calculate land values, i.e. the requirement that land values equal the present
value of expected future rents on land, but they are more interested in carefully
specifying the structure of cities in order to explain the variation of land
prices within cities.
This contrasts with our emphasis upon the structure of
production, technology and preferences.
In the former case transport costs,
agglomeration economies, and other externalities give rise to differences in
urban and rural land values.
But again these must be reflected in rents, and
so higher land values in Japan due to variations in such "urban factors" would
generally predict that rents in cities in Japan would be much higher than in
－11－
major foreign cities. 5
As we discuss below, most arguments explaining high land
prices in Japan due to such urban factors have focused upon particular problems
within the Tokyo region.
To properly integrate the role of urban factors within our framework we
would need to model the determinants of city size, the number of cities, and rent
gradients, according to the structure of production, technology and endowments
in the economy.
But such a framework is beyond the scope of our work, and there
certainly is no consensus of opinion in the literature concerning the nature of
these linkages.
As an alternative our approach is to separate the urban issues from the
macro issues.
In section four we examine the gradient of land prices across
cities in the U.S. and Japan with the intention of determining whether there are
major disparities in urban structure across the two countries.
For example, if
the concentration of population, and poor urban planning has somehow driven land
prices in Tokyo far above that in other major cities in the world and Japan, then
the
variation
of
land
prices
across
cities
in
Japan
should
reflect
this
"abnormal" variation in Tokyo, and this should be readily apparent when we
compare variations in land prices across cities in the U.S. with cities in Japan.
In summary, in the models we discussed a high land value relative to GNP
and a high average land price may be simultaneously explained by rents in Japan
being high and/or the capitalization rate on land being low.
5 For
The high rents may
example, the standard urban model relates the rent gradient in a city
to transport costs. Land near to the center of the city carries a higher rent
because the owners need to travel less distance to work and hence save on
transportation costs.
This higher rent is then capitalized into a higher land
price.
－12－
be explained by an orientation of production towards sectors which are land
intensive,
differing
production
costs
technologies
being
relatively
which
high
result
in
in
Japan,
the
or
share
an
of
rents
inefficient
in
urban
organization, all in combination with a high population density in the country.
A low capitalization rate could be the result of tax policies, differing required
returns on assets or differing expected growth rates.
In the next section we discuss the potential arguments in the literature
for the cause of the high land prices, and in the following sections we make a
preliminary attempt to isolate the relative importance of these explanations.
Part 2: Potential Causes of the Land Problem
A number of potential factors which would lead to cross-country differences
in urban structure, production structure, and capitalization rates have been put
forward as potential explanations for the land problem.
In this section we
outline these arguments, and in the following sections we examine to what extent
their predictions are verified in the data.
2.1 Taxation
The complex set of tax codes governing the taxation of property, capital
gains and inheritance is often argued to be at the root, or at least potentially
a solution to the high price of land in Japan.
into two categories.
The arguments can be separated
The first is that the tax codes cause distortions in land-
use, and these tend to bias land-use towards agricultural production and to cause
inefficiencies in allocation of land in the urban areas, particularly in regions
where growth would otherwise result in replacement of the existing stock of
－13－
structures with buildings which make more efficient use of the land.
The second
argument is that the tax codes cause the pre-tax required return on land to be
low both relative to other assets within Japan and relative to other countries.
We present a detailed examination of the main issues, and we refer the interested
reader to Iwata and Yoshida(1987) for a theoretical analysis of the effects of
various tax policies.
The distortionary effects of tax policies on land usage are rooted in the
differing treatment of agricultural land and urban land for inheritance tax and
property tax purposes.
The effective property tax on agricultural land, because
agricultural property is assessed at less than ten percent of its true value,
is much lower than that on urban land.
And because the inheritance tax on farm
land is effectively much lower than the inheritance tax on land used for other
purposes, there is an incentive for those receiving farms as a bequest to
maintain them as farms rather than convert them to urban usage.
Each of these
tax loopholes can be maintained by the owner of the land if they actually farm
the land.
This is one explanation for the apparently large amount of farmland
found within the urban region of Tokyo. 6
The second major influence of tax policy is to reduce the required return
6 It
should be noted that there exists an alternative argument in the urban
economic literature for the existence of farmland within fast growing cities.
There it is reasoned that in growing cities the optimal size and type of building
at any particular location will vary over time.
Given the costliness of
rebuilding or renovating there is an incentive for landowners to hold parcels
of land throughout the city for later development. This implies that growth of
cities will be "patchy" with vacant lots in the midst of the urban sprawl.
Fujita and Kashiwadani (1989) apply this to land development in Tokyo and argue
that the existence of farmlands in the city center is actually an efficient
market response to expected future growth. Note how this interlinks with rental
market distortions. One of the reasons it is costly to build in Tokyo is that
if you build a smaller building for rental with the intention of rebuilding
later, you will have great difficulty evicting your tenants at the opportune
moment. This would encourage landowners not to build.
－14－
on land.
This is caused by the low effective inheritance tax rate on land, and
the often claimed low property tax rate.
Both derive from the fact that the
assessed value of land is well below the true market value.
For the inheritance
tax it is claimed that urban land is assessed at roughly seventy percent of its
market value.
Further, a person can reduce their net wealth for tax purposes
to zero by purchasing land through debt financing.
By buying one hundred dollars
worth of land using loans, one can increase their liabilities by one hundred
dollars, and yet only increase their assets according to tax assessors by seventy
dollars.
Thus total assets fall by thirty dollars, and if one purchases enough
land net assets for tax purposes can be effectively reduced to zero.
The valuation of land for property tax purposes is even lower, and is
roughly thirty percent of market value for urban land and less than ten percent
of market value for farm land.
But it is not often noted that such under
assessment is also widespread in other countries.
Table 2 presents a summary of property tax rates in a number of OECD
countries.
As in the case of Japan the rules are very often complex, and this
table must be treated with caution.
The table shows the average rate of taxation
after adjusting for the portion of the tax which is assessed on land versus
capital.
At best they are crude approximations to the marginal tax rate on land
but they clearly show that while Japan does have relatively low tax rates, they
are similar to those found in some other industrialized countries.
2.2 Agricultural and Trade Policy
The history of Japanese agricultural policy represents a fascinating study
of the interaction between political and economic interests.
－15－
The rapid growth
of
industry
after
the
Meiji
restoration
caused
income
disparities
between those working in the rural and urban sectors of the economy.
to
arise
With this
the battle over agricultural policy came to a head in the late nineteenth
century.
The contemporaneous debate was between advocates of greater imports of food
in
order
to
reduce
food
costs
and
those
production under the auspices of food security.
favoring
self-sufficiency
in
food
In this first battle it was the
protectionists who won and from 1903 to 1918 Japan remained more or less selfsufficient in its staple food rice.
In 1918 and 1919 riots broke out throughout the country in response to
sharp rises in the price of rice as a result of the boom in consumer expenditures
during world war 1.
As a result the government reversed the food self-
sufficiency policy and by the 1930' s the self-sufficiency ratio had fallen to
80%.
But at the end of the second world war a policy of self-sufficiency, again
under the auspices of food security was reenacted.
Currently, the most highly
protected items are rice, beef and pork products and vegetables, and the selfsufficiency
ratios
in
these
products
are
nearly
100
percent.
Sachs
and
Boone(1988) argued that this policy may be a key factor in causing a high price
of land.
Under the standard predictions of trade models, Japan should specialize
in goods which are not land intensive.
But since instead she prohibits trade
in these goods (or in the case of housing the goods are non-tradeable) there is
an increased demand for land and hence higher rents and higher land prices than
would be the case under free trade.
In fact the price of rice in Japan is
currently over seven times the world price and this is certainly reflected in
the standard of living.
And measured against the wholesale price index, the real
price of rice has doubled from 1960 to 1985.
－16－
The hypothesis has three parts.
The first is that the higher price of rice
results in a higher rental payment on land.
The second is that arbitrage from
the rural regions to the urban regions results in higher urban rents also.
And
the third is that these high rents are capitalized into a high land price.
The weakest link in the chain is the second, where we argue that urban and
rural land rents are closely linked.
While in theory this is possible, in
preliminary
this
prefectural
empirical
data
we
research
have
found
into
that
urban
issue
land
(see
prices
Boone(1989))
are
affected
using
only
marginally by changes in rural rents, and so we suspect that while agricultural
policy does contribute partially to the rise in Japanese land values it is not
the major source of international differences.
2.3 Fast Growth and Low Discount Rates
Of the G-7 countries, Japan has grown the fastest since the second world
war, and it is anticipated that in the future this trend will continue.
This
would imply that rents on land can be expected to grow in the future and this
in turn, as measured by a low capitalization rate, would drive up the price of
land today.
It is also argued that the discount rate in Japan is lower than that in
other countries.
This argument can be used to explain both why the price-
earnings ratio on stocks in Japan is so high, and why land values are high.
But
a major criticism is that the increasing integration of world capital markets
should be leading to convergence of world interest rates. 7
7 Note
that high expected growth results in a low capitalization rate on land
since it is a non-reproducible asset, but will not in general affect the
capitalization rate on dividends from capital since capital is reproducible and
the return to capital is tied down by world interest rates.
－17－
Table 2
International Land Taxation in the OECD
Country:
Tax Revenue as %
GDP Land Value
Australia
Denmark
France
Germany
Ireland
Japan
Netherlands
New Zealand
Spain
Sweden
Switzerland
United Kingdom
United States
1.42
1.10
0.85
0.40
2.08
1.21
0.58
2.08
0.27
0.43
0.19
3.16
3.41
2.54
0.92
0.45
0.19
1.04
0.39
0.58
1.00
0.18
0.40
0.10
1.80
1.73
Notes:
1. For those countries where property taxes apply to both
capital and land we assumed that one half of the revenue
applied to land. For Japan we assumed 3/4 of the revenue
applied to land reflecting the higher proportion of land
in total assets. For Ireland, Netherlands, New Zealand
and Spain no land value numbers were available. We assumed
land value equals GNP in these countries.
