Metropolitan Land Values∗
David Albouy
Gabriel Ehrlich
University of Illinois and NBER
University of Michigan
Minchul Shin
University of Illinois
March 3, 2017
∗ We
would like to thank Henry Munneke, Nancy Wallace, and participants at seminars at the AREUEA
Annual Meetings (Chicago), Ben-Gurion University, Brown University, the Federal Reserve Bank of New
York, the Housing-Urban-Labor-Macro Conference (Atlanta), Hunter College, the NBER Public Economics
Program Meeting, the New York University Furman Center, the University of British Columbia, the University of California, the University of Connecticut, the the University of Georgia, the University of Illinois,
the University of Michigan, the University of Rochester, the University of Toronto, the Urban Economics
Association Annual Meetings (Denver), and Western Michigan University for their help and advice. We
especially want to thank Morris Davis, Andrew Haughwout, and Matthew Turner for sharing data, or information about data, with us. We thank Nicolas Bottan for outstanding research assistance. The National
Science Foundation (Grant SES-0922340) generously provided financial assistance. Please contact the authors by e-mail at [email protected] or by mail at University of Illinois, Department of Economics, 1407
W. Gregory, 18 David Kinley Hall, Urbana, IL 61801.
Abstract
We estimate the first cross-sectional index of transaction-based land values for every U.S. metropolitan area. The index accounts for geographic selection in location, and incorporates novel shrinkage
methods using a prior belief based on urban economic theory. Land values at the city center increase with city size, as do land-value gradients; both are highly variable across cities. Urban
land values are estimated at 1.5 times GDP in 2006. These estimates are higher and less volatile
than estimates from residual (total - structure) methods. Average values in the five most expensive
metros are 35 times higher than in the five cheapest.
JEL Codes: C43, R1, R3
Keywords: land values, monocentric city, hierarchical modeling
1
Introduction
We estimate the first index of directly-observed market values for land that is cross-sectionally
comparable across U.S. metropolitan areas. Standard economic theory, e.g. Roback (1982),
Brueckner (1983), and Albouy (2016), suggests that this index captures differences in the private
value of household amenities, employment, and building opportunities combined across metro areas. Urban land values have been central to questions of wealth, income, and taxation since the
seminal works of Ricardo (1821) and George (1884).
Unfortunately, market data on land values have been notoriously piecemeal and subject to
numerous measurement challenges. Flow of Funds (FOF) accounts of the Federal Reserve stopped
publishing series for land value in 1995 because of accuracy concerns made plain by