2. Notes on regime in Australia is for New South Wales only.
3. New Zealand tax rate may be applied only to land valuation
depending on municipality, or to land value plus fixed capital.
We have no information concerning a proper adjustment for
this in column B.
Source: OECD(1983b)
Golsmith(1985)
－18－
2.4 Population Density and Intensity of Land-Use
The
most obvious
distinction
between
Japan
and other
industrialized
countries is her high density of land usage as reported in Table 1.
We have
already discussed in Section 1 how this can affect average land prices, and how
the effects upon land values when measured relative to GNP depend upon the nature
of the aggregate production function.
We should further note that average land prices in rural and urban areas
depend not only on the aggregate intensity of land use, but also on the
allocation of land between the two sectors.
In Japan roughly one-quarter of all
arable land is defined as being in urban areas, whereas in the U.S. only five
percent of land is defined as being urban.
This implies that the average land
price in urban regions in Japan will be less than seventy times that in the U.S..
By focusing on the explanation of average land prices we ignore the
distribution of land between urban and rural regions.
For policy analysis it
is important to analyze how this distribution is determined. 8
2.5 Bubbles on Land Prices
Amongst possible explanations for why land prices are so high, a recurrent
theme is that speculative bubbles may be the cause.
8 Some
land prices
little across
all types of
is evidence
land.
We are skeptical as to the
may argue that this redistribution of land, by lowering relative urban
in comparison to average prices, could explain why rents vary so
countries in urban areas. But as we show below, land prices for
land in Japan are well above those in other countries, and there
that cross-country rents vary by less than prices for all types of
－19－
importance of this factor in explaining the long-run trend of high land values.
The main reason for our skepticism is that there are alternative explanations
which are consistent with the observed facts and could just as easily explain
the high land values without resorting to the assumptions underlying models with
bubbles.
Further, while there is little reliable data showing time series for
rents and prices of land, what data there is for agricultural land suggests that
rents and prices have increased at similar rates, and so we could rule out many
of the models with bubbles which predict that the bubble component should grow
relative to the fundamental such that the rent/price ratio will fall.
2.6 Inefficiencies in the Tokyo Region
Tokyo is not considered to be a well planned city.
There is a noticeable
lack of tall skyscrapers and large office buildings in the downtown area.
The
land area has been subdivided into many small lots which are difficult for
developers to amalgamate into substantial lots for building.
find plenty of "pencil buildings" throughout downtown Tokyo.
As a result you
These are buildings
of five to ten stories with an elevator, stairway, and from all appearances no
room left for office space.
The transportation system, while reasonably fast,
is highly congested and roads into Tokyo have expensive tolls and travel can be
slow.
There are numerous causes for these perceived inefficiencies.
These
include market regulations, the fast growth of the city itself, lack of urban
planning, and a poor inner-city transportation system.
We will briefly outline
some of these without claiming to compile a complete list.
Amongst the market regulations the sunshine laws and certain zoning laws
－20－
are considered most important.
The sunshine laws are a complex and highly
detailed set of codes which are aimed at preventing builders from blocking the
sun from reaching city streets and other buildings.
The regulations include
strict guidelines detailing the exact number of hours of sunshine which must meet
the road, neighboring buildings, etc. each day.
The effect is to have large
numbers of buildings with an optimally angled roof which just permits the
regulations
to
be
met.
In
these
areas
the
regulations
have
prevented
construction of more efficient sized buildings.
Municipal zoning also contributes to the problem.
for each building in particular zones are regulated.
The number of stories
There are also regulations
imposed on new buildings and renovations which discourage rebuilding.
Poor urban planning is also to blame for the inefficiencies.
Many of the
roads throughout Tokyo are too narrow to allow passage of fire engines.
The fire
regulations do not permit tall buildings to be built on these roads, so the space
for more efficient building is limited.
The rapid speed of development has also played a role.
As the city has
grown the stock of buildings is argued to become outdated rapidly and so
neighborhoods of small wooden houses which predate the war remain in downtown
Tokyo.
Further,
tenants
rights
are
strong
and
there
are
effectively
rent
controls which prevent rents on leased buildings to follow the path of rents on
new units.
This is argued to lock in tenants to their existing homes and prevent
urban adjustment.
2.7 Summary
The explanations for the high price of Japanese land can be sorted into
－21－
those which focus on urban spatial structure and those which focus on macrophenomena.
The former includes arguments that production and the population in
Japan are excessively concentrated, and that the urban structure of the Tokyo
region is highly inefficient.
All these factors are presumed to lead to high
rents in these urban regions and are presumed further to lead to high average
land prices.
The macro-phenomena can be divided into those explanations which
predict that rents throughout the country are higher in Japan than elsewhere,
e.g. the role of agricultural policy and high population densities, and those
which predict that the capitalization rate on land is low.
These include the
low property tax rates, the favorable treatment of land for inheritance tax
purposes, a high expected growth rate in the economy, and a low discount rate.
3 Historical and Regional Patterns of Land Values
3.1 Historical Conditions
Figure 1 shows the movement in the real price of various categories of land
over three decades.
The price of all types of land has risen consistently over
this time, with commercial and housing land showing the largest rises with
similar time paths.
Figure 2 shows the pattern of land values measured relative
to GNP over the same period.
Both reflect an important fact that land earned
high and persistent returns throughout the rapid growth era.
With the real
capital gains (in terms of the GNP deflator) averaging over 14% per year from
1955 to 1972 it is clear that the growth in prices was not fully anticipated.
－22－
Figure 1
Real Urban and Rural Land Prices
1955 － 1987
Source: National Land Agency
Note: 1955=1, deflated by WPI.
Figure 2
Japanese Land Value Divided by GNP
1955－1986
Source: NIA
－23－
In correspondence with this transition the economy has undergone the period
of structural adjustment.
In 1900, 35 percent of output was in the primary
sector, but by 1985 this share had fallen to under 5 percent with the rest
falling into manufacturing and tertiary sectors.
And in 1900, 15 percent of
total land value was in urban areas whereas by 1985 the urban share reached 90
percent.
The period has been one of change from a rural to urban society.
With the shift towards production of manufactures, apart from the period
just after the second world war, there has been steady migration from all regions
of Japan to the urban areas of Tokyo, Osaka and Nagoya.
By 1985 the region
surrounding Tokyo had a population of 24 million with 22% of the population in
Japan.
How does this compare with structural transformation in other countries
? The movements in land value relative to GNP for the G-5 countries are shown
in Table 3. 9
Here we find an interesting characteristic of almost all other
industrialized countries.
As the economy developed, land values relative to GNP
fell quite dramatically.
Only in the United States do we find that this is not
true, and here the different process of adjustment may be due to the relatively
abundant and growing supply of land.
One explanation for the declining trend
in other countries is the shift of production away from sectors which are land
intensive (agriculture) and towards the manufacturing and service sectors.
Thus
the share of rents in GNP falls over time and movements in land value relative
to GNP reflect this.
Why do we find then that there is an upwards trend in Japan
in the post-war period ?
Part of the reason may be that during the rapid growth
9 The
data provided by Goldsmith for Japan proved to be quite different from
that recorded in the National accounts for the years 1950 to 1970. For that
reason we report the national accounts data here. We cannot judge the accuracy
of the data for other countries.
－24－
Table 3
Historical Pattern of Land Values in G-5 Countries
Japan
France
Germany
U.K.
U.S.
3.5
1.1
1.2
Range:
1800
1850
5.6
4.3
2.2
2.9
2.9
2.3
4.6
2.9
1.9
2.9
1.9
2.2
1.8
0.6
1.4
1.3
0.7
1.4
1.1
1960
1.9
0.6
0.5
0.5
0.6
0.5
1970
2.2
0.8
1977
2.3
0.9
1985
2.9
1875
1990
1925
1950
Source:
1.0
1.1
1.1
1.6
1.5
1.2
0.9
0.7
0.5
0.9
0.6
0.9
0.9
1.0
1.1
Goldsmith (1985)
1950 to 1985 figures from NIA for Japan, and 1985 figure from
Sachs and Boone(1988) for the U.S.
－25－
era land was undervalued, as witnessed by the large capital gains.
Had these
gains been anticipated speculators would have purchased land and driven the price
upwards.
Another cause may be a decling real interest rate which lowered the
required return on land over the period.
More analysis of the time series data
needs to be done.
3.2 Regional Variations in Land Values
Japan is composed of four islands of which the major island is Honshu.
These are in turn divided into forty-seven prefectures which are numbered in
ascending order running from north to south (see map, figure 3).
The most
northern prefecture is the island of Hokkaido and the most southern is Okinawa.
Many of the prefectures are named after their capital city, for example Tokyo
is centered in Tokyo prefecture and likewise for Osaka, the second largest city
in Japan.
The third largest city, Yokohama, neighbors Tokyo. Nagoya, the fourth
largest city, is in Aichi prefecture.
The northern and southern prefectures have
the greatest concentration of agricultural production and a mountain range runs
down the center of Japan.
Most of the population is found along the eastern
side of this mountain range with the highest density in the central regions.
Figure 4 shows the concentration of production, as measured by prefectural
GDP per unit of arable land.
Here we observe the concentration of production
in the Tokyo region, which is composed of the four prefectures: Saitama, Chiba,
Tokyo and Kanagawa (labelled S, C, T, and K on the figures respectively).
But how does this regional structure affect land values ?
Figures 5 and
6 show the average price of land and land values relative to GDP for each
－26－
Figure 3
－27－
Figure 4
GNP per Unit of Arable Land by
Prefecture: 1985
(Normalized: Hokkaido = 1)
Prefecture
Source: NIA
Figure 5
Average Land Price by Prefecture
1985
(Normalized: Hokkaido = 1)
Prefecture
Note: Average land value calculated as
land value of prefecture divided by
total arable land. Source: NIA
－28－
Figure 6
Land Value Divided by GNP
by Prefecture － 1985
Prefecture
Source: NIA
Figure 7
Average Price of Residential and
Commercial Land by Prefecture
(Normalized: Hokkaido = 1)
Prefecture
Source: Ministry of Construction
－29－
prefecture.
Again the land value shares correlate highly with densities, and
we find that nearly 40% of land value is found in the regions around Tokyo
(compared with roughly 25% of GNP and 8% of arable land).
But there is much less
variation in regional land values when normalized by GNP.
This points to two
observations which will be supported further in the following sections.
Firstly,
that the phenomenon of a high measure of land values relative to GNP in Japan
is not due to factors specific to Tokyo.
The rural prefectures also have
measures which exceed those in other industrialized countries.
Second, that the
sharp variations in density of production have only minor effects upon land
values measured relative to GNP.
In the first section we showed that this would
be the case when the production function is reasonably well approximated by a
Cobb-Douglas production function, and in empirical work in section six we shall
further examine this.
4 Japan vs. the U.S.: Rent and Price Gradients
In section two we discussed how urban phenomena which are, as claimed in
the popular press, deemed to be unique to Japan could be responsible for high
land prices.
With 40% of land values in the Tokyo area it is claimed that excess
concentration and other urban inefficiencies have distorted land prices here and
caused average land values in Japan to be higher than in other countries.
The argument that Tokyo is somehow inefficient while New York or London
are efficient has a number of testable implications.
What we do in this
section is compare the gradient of land prices in the U.S. with that in Japan
under the assumption that if it is truly urban concentration and high land costs
in Tokyo that is the key factor, then we should find much sharper relative price
－30－
movements across regions in Japan than we do in the U.S..
The major finding is
that variations in land prices across regions in Japan are if anything slightly
smaller than variations in comparable land prices across regions in the U.S.
Combining this with the observation that land values relative to GNP in most
prefectures in Japan are well above those in other countries, the evidence
clearly points to macro-phenomena being at the heart of the Japanese land
problem.
But before doing this, we first ask whether there is any strong
evidence of excess urban concentration in Japan.
4.1 Is Japan Excessively Concentrated ?
Table
4
gives
a
cross
country
comparison
of
one
measure
of
urban
concentration: the proportion of population in cities over 500,000 and the
proportion of the population which is in the largest metropolitan area (where
metropolitan areas are defined broadly to include urban sprawl).
Given the
arguments that Japan is excessively concentrated it is surprising to find that
using these measures she is really in the middle of the pack.
Part of the reason
is that her population is larger than these european countries, so with a roughly
equal percentage of persons in the Tokyo region we find a greater level of
population concentration than in London.
On the other hand New York has this
same level of concentration (the figures here are for the New York/New Jersey
sprawl).
With only a small sample of countries there is no clear definition of
what is normal.
The population density within cities does appear to be higher in major
metropolitan areas in Japan than in the U.S..
estimates of population densities.
Table 5 shows some crude
The problem with these is that the definition
－31－
Table 4
Measures of Urban Concentration
in Selected OECD Countries
Percentage of Total Population:
in Cities over 500,000
in Largest City
Country:
Belgium
Canada
France
Germany
Italy
Japan
Netherlands
United Kingdom
United States
1980
1980
18
24
27
38
36
33
18
50
60
10
18
15
12
17
7
18
14
Source: World Bank Report (1984)
Note: Largest city is defined broadly to include urban sprawl
not officially included in city populations.
Calculated
using measures of percentage of urban population in total
population in 1982 multiplied by percentage of urban
population in cities over 500,000 and percentage of
urban population in largets city.
－32－
Table 5
Population per Sq. km in Various Japanese and U.S.
Metropolitan Areas
New York
Boston
Chicago
Honolulu
Phil
Detroit
Wash
San Fran
Cinn
Houston
Dallas
LA
SMSA
City
796
524
490
466
371
303
299
265
221
172
165
126
8306
4191
4550
1476
4166
2762
NA
7545
1635
1041
1044
2412
D.I.D
Tokyo:
Tokyo
Kanagawa
Chiba
Saitama
Osaka
Aichi
8881
11531
7643
6948
7359
9876
6057
Note:1. SMSA is Standard Metropolitan Statistical Area.
City: Designated city area according to municipal divisions.
DID: Densely Inhabited District
2. City is generally a much smaller region than the SMSA, and is
confined to the most dense region of the SMSA. It is unclear
whether DID is more closely related to city or SMSA, but it
reasonable to presume that it is between the two definitons.
Source: U.S. : County and City Data Book: 1988 and State and Metropolitan
Area Data Book: 1986. Both published by U.S. Dept. of Commerce,
Bureau of Census. Japan: Japan Statistical Yearbook 1984.
－33－
of the urban region, and the definition of "land" are not strictly comparable.
The measured population densities vary greatly with these definitions.
For
example, the SMSA is a broadly definied urban region, while the "city" area is
usally much smaller and centered around the downtown region.
In Japan the urban
regions are defined as the "densely inhabited districts", but it is unclear
whether this would be more closely related to the "SMSA" or "city" definiton for
the U.S..
We can only suggest that the density of land use in Japanese cities
is greater than in the U.S., but the exact difference is unclear.
With personal
income per capita, when converted at 125 Yen/dollar, being roughly equal in these
regions, we suggest that the density in Tokyo, meaured by income per unit of
land, is at most ten times that in New York.
4.2 Land Price Gradients in Japan and the U.S.
Appendix 1 describes some urban land price models, and examines what we
should expect to find when a country consists of numerous cities with spatial
advantages and arbitrage of land prices across these cities.
The prediction is
that across cities, land prices should be arbitraged to the point that it is just
as expensive to produce goods in one city as it is in another.
The second
prediction is that within cities, prices should adjust just enough so that a firm
(or household) is just as well off locating near the center as they are near the
fringe.
The result is that rural land prices throughout the country are
identical, and that around cities there are rent gradients which are like cones
rising to the center.
The relative price of land for each city at the center
of the cone represents the extent to which the particular spatial advantages of
that city are greater than other cities.
The key point is that these arbitrage
－34－
conditions place bounds upon movements in prices across cities.
The inter-city bounds on price movements will depend upon the nature of
the locational advantage inherent to cities.
As these advantages are not well
understood it is difficult to confirm the functional nature of the bound.
In
particular there could either be restrictions on the differences in the level
of rents across cities, or restrictions on the proportionate movements in rents.
The conditions under which the latter occurs are described in appendix 1.
Now lets suppose that we are faced with two countries where one country
has a much more inefficient usage of land than another.
The Tokyo argument rests
upon the notion that there is excess urban concentration and inefficient land
use.
Both factors imply a shift upwards in the land price in the center of the
city.
It is also often said that other cities, and in particular Nagoya, are
relatively efficient as compared to Tokyo.
Thus we should expect to see a very
high price of land in Tokyo and a relatively low price of land in Nagoya.
What about relative to other countries ?
If we compare with a country
which has an efficient allocation of land across city centers then we should
expect that as long as spatial advantages are distributed in a similar manner,
the relative movement in prices across city centers in this second country should
be less severe than in the inefficient country.
Thus not only should we find
that Nagoya has less expensive land than Tokyo, but also that the relative
movement of land prices from Tokyo to Nagoya is much larger than the relative
movements in land prices from say New York to Philadelphia.
The last question
is to ask what is the correct measure of a relative change in prices.
We shall proceed using proportionate measures here, where proportionate
is defined as the ratio of the highest to lowest priced land across states within
the U.S. and across prefectures in Japan.
－35－
It is important to note that similar
cross-country proportionate variations in land prices in the face of different
levels
of
differences
land
in
prices
the
may
rate
of
result
from
capitalization
proportionate
of
rents.
shifts
As
in
we
rents,
shall
or
show
proportionate movements in land values across regions in Japan are no larger than
proportionate movements in the United States.
We begin by characterizing the rural price gradient in both the United
States and Japan.
In Table 6 we show the movement in average cropland prices
across states in the U.S. and the movement in average field prices (which does
not include paddy fields which are roughly twice the price of ordinary fields)
across prefectures as calculated by the Japanese Real Estate Institute.
In
general, rural land prices should vary with the quality of land, with property
tax rates and with the distances from major centers where the goods are sold due
to transport costs.
We find that the ratio of the highest to lowest priced land
in the U.S. is 11, while in Japan this ratio is ten.
differs substantially.
But the level of prices
The price of rural land in all prefectures is well above
that in the U.S. and on average is seventeen times the U.S. values.
Of course
the difference in the quality of land may also be substantial, but what is
interesting is the consistent level difference in prices when compared with all
states.
A second set of roughly comparable data is found in the price of home
sites.
In Japan the National land agency defines a certain category of land
which is considered to be in an ideal location for future housing. In the U.S.,
the federal housing association publishes data on the average site price of lots
for homes which they build.
These homes are federally subsidized and usually
locate in relatively inexpensive areas of cities.
The categories are comparable
except for the fact that the U.S. site prices are improved and ready for building
－36－
Table 6
Price per 10 Are of Crop Land
( Yen/$ = 125)
United States
(1985)
Japan
Vermont
Massachusetts
New York
New Jersey
Pennsylvania
Delaware
Maryland
Michigan
Wisconsin
Minnesota
Ohio
Indiana
Illinois
Iowa
Missouri
North Dakota
South Dakota
Nebraska-nonirr
Nebraska-irr
Kansas-nonirr
Kansas-irr
Virginia
North Carolina
Kentucky
Tennessee
South Carolina
Georgia
Alabama
Mississippi
Arkansas
Oklahoma-nonirr
Oklahoma-irr
Texas-nonirr
Texas-irr
U.S. Average
26
84
26
145
64
65
87
34
31
30
50
49
57
45
25
15
13
20
36
17
31
46
77
36
35
29
26
23
29
30
24
22
42
19
41
Hokkaido
Aomori
Iwate
Miyagi
Akita
Yamagata
Fukushima
Nigata
Ibaraki
Tochigi
Gumma
Saitama
Chiba
Tokyo
Kanagawa
Yamanashi
Nagano
Shizuoka
Toyama
Ishikawa
Gifu
Aichi
Mie
Fukui
Shiga
Kyoto
Osaka
Hyogo
Nara
Wakayama
Tottori
Shimane
Okayama
Hiroshima
Yamaguchi
Tokushima
Kagawa
Ehime
Kochi
Fukuoka
Saga
Nagasaki
Kumamoto
Oita
Miyazaki
Kagoshima
Okinawa
Japan Ave
227
785
474
762
576
612
737
489
1092
637
898
1026
940
113629
20044
914
690
742
767
427
794
686
843
920
720
594
39943
881
1109
914
725
471
517
507
377
821
878
844
554
847
492
585
714
467
826
713
948
717
Source: See notes at end of table 17.
Note: Average figures are arithmetic mean. Japanese average
does not include Tokyo, Osaka or Kanagawa.
－37－
while this is not true of the Japanese housing data.
This implies that the
inter-regional differences in land values should be comparable, but the level
difference may be biased closer together.
These prices are shown in Table 7.
Again we find the same key fact. The ratio of the highest to lowest land
prices across Japan is 5, while in the U.S. this ratio is 17 (not including
Hawaii).
Thus the pattern and range of price movements on this land in Japan
does not appear to be out of the ordinary.
And once more we do find that the
level difference in prices is substantial even if it may be biased downward.
The next question is to ask how commercial and downtown residential land
prices vary.
Unfortunately similar data does not exist for each state in the
U.S. but we can get figures on relative changes in prices across cities.
The
price of commercial land is highly variable within all cities, for example in
downtown New York the price varies by large magnitudes depending on location.
But a number of studies do attempt to find comparable indexes and these are shown
in Tables 8 and 9.
The average residential land prices in Tokyo and New York
differ by a factor of over 100 according to Iwata and Yoshida(1988), whom cite
Nomura Research Institution.
To show that the relative price movements across
commercial land in Japan are not substantial, in Table 9 we have taken the price
of downtown commercial land in three cities of comparable relative size for the
U.S and indexed them in the same table with New York equal to one hundred.
Clearly there is no special pattern of relative price movements which is unique
to Tokyo.
Figure 7 shows the price gradient of the average price of commercial and
residential land across prefectures in Japan.
Here we normalize Hokkaido to one.
The price of residential land in Tokyo is less than ten times that of residential
land in the rural prefecture of Hokkaido.
－38－
If on average residential land in New
Table 7
Average Home Site Price Per Square Foot
Site Price Of Housing
(1984)
United States
Hawaii
South Carolina
Nevada
Maryland
Puerto Rico
Louisiana
Minnesota
North Carolina
California
Georgia
New Mexico
Pennsylvania
Montana
Missouri
Utah
Washington
Ohio
Oklahoma
Texas
Nebraska
Illinois
Kentucky
South Dakota
Florida
Tennessee
Mississippi
Alabama
22031
11715
5824
5488
4291
4170
4048
3914
3524
3470
3120
3013
2408
2273
2179
2179
2085
1977
1950
1789
1614
1560
1466
1372
1305
1063
686
Japan
Saitama
Kochi
Osaka
Tokyo
Wakayama
Kanagawa
Gifu
Tokushima
Chiba
Nagano
Ehime
Hyogo
Kyoto
Nara
Kagawa
Aichi
Shizuoka
Gumma
Fukuoka
Hiroshima
Shiga
Shimane
Kagoshima
Yamanashi
Ibaraki
Kumamoto
Ishikawa
Saga
Miyazaki
Tochigi
Yamagata
Okayama
Okinawa
Toyama
Iwate
Tottori
Fukui
Miyagi
Nagasaki
Yamaguchi
Fukushima
Nigata
Akita
Aomori
Oita
Hokkaido
Mie
Source: Federal Housing Authority and
Japan Statistical Yearbook.
－39－
57000
56500
52400
47800
47400
46000
44600
43500
42500
41000
41000
40600
39100
38700
36000
34300
34200
32800
29800
28700
28100
27900
27100
26000
26000
25600
25200
24000
22900
22600
22300
22200
21500
21300
21200
21100
20900
20500
18700
18700
18600
17800
16900
16000
15500
13800
12300
Table 8
International Comparison of Residential Land Prices
Cities
New York
Price
100
3400
14956
508500
London
938
31900
Paris
426
14500
1371
46600
Tokyo
Munich
Source:
Index
Nomura Research Institute: "Land Price and Land System"
cited in Iwata and Yoshida (1988).
－40－
Table 9
Commercial Land Values in the U.S. and Japan
(price per square foot)
Land
Price
Index
Land
Price
New York
$1750
Index
Tokyo
338,000
100
100
Osaka
180,800
53
Chicago
$ 800
45
Nagoya
102,000
30
Philadelphia
$ 450
23
Cities with
Population
over 500,000
77,400
23
Cities with
Population
over 300,000
75,100
22
Other Local
Cities
55,700
16
Note: Figures for Japan are average for prefecture while figures U.S.
are highest priced land in center-city. In central Tokyo, the
highest priced land will be over ten times this.
Source: Japanese data NLA (1986)
U.S. Data: Urban Land Institute (1988).
－41－
York city is one hundredth the price of residential land in Tokyo, then it must
be that residential land in Hokkaido is ten times the price of the same land in
metropolitan New York!
The implication is that the gradient of the price of land across Japan
is very similar to the gradient across the U.S., but the actual price of land
in Japan is a proportionate jump above that in the U.S..
Thus once again we find
evidence that the land problem could not be simply one of excess concentration
and urban inefficiencies in the Tokyo region, but rather an issue which is
economy wide.
4.2 Variations in Rents Across Cities
While there are not a lot of good data on rents in Japan, we present data
from commercial office rents and agricultural rents which agrees with the
findings of other studies (National Land Agency(1986) , and Noguchi(1987) as cited
in Ito(1988)) that the relative movement in land prices across countries is not
reflected fully in the cross-country movements in rents.
Table 10 shows estimates of office rents provided by the OECD for
commercial buildings in downtown areas.
We should not expect office rents to
vary in proportion to the underlying land price.
reflect
capital
costs,
and
so
rental
proportionate shifts in land prices.
actually quite similar.
rates
The reason is that rents also
should
differ
by
less
than
the
But the finding here is that rents are
They vary by a factor of two to three between Tokyo and
New York, and by even less between Tokyo and London.
A recent survey by Colliers International shows the same findings as these
other studies.
Their study estimates the cost of a five year lease on a new
－42－
Table 10
Office Rents in Major Cities
City
Tokyo
New York
London
San Fransisco
Singapore
Hong Kong
Rent per3.3 m 2
1985
1986
41,400
24,000
32,400
12,600
13,200
19,800
53,570
17,300
23,300
9,100
9,500
14,200
Source: "Outline of the FY 1986 White Paper on National Land Use"
－43－
commercial building where the lease starts in May 1989.
in Table 11.
The results are shown
The difference in relative rents across surveys is likely due to
variations in exchange rates, but the qualitative observation remains that rents
do not vary nearly so much as land prices.
Lastly, a survey of apartment rents by Fiabci International is shown in
Table 12. This actually places aprtment rents in Tokyo below London and New
York. The general conclusion has to be that rents are not all that different
across major cities.
Another measure of rental rates and perhaps a more reliable measure is on
land in rural areas.
The Japanese Real Estate Institute has published a series
on the rental cost of paddy fields and fields, and the average price of paddy
fields and fields by prefecture for over thirty years.
The U.S. government
publishes data on the rental rate on cropland by state, and the price of cropland
by state.
By comparing the rental rate relative to the market value of land we
should be able to measure the rate of capitalization of rents in rural areas.
Table 13 shows the rents on cropland by prefecture in Japan and compares
them with rents on cropland in the U.S..
Again there is a level shift of roughly
7 times on average, but the shift is not proportional to prices.
Since there
are no capital costs to this land the finding that the shift is not proportional
must reflect a difference in required returns.
The finding is quite striking.
Table 14 shows this.
The rental return on land in Japan is
lower than that in the U.S. by over 3% on average.
Since taxation can only
explain one percent of this difference, there remains a substantial difference
in returns which cannot be explained by the property tax differential.
The fact that this capitalization rate on agricultural land is so low in
Japan may either be a reflection of a lower discount rate, greater expected
－44－
Table 11
Colliers International Survey of Commercial Office
Space Base Rents and Occupancy Costs 1
Base Rent
Tokyo
London
Hong-Kong
Paris
Sydney
Zurich
Boston
New York
Washington
Philadelphia
Toronto
Los Angeles
Occupancy Cost 2
177
124
92
52
49
48
36
33
36
28
29
22
186
152
102
59
58
50
48
47
47
41
39
31
Note:1. The difference between base rents and occupancy
costs are operating expenses which vary across
countries depending upon the responsibilites for
upkeep, etc. as determined in the lease.
2. Annual occupancy cost per square foot in $US
include es base rent and operating expenses for a
credit worthy client occupying 10,000 square
feet in a prime, downtown location for a period
of five years beginning May 1989.
－45－
Table 12
Rents for New Apartments in Major Cities
City
Tokyo
New York
Paris
Hamburg
Zurich
Brussells
Milan
Oslo
Tel Aviv
Sydney
Auckland
Rent $ per m2
1988
262
350
170
84
158
102
140
102
160
114
39
Source: Fiabci International, Survey September 1988.
－46－
Table 13
Rents per 10 Are on Crop Land
( Yen/$ = 125, 1984)
United States
Japan
Vermont
Massachusetts
New York
New Jersey
Pennsylvania
Delaware
Maryland
Michigan
Wisconsin
Minnesota
Ohio
Indiana
Illinois
Iowa
Missouri
North Dakota
South Dakota
Nebraska-nonirr
Nebraska-irr
Kansas-nonirr
Kansas-irr
Virginia
North Carolina
Kentucky
Tennessee
South Carolina
Georgia
Alabama
Mississippi
Arkansas
Oklahoma-nonirr
Oklahoma-irr
Texas-nonirr
Texas-irr
1.0
1.3
1.3
1.6
1.6
2.5
2.4
1.9
2.0
2.3
2.7
3.5
4.1
3.8
2.1
1.2
1.1
1.7
3.4
1.2
2.3
1.4
1.5
1.9
1.7
1.0
1.1
1.1
1.5
1.9
1.2
1.2
0.8
0.9
Source: See notes at end of table 17.
－47－
Hokkaido
Aomori
Iwate
Miyagi
Akita
Yamagata
Fukushima
Nigata
Ibaraki
Tochigi
Gumma
Saitama
Chiba
Tokyo
Kanagawa
Yamanashi
Nagano
Shizuoka
Toyama
Ishikawa
Gifu
Aichi
Mie
Fukui
Shiga
Kyoto
Osaka
Hyogo
Nara
Wakayama
Tottori
Shimane
Okayama
Hiroshima
Yamaguchi
Tokushima
Kagawa
Ehime
Kochi
Fukuoka
Saga
Nagasaki
Kumamoto
Oita
Miyazaki
Kagoshima
Okinawa
7.3
15
11
17
10
15
14
10
12
8.9
14
14
15
12
14
10
11
8.7
6.6
13
10
11
11
11
9
7.7
12
8.1
11
11
11
10
8.9
9.1
6.8
14
8.4
12
7.5
16
12
11
15
10
13
12
9.1
Table 14
Rents as a Percentage of Crop Land Price
(1985)
United States
Japan
Vermont
Massachusetts
New York
New Jersey
Pennsylvania
Delaware
Maryland
Michigan
Wisconsin
Minnesota
Ohio
Indiana
Illinois
Iowa
Missouri
North Dakota
South Dakota
Nebraska-nonirr
Nebraska-irr
Kansas-nonirr
Kansas-irr
Virginia
North Carolina
Kentucky
Tennessee
South Carolina
Georgia
Alabama
Mississippi
Arkansas
Oklahoma-nonirr
Oklahoma-irr
Texas-nonirr
Texas-irr
4.1
1.6
5
1.1
2.5
3.8
2.7
5.5
6.3
7.8
5.4
7.3
7.2
8.4
8.5
7.6
8.3
8.6
9.6
7.2
7.5
3
2
5.2
4.8
3.5
4.3
4.7
5.2
6.4
4.8
5.4
1.9
4.6
－48－
Hokkaido
Aomori
Iwate
Miyagi
Akita
Yamagata
Fukushima
Nigata
Ibaraki
Tochigi
Gumma
Saitama
Chiba
Tokyo
Kanagawa
Yamanashi
Nagano
Shizuoka
Toyama
Ishikawa
Gifu
Aichi
Mie
Fukui
Shiga
Kyoto
Osaka
Hyogo
Nara
Wakayama
Tottori
Shimane
Okayama
Hiroshima
Yamaguchi
Tokushima
Kagawa
Ehime
Kochi
Fukuoka
Saga
Nagasaki
Kumamoto
Oita
Miyazaki
Kagoshima
Okinawa
3.22
1.91
2.32
2.23
1.74
2.45
1.90
2.04
1.10
1.40
1.56
1.36
1.60
0.01
0.07
1.09
1.59
1.17
0.86
3.04
1.26
1.60
1.30
1.20
1.25
1.30
0.03
0.92
0.99
1.20
1.52
2.12
1.72
1.79
1.80
1.71
0.96
1.42
1.35
1.89
2.44
1.88
2.10
2.14
1.57
1.68
0.96
capital gains, or it may be an effect due to the particular characteristics of
the farms offered for rent. For example, it is argued that tenant farming, or
the size of the farms (due to fixed costs for persons using the land) correlates
with the rental return. But there is no careful study which we are familiar with
which examines this question.
4.4 Summary
In summary, it appears that the price of land throughout all regions and
sectors in Japan is high in absolute terms relative to other countries, and the
gradient of prices across the country has roughly the same pattern as that in
the U.S..
With this we found that rents also appear to be higher in Japan, both
in urban areas and in the rural areas, but we found that these rents did not
appear to move in proportion to the difference in land prices and in fact moved
by much smaller proportions.
This can only lead to the conclusion that rents divided by the price of
land are low relative to the U.S., which is what we observed directly from the
agricultural data.
And with the rents on commercial buildings being only a
fraction above the rents in other major cities, we may also infer that land rents
in
commercial
areas
of
Japan
are
low
relative
to
the
price,
i.e.
the
capitalization rate is low there also.
These findings suggest that differences in urban structure are not at the
heart of the explanation of the land problem, and instead that we should focus
upon macro-phenomena.
In the next sections we take this route.
－49－
5 International Comparison of Land Price and Rent Determinants
In this section we continue the strategy begun in section four.
In section
four we argued that urban structure and the pattern of land prices across regions
in Japan was roughly similar to that in other nations.
The remaining factors
which could cause sharp differences in land values and prices are differences
in
the
structure
capitalization.
of
production,
technology,
densities
or
the
rate
of
We now briefly show that the structure of production across
countries is roughly similar, and that there does not appear to be strong
evidence
of
technology
differences.
On
the
other
hand
growth
rates
are
substantially different.
5.1 Share Coefficients and the Structure of Production
The two goods which are most often considered to be land intensive are
housing and food.
These are the items we examine.
Table 15 shows the share of
expenditures on housing rents in consumption, and the share of agriculture (food
only) in GNP as calculated by the OECD.
As usual, the housing rent figures
should be treated with some skepticism since they often will not properly reflect
owner occupancy costs, and they include both rents on capital and land.
In any case the available data does not show a substantial difference in
total shares of production either for housing or food products.
Further, the
food items represent a more important share of total production for most european
countries than they do for Japan.
The share of land costs in total housing costs is not possible to determine
since we have no way of dividing rents between payments on structures and land
－50－
in a reasonable manner. According to Horioka (1988) the share of land value in
the average purchase price of a home in cities in Japan is 70%.
And the share
in rural areas, according to anecdotal evidence from the Ministry of Construction
is 30%.
In the U.S. the share for city housing is from 20 to 25% land cost.
The fact that these shares vary so substantially, while rental rates on buildings
do not vary, must imply that the differences in cost shares is not reflecting
a difference in the rental costs of land, but rather both a difference in rental
rates on land and a difference in capitalization rates. 10
Hayami and Ruttan (1985) have estimated shares of rents in value added in
agriculture for Japan and the U.S..
They find that land' s share in agricultural
output in Japan is roughly 23% in 1985, while in the U.S. it is near 30%.
Thus
it does not appear that differences in share coefficients for rents in the
agricultural production function could explain the high land values relative to
GNP in Japan.
5.2 Growth Rates
Table 16 shows the rate of growth of population and real GNP in nine
industrial countries.
From 1965 to 1980 Japan had the second highest rate of
growth of population, and from 1980 to 1986 she had the third highest rate of
growth.
Over the entire period her growth rate in real GNP was higher than that
10 The
rental rate on a building equals the implicit rents on the underlying
land plus the rental rate on capital used in the structure. If the high price
of land in Japan reflects only a high implicit rent on land, while the cost of
capital in all countries is similar, then the high share of land costs in the
purchase price of a building should imply that rents on buildings in Japan would
be well above that in other countries. Given that we don' t find this, it must
be that the high price of land is not due solely to a high rental rate on land.
－51－
Table 15
Share of Land Intensive Goods in Production
and Consumption in Nine Industrialized Countries
Country:
Belgium
Canada
Denmark
France
Italy
Japan
Netherlands
United Kingdom
United States
Share of Housing
in Total Private
Consumption
12.5
17.2
18.8
13.5
11.1
14.1
13.0
15.5
15.6
Share of Agriculture
in Nominal GNP
2.1
na
4.4
3.5
4.3
2.2
4.2
1.6
2.1
Definition: Housing is rents paid on housing as computed by each country
in their national accounts. Agriculture is food products not
including forestry or fidheries.
Notes: 1. The housing share for Canada and Belgium were only available
with fuel and power included. The Denmark rents were estimated
using the Netherlands share of rents in total rents plus fuel
and power. The Canadian share of rents was assumed to be .75.
Source: OECD(1987)
－52－
in any of the other countries listed.
From 1965 to 1980 growth in Japan was on
average 1.5% higher than that in the second fastest growing country, Canada, and
3.4% higher than in the U.S. .
that in the U.S..
From 1980 to 1986 growth was 1.3% higher than
This trend appears to be continuing, in 1988 real GNP in Japan
grew by 5.6% while in the U.S. it grew by 3.4%.
6 Regional Variations in Land Prices
So far we have examined the cross-country differences in variables which
potentially can explain the international differences in land prices.
But we
have not quantified the relative importance of these differences in any explicit
manner.
We now turn to an econometric investigation aimed at determining the
relative importance of these various factors.
Ideally we would like to carry
out a cross country investigation, but to date we have been unable to obtain
reliable and sufficient cross-country observations on land values, and righthand side variables, to carry this out.
As an alternative we focus here on the
determinants of regional land prices, and measures of land value relative to GNP,
across prefectures in Japan.
The Economic Planning Agency publishes national
accounts data for each prefecture with measures of land values, population, and
breakdowns of the production structure by prefecture.
This allows us to isolate
the empirical importance of the explanatory factors listed above.
And given our
previous discussion of international differences in these variables, using the
estimated empirical importance we can predict which factors are most likely to
explain the international differences in land prices.
6.1 Production Structure and the Recent Pattern of Growth Across Prefectures
－53－
Table 16
Measures of Income and Population Growth
in Selected OECD Countries
Country:
Belgium
Canada
France
Germany
Italy
Japan
Netherlands
United Kingdom
United States
Average Annual Growth
in GNP
Total Population Growth
1965-1980
1980-1985
1965-1980 1980-1985
3.8
4.8
4.3
3.4
3.8
6.3
3.9
2.2
2.9
0.7
2.4
1.1
1.3
0.8
3.8
0.7
2.0
2.5
0.3
1.3
0.7
0.3
0.6
1.2
0.9
0.2
1.0
Source: World Bank Report (1988)
－54－
0.0
1.1
0.5
-0.2
0.3
0.7
0.5
0.1
1.0
As we discussed in section 3, the central prefectures in Japan are in
general the most highly concentrated in terms of population densities and the
density of production per unit of land.
structures
more
oriented
towards
the
manufacturing, then do the other prefectures.
They also tend to have production
secondary
sector,
most
notably
Real growth rates have also been
fastest in the central prefectures. And there has been a trend towards growth
in prefectures surrounding the most concentrated regions.
6.2 Explaining the Variations in Land Values relative to GNP
The model derived in section 2 serves as the basis for the analysis in this
subsection.
There we argued that the value of land is a function of the
structure of production, technology and demand through their influence upon the
share
of
rents
in
total
costs,
the
intensity
capitalization, and urban spatial structure.
of
land
usage,
the
rate
of
In this section we use variables
proxying for these measures to examine correlations between these and measures
of land values relative to GNP.
The equation we estimate is motivated by (7) and is specifed as follows:
(9) LV/GNP
= Σ
i
a *SHARE
+ e*GROWTH
where:
ISLAND:
+_b*ISLAND + c*POPDEN
+ d*POPDEN 2 + f*NEBDEN
+ g*TIME +
dummy equal to one if the prefecture is not
Honshu
POPDEN: population in thousands divided by prefecture' s
arable land area measured in thousand hectares
POPDEN 2 : popden squared
－55－
GROWTH: proxy for expected growth. As described in text
these are (all in percent):
i) population growth rate (POPGR)
ii) real gnp growth rate (GNPGR)
iii) "fitted", a proxy calculated using a
linear forecast rule (FITTED).
NEBDEN: measure of density of production per unit
of land in neighboring prefectures. Calculated as average of "popden" in neighboring
prefectures.
SHARE :
percentage share of sector i in prefecture' s GNP
TIME :
time dummy equal to one in year t
for all prefectures
i = primary, secondary, tertiary.
j = 1, . . , 47 prefectures
The effect of the structure of production will be captured by the sectoral
breakdown of output by prefecture into three sectors. Since the three variables
sum to one each period, the coefficients will reflect the relative importance
of each sector in determining total land values.
Since the share of rents in
costs of production is presumably highest in the agricultural sector, we would
expect the coefficient on the primary sector to be highest amongst the three.
The population density variables are aimed at capturing the effects of
changes in density, and hence average land prices, upon the share of land in
total costs.
If the elasticity of substitution in the production function is
less than one, then as we discussed above the share of rents in costs will tend
to rise with density, and hence land values relative to GNP will tend to rise.
This would imply that the coefficient on this variable would be positive.
It
may also capture urban spatial factors which we are not able to control for.
－56－
The square of the population density term is included to determine whether there
is a non-linear relationship between population density and the dependent
variable.
Amongst those who argue that excess urban concentration is at the
heart of the Japanese land problem, it is sometimes argued that by moving
individuals out of the most dense prefectures average land values would fall.
This would be true in our model if the coefficient on the squared term is
positive, if it is zero or negative then, ceteris paribus, smoothing out the
population would result in aggregate land values staying the same or actually
rising.
As
the
required
return
on
land
is
presumably
similar
across
all
prefectures, variations in the capitalization rate on land will be due solely
to variations in expected rates of growth.
this variable.
We consider a number of proxies for
The first two are simply the current rate of population growth
and the current growth rate of GNP.
As it turns out there is a high degree of
serial correlation by prefecture in the measured population growth variable but
not in the GNP variable.
The lack of serial correlation in GNP growth may be
due to a greater importance of transitory shocks in this variable.
As an
alternative we have regressed GNP growth by prefecture over the sample on righthand side variables including measures of sectoral production, past growth rates,
and the right hand side variables used in (9).
We then constructed a "fitted"
growth rate (which we label: FITTED).11
11 If
we interpret this as an exact measure of the optimal forecast of growth
rates than t-statistics would not need to be adjusted. We do not adjust the tstatistic as the high significance of population growth in other equations and
the high value of the unadjusted t-statistic clearly point to the explanatory
power.
－57－
Finally, the role of urban structure is captured in the two remaining
variables.
The first is "island", which is a dummy equal to one when the
prefecture is not located in Honshu.
This should capture spatial advantages of
location on the main island and differences in transportation costs.
The
variable "npopld" is a measure of the density of population in neighboring
prefectures.
When a prefecture is adjacent to highly populated regions then land
values will tend to rise because transport costs for products to this market will
be reduced.
manufactures.
This will be particularly true for agricultural products and some
There will also be anticipation of more rapid growth if urban
concentration continues over time, so this variable may also proxy for expected
growth.
The equation is estimated over the sample period 1977 to 1985 using annual
data for all 47 prefectures.
Estimates using a restricted sub-sample which
exclude the Tokyo region, Aichi and Osaka gave similar results.
We use OLS to
estimate the coefficients and make no adjustment to the variance-covariance
matrix of the error terms.
The results for various specifications are given in
Table 17.
The equations fit reasonably well with the R 2 ranging from .56 to .69.
The signs of the coefficients appear reasonable given our previous discussion.
The sectoral share coefficients in general give the primary sector the highest
share, though services have a coefficient which is near and sometimes equal to
that in the primary sector.
The latter may be due to some services having a
fairly high share of rents in their costs, such as housing and retail trade, but
－58－
Table 17
Regression Results from Equation (9) in Text
DEPENDENT VARIABLE: Land Value divided by GNP
SAMPLE: Annual data 1977-1985, 47 prefectures
Coefficient Estimates
(t-statistics in parenthesis)
(1)
(2)
(3)
(4)
POPLD
.017
(0.5)
.074
(2.0)
.007
(0.2)
.237
(6.3)
POPLD 2
-.0035
(0.2)
-.0054
(2.5)
-.0001
(0.4)
-.011
(5.1)
POPGR
.339
(7.9)
RHS Variables:
.564
(12.0)
GNPGR
.017
(1.8)
FITTED
.269
(7.9)
PRIMARY
.026
(3.0)
.018
(2.1)
.022
(2.4)
.027
(2.7)
SECONDARY
-.007
(3.2)
-.027
(7.1)
-.003
(1.2)
-.007
(2.7)
TERTIARY
.020
( 10 . 3 )
.018
(8.9)
.021
(9.9)
.019
(8.1)
ISLAND
-.302
(6.3)
-.356
(7.3)
-.267
(5.2)
-.416
(7.4)
NPOPLD
.218
( 13 . 0 )
.259
( 16 . 8 )
.271
( 16 . 5 )
R2
SSR
.69
52.2
.69
52.1
.64
59.7
.56
73.9
Note: All equations is estimated with time dummies for each
year. These are not reported.
－59－
it may also be due to simultaneity bias. 12
The magnitude of the effect of
production structure upon this variable is estimated to be small.
For example,
a reduction in the agricultural sector' s share by ten percent and an equal
increase in manufacturing correlates with a decrease in land value to GNP of
roughly 0.3.
a
sizable
Over a long history of structural adjustment this could amount to
reduction
in
this
variable,
but
given
the
general
similarity
of
production structures across countries today it is hardly evident that this would
explain the international differences in this variable.
The population density variable in all cases has a positive coefficient
on the level but a negative coefficient on the squared term, suggesting that a
"smoothing" of the population across prefectures will not lower the average land
value relative to GNP.
The population density variable has a range of roughly
0.8 to an average of 6 in the Tokyo region.
But the importance of the variable,
in terms of both significance and the magnitude of the effect, depends upon
whether the term capturing the density of population in neighboring prefectures
is included in the regression.
When it is not, population density explains a
maximum variation of about 0.9 in the dependent variable across prefectures,
while when we include the neighbor term this falls to around 0.2, and in terms
of statistical significance the latter clearly dominates when it is included.
The economic justification of the latter term, perhaps most importantly due to
12 In
an earlier version of this paper we divided production into seven
sectors rather than three.
But here a definite simultaneity problem arose
because the location choice of particular sectors correlates with the price of
the land.
By reducing the number of sectors we hope to capture the major
differences in production technologies without introducing the simultaneity bias.
In any event, the qualitative results are not changed.
－60－
expectations of future growth 13 , along with the observation that farmland in
neighboring prefectures of the concentrated regions is quite expensive relative
to rents, further justifies the inclusion of this variable.
The implication is
that the own population density by prefecture contributes little to measures of
land values relative to GNP.
The island term in each case has the expected sign, and is highly
significant.
We report regressions with three different measures of growth rates across
prefectures.
Both the population growth rate and the "fitted" growth rate
described above are highly significant.
The GNP growth term does not do so well
but this may partially reflect a greater weight on the transitory component of
shocks, as compared to population growth, or measurement error.
proxy
is
the
expenditures
fitted
in
characteristics.
the
variable
region
since
based
it
serves
upon
as
a large
a
set
forecast
of
Our favored
of
relevant
growth
in
prefecture
The explanatory power of the growth variables is high.
For
example, the fitted variable ranges from 2.4 to 6.3(Tokyo) in 1985, so it
explains a variation in the dependent variable of 1.1.
6.3 Explaining the Variations in Average Land Prices
In order to verify some of the findings above, and also to examine the
importance of density in determining land prices we now try to estimate
empirically the variation in land prices across prefectures.
13 As
The equation we
we discussed above growth over the last ten years has been most rapid
in the prefectures surrounding the concentrated regions in Japan.
The
expectation of growth in these regions is reflected in the price of farmland
which is often many times higher than could be justified by rents on the land.
－61－
use is based upon (8) above and is specified in log-linear form as follows: 14
(10) log(LV/LD)
= Σ a *SHARE
+ b*ISLAND + c*log(GNPLD)
i
+ d*log(GNPLD 2 )
+ g*TIME +
+ e*GROWTH
+ f*log(NEBDEN)
where: GNPLD : real gross domestic product per unit of arable
land.
LV/LD
: real value of land divided by arable land area
and remaining variables are as described above.
The average land price is calculated as land value divided by arable land area.
The role of the right-hand side variables is the same as that described above,
and the only difference here is the use of GNP per unit of land rather than
population per unit of land as the RHS variable.
If the coefficient on GNPLD
is restricted to unity, then the dependent variable would be land value over GNP,
as in the previous regression.
From our derivation in (8) above it is clear that
the average price of land should depend upon the expenditures in the region
divided by land area.
The estimated coefficients are shown in Table 18. 15
In every case we
cannot reject the hypothesis that the GNP density variable is equal to one and
14 Note
that this equation and (9) are similar. When the coefficient on
GNPLD is restricted to equal one then equation (10) is a variation of (9) with
the log of LV/GNP on the left-hand side and GNPLD2 on the right hand side instead
of the population density terms.
The similarity of the results shows the
robustness of our findings to differences in specification, and confirms that
the coefficient on GNPLD is insignificantly different from one.
15 The
results were similar when we excluded the prefectures surrounding
Tokyo and Osaka and Aichi. These are not reported.
－62－
Table 18
Regression Results from Equation (9) in Text
DEPENDENT VARIABLE: Natural Log of Average Land Price
SAMPLE: Annual data 1977-1985, 47 prefectures
Coefficient Estimates
(t-statistics in parenthesis)
(1)
(2)
(3)
(4)
log(GNPLD)
1.02
(20.8)
1.08
(20.7)
0.99
(18.5)
1.08
(19.6)
log(GNPLD 2 )
.005
(0.5)
-.020
(1.9)
-.001
(0.1)
-.012
(1.1)
POPGR
.181
(9.1)
RHS Variables:
GNPGR
.459
(1.0)
FITTED
.120
(7.1)
log(FITTED)
.269
(5.1)
PRIMARY
.011
(2.1)
.007
(1.3)
.005
(1.0)
.007
(1.3)
SECONDARY
-.005
(4.1)
-.014
(6.6)
-.002
(1.4)
-.010
(4.8)
TERTIARY
.008
(7.7)
.007
(6.8)
.009
(8.2)
.007
(6.3)
ISLAND
-.170
(7.2)
-.196
(7.9)
-.153
(6.0)
-.181
(7.1)
NPOPLD
.159
(7.2)
.232
( 10. 7 )
.225
(9.8)
.232
( 10. 4 )
R2
SSR
.97
12.2
.96
14.7
.97
13.1
.96
13.8
Note: All equations are estimated with time dummies for each
year. These are not reported.
－63－
the square of the density variable equals zero at the one percent significance
level.
This is again evidence that there is no apparent non-linear relationship
between density and land prices across regions in Japan.
The production terms show similar patterns and do not explain large
variations in prices, on the other hand growth variables again are highly
significant and do have high explanatory power.
In the last column of Table 18
we use the log of the "fitted" variable (we could not take the log of the other
growth variables as they had negative observations) so the coefficient may be
interpreted as an elasticity.
This implies that a doubling of this proxy for
expected growth in the prefecture results in a 27 percent increase in the average
price of land.
6.4 Implications
The finding that land prices move roughly one to one with GNP per unit of
land, and the lack of any strong non-linear relationship with respect to density,
implies that the Cobb-Douglas model given in section 1 serves as a reasonable
approximation to the determinants of regional variations in land values.
This
implies a simple rule: a one percent increase in density, as measured by GNP per
capita, results in a one percent increase in average land prices.
But the lack
of any significant effect of density measures upon land values measured relative
to GNP suggests that density alone will not explain the differences in this
measure across countries.
Neither can the role of production structure explain the cross country
differences in measures of land values relative to GNP.
The industrialized
countries all have relatively similar compositions of sectoral output, and we
－64－
found that only for large shifts in production structure, such as over time
during the early stages of industrialization, do we find that production might
matter in affecting land values relative to GNP.
The variables which did explain the variations in land values relative to
GNP
were
factors
which
captured
spatial
structure
(island
and
neighbor' s
population density) and factors which capture future growth expectations (the
growth proxy and neighbor' s population density).
As we have shown that urban
spatial structure does not appear to vary considerably across countries, and that
the phenomena of high measures of land values relative to GNP was consistent
throughout the country, we argue that the growth term will be a key determinant
of the variations in this measure across countries.
The finding that expected growth plays a key role in affecting land values
is consistent with a number of the puzzles we presented above.
The finding
that the rent price ratio is low in Japan, and the observation that land costs
make up a large share of the cost of a building while rents on buildings in Japan
are not a lot different from those in other countries, are both consistent with
capitalization rates being relatively low in Japan. 16
Further, to the extent
that expected growth in rents is a function of past growth rates, it is not
surprising that Japanese investors are willing to pay a high price for land.
The substantial returns throughout the historical periods would suggest high
returns in the future.
The comparison of the price gradient across regions in Japan and the U.S.
leads to a more subtle argument in favor of the capitalization rate playing an
16 According
to Colliers International, in Hong-Kong, another fast growing
country, it is also true that the cost share of land in new buildings (until
recently perhaps) is roughly 80 percent of the price of total construction costs.
－65－
important role in explaining the international variations in land prices and
values.
In standard urban economic models the choice of locations of firms
across cities depends upon the differences in relative costs of production.
One
source of these differences in costs is the difference in transport costs, say
to move goods to a port.
If there are vary large differences in rents on land,
then they will outweigh transport costs and firms will chose to locate in the
low-rent cities.
In our comparison of urban structure across Japan and the U.S.
we found that the variation in relative land prices across Japan was perhaps
somewhat smaller than that in the U.S.
variations in the level of prices.
But underlying these were large
If these variations in the level of prices
were due solely to differences in rents, then the level differences in rents
across prefectures in Japan would have to be much greater than that in the U.S..
This implies that there would be great savings, when compared with the U.S., for
Japanese firms to move to the regions where land prices are cheaper.
The
observation that firms are not moving to these regions suggests that the level
shift in rents is not a sufficient inducement, and yet this level shift is very
large (e.g, according to Table 8 and Figure 8 the difference between the average
residential land price in Hokkaido and Tokyo was roughly 140 times the average
residential land price in New York, so the difference in rents would be of the
same order under assumptions of identical capitalization rates across regions
and countries.) if the variation in land prices is due solely to variations in
rents.
If we drop the assumption that capitalization rates are similar than this
observation makes sense.
When capitalization rates are different, the level
difference in rents can be much smaller than that implied by the difference in
prices, so large variations in price levels can coincide with little inducement
for firms to relocate.
－66－
With the production structure, technologies, and urban structure being
similar across countries, a crude application of the Cobb-Douglas model above
provides a simple framework within which we may account for differences in land
values.
First, the share of rents in GNP across countries should be similar,
and this implies that capitalization rates will determine the fluctuations in
land values relative to GNP, while differing densities (GNP per unit of land)
will explain the movement in average prices when we hold fixed the capitalization
rates.
Thus the observation that most industrialized countries have land values
relative to GNP equal to one is consistent with similar capitalization rates.
The fact that Japan' s measure is nearer to three implies that her capitalization
rate is, say, one quarter to one half that of other nations.
A difference of this magnitude in capitalization rates is easy to justify
given
our
previous
discussion
growth rates across countries.
of
differences
in
tax
policies
and
historical
The average return on land in the U.S. has
averaged three percent over the last century, though in the last two decades it
has been somewhat higher.
Suppose the required return is five percent plus one
percent for property taxes.
percent,
while
expected
Then if expected growth in the U.S. is three
growth
in
Japan
is
four
percent,
the
rate
of
capitalization on rents in the U.S. will be .03, while if Japan has the same
required return on land but no property tax, the rate of capitalization will be
.01. Thus the capitalization rate in Japan will be one third that in the U.S..
The key point is that small variations in expected growth and the observed tax
parameters can easily explain both the low capitalization rate on rents in Japan
and the high measure of land values relative to GNP.
The remaining difference
in average land prices is allocated to the difference in the average intensity
of land usage across countries.
－67－
Conclusions
This paper has argued that the cause of high land prices in Japan can be
factored into two main sources.
The first is a high density of population which,
under a simple rule of thumb, raises the value of land relative to other
countries by the same proportion as the relative densities (as measured by GNP
per unit of arable land) of the countries being compared.
And second, by a low
rate of capitalization on rents which is due primarily to expectations of future
capital gains on land due to an expected faster growth of the Japanese economy
in comparison to other industrial countries.
This difference is further widened
by tax policy.
Our argument derives from the following, simple accounting relation for
land prices:
Average Land Price =
Land Value
Land
=
Land Value Rent
GNP
.
.
Rent
GNP Land
The average land price is high only if the capitalization rate is high, which
equals land values divided by total rents, if the share of rents in GNP is high,
or if the density of land use is high.
It is clear that the density of land use in Japan is well above that in
any other industrialized nation, and this itself leads to a high average price
of land, but we have argued that density alone cannot explain the observed
differences
in
land
prices
across
countries.
If
densities
were
the
sole
explanation, then rents on land would move in proportion with the price of land
when compared across countries.
But in fact rents appear to vary across
－68－
countries by much less than land prices in both rural and urban areas.
The role of production structure, tastes, and urban spatial factors will
affect the term Rent/GNP.
But from the evidence we have presented, we find no
support for the view that the share of rents in GNP is much higher in Japan than
in other countries.
This evidence included a careful comparison of urban spatial
structure in Japan and the U.S., and an examination of the production structure
in land intensive sectors across countries.
In both cases we found that the
patterns observed in Japan were similar to those found in other nations.
The view that differences in capitalization rates are responsible for the
remaining differences in land prices across countries is supported by various
direct, and indirect arguments.
We have observed directly that capitalization
rates on rural land in Japan are much lower than in the U.S..
And the
observation that rents on buildings in major world cities are similar further
supports this view.
For example, the price of land in Tokyo was estimated to
be 150 times that in New York, and yet there was little difference in rental
rates on apartments and buildings.
Further, the density of land use in Tokyo
was found to be at most ten times that in New York, so from the accounting
relation above it is clear that the capitalization rate must be at the cause of
the large variations in these land prices.
These
conclusions
were
supported
by
an
examination
of
land
price
determinants across prefectures in Japan. We found that fundamental factos suchs
as the structure of production, expected growth rates, density, and urban spatial
structure explained a large proportion of the variance in regional land prices.
And we found that prices varied in proportion with the intensity of land usage,
and they were highly sensitive to expected growth rates.
Extending this to
cross-country data, we have shown that differences in densities of production
－69－
across countries, tax policies, and reasonable beliefs for growth expectations
are fully sufficient to explain the international differences in land prices,
and in measures of land values divided by GNP.
This implies that many of the "popular" explanations for the high price
of Japanese land are misdirected.
In particular the observations of excess
concentration in the Tokyo region, inefficiencies in Tokyo land-use, and other
particular aspects of zoning regulations and urban planning are not at the heart
of the problem.
This is not to say that these problems do not cause land to be
somewhat more expensive in the Tokyo region, but rather that in order to explain
high land values and prices throughout Japan we must look to other explanations.
－70－
References
Adam, M.C. and A. Szafarz (1988) "On the Definition of Speculative Bubbles,"
Centre d' Economie Mathematique et d' Econometire, Discussion Paper No. 8806.
Universite Libre de Bruxelles.
Boone, Peter (1989) "City Economies and Macroeconomic Policy: How Agricultural
Policy affects Land Values", unpublished manuscript. Harvard University.
Dasgupta, P. and G. Heal (1974), "The Optimal Depletion
Resources," Review of Economic Studies (symposium), pp. 3-28
of
Exhaustible
Eaton, Johnathan (1988) "Foreign Owned Land" American Economic Review, vol. 78,
no.1, March 1988 pp. 76-88.
Economic Planning Agency (1988) Annual Report on Prefectural Accounts, Economic
Research institute, Economic Planning Agency. pp.734
Federal Housing Authority (various) FHA Homes.
Fujita, Masahisa and Masuo Kashiwadni (1989), "Testing the Efficiency of Urban
Spatial Growth: A Case Study of Tokyo" Journal of Urban Economics, vol.25, pp.
156-192.
Goldsmith, Raymond W. (1985) Comparative National Balance Sheets, University of
Chicago Press, Chicago. pp. 353
Hayami, Yujiro and Vernon W. Rutan (1985). Agricultural Development: An
International Perspective. John Hopkins University Press. pp.506
Hayashi, Fumio, Ito, Takatoshi, and Joel Slemrod (1988) "Housing Finance
Imperfections, Taxation, and Private Saving: A comparative Simulation Analysis
of the U.S. and Japan." Journal of the Japanese and International Economies, Vol
2, no. 3.
Horioka, Charles Yuji (1988) "Tenure and Housing Demand," Journal of Urban
Economics vol. 24 no.3
Ito, Takatoshi (1988) "Japan' s Structural Adjustment: the Land/Housing Problem
and External Balances." unpublished manuscript.
Iwata, K. and Atsushi Yoshida (1988) "Housing, Land and Taxation System in Japan"
manuscript presented at the Fourth EPA International Symposium.
Economic
Research Institute, Economic Planning Agency, Tokyo, Japan.
Lincoln, Edward J. (1988) Japan: Facing Economic Maturity.
Institution. Washington D.C. pp. 298.
The Brookings
Manning, Christopher A. (1988) "The Determinants of Intercity Home Building Site
Price Differences". Land Economics. vol. 64 No. 1 pp. 1-14
－71－
McCallum, Bennett T. (1987) "The Optimal Inflation Rate in Overlapping Generation
with Land." New Approaches to Monetary Economics (eds) William Barnett and
Kenneth J. Singleton. New York. Cambridge University Press. pp. 325-339
Mills, Edwin S. (1984) Urban Economics 3rd edition. Scott, Firesman and Company.
Illinois, pp. 420
National Land Agency (1986) "Outline of the FY 1986 White Paper on National Land
Use." Foreign Press Center Tokyo, Japan. May.
Noguchi, Yukio. (1987) "Land Price Swollen by Bubbles" (in Japanese as cited by
Ito (1988)). Tokyo Kezai Modern Economics, no. 77, November. pp. 38-45
OECD(198?) Unpublished Survey of Japanese Land Use provided by Ministry of
Construction, Tokyo.
OECD (1983a) Issues and Challenges for OECD Agriculture in t he 1980' s. Paris
OECD (1983b) Taxes on Immovable Property. Paris.
Ogura, Takekazu (1982) Can Japanese Agriculture Survive ? Agriculture Policy
Research Center, Tokyo. pp. 280
Rhee, Chnagyong (1988) "Dynamic Inefficiency in an Economy with Land" unpublished
manuscript, Harvard University
Sachs, Jeffrey and Peter Boone (1988) "Japanese Structural Adjustment and the
Balance of Payments." Journal of Japanese and International Economics, vol 2.,
no. 3.
Solow, R.H. (1974) "Intergenerational Equity and Exhaustible Resources," Review
of Economic Studies (symposium), pp. 29-45
United States Department of Agriculture (1986) Agricultural Land Values and
Markets Situation and Outlook Report. June.
Urban Land Institute (1988) Market Profiles.
Washington, D.C.
World Bank (1988) World Bank Development Report.
－72－
Appendix 1
Urban Rent and Price Gradients in the Macroeconomy
In this appendix we illustrate how the price gradient of land is determined
across cities in the macroeconomy.
While we do not solve for a full, general
equilibrium model of output and price determination, we do show under what
conditions the price gradient across countries will be similar. 1
The model is
highly stylized, and serves only as an example of the nature of the arbitrage
relationships across cities rather than a complete explanation of their
determinants.
Differences in the price of land acrros cities can depend upon
spatial advantages, transport costs, externalities, returns to scales, public
goods, etc.
Here we consdier the interaction of spatial advantages with
transport costs.
This simplification allows us to illustrate the issues
involved.
Consider a country with m urban regions where each region has a production
function for the single composite good defined as:
f = a z k l (A1)
i = 1, . . . , m
The parameter a is a technology shift parameter which measures the relative
productivity of a particular region.
It is meant to capture differences in
urban amenities, public goods for production, and other spatial advantages
such as the availability of port facilities.
price takers.
Firms maximize profits and are
Based on cost minimization they chose which city to locate in,
the distance from the center of the city in which they do locate, and the
inputs used in production given these choices.
1I
wish to thank Professor John Kain for pointing this issue out to me.
－73－
As with the standard urban model we assume costs of production involve the
usual factor payments to inputs, and we assume there are transport costs which
rise as a firm locates farther from the center of the city.
costs are proportional to output.
PR
(A2)
= ( 1 - t c ) pf -
The transport
Thus the profits of the firm are:
z - wl - rk
where PR is profits of the firm, p is the price of the firm' s output,
is the
rental rate on land, t is the distance from the center of the city, c is the
transport cost per unit of output, and w, l , r, k are wages, labor input, the
rental cost of capital and the capital stock respectively.
Firms maximize
profits over the choice of location, i.e. t, and the input choices.
The first
order conditions are:
z (A3)
/
t = -cpf p f / z =
p f / l = w
p f / k = r
From these we may determine that the amount of land used, z, at distance t
from the center of the city is as follows:
(A4)
z =
p f /
And substituting this into (B3) we may write:
/
t (1/
)
= -c/
－74－
We then integrate this and solve for the constant term by evaluating the
expression at the urban fringe, where the rental rate on land is determined by
the marginal product in farming.
(A5)
=
e
This gives the urban rent gradient as:
/
is the rental rate at the boundary of the city.
where
Since firms are
profit maximizing the cost of production at every point in the city will be
identical.
Thus for the city i we may derive the cost function as:
c =
(A6)
where
r w 1
is a constant term and
of city i.
is the rental rate for land in the center
Since firms will locate in the city where costs of production are
the least, we will in equilibrium have the costs of production across all
cities where production occurs being identical, i.e. c = c ∀ i,j.
From
this and (B6) above we may derive the following condition for rental rates
across cities:
(A7)
/
= (a / a ) /
Thus the ratio of rental rates across cities will be a fixed constant.
And if
the price of land is simply future discounted rents, then the prices too will
be proportional as long as cities grow at identical rates:
(A8)
P / P = (a / a ) /
－75－
where pl is the price of land.
We may then conclude that if the change in
spatial advantages across cities in different countries is similar, then the
gradient of prices across cities in the different countries should also be
similar when we measure proportionate movements in prices as long as the
countries have an efficient allocation of urban production.
－76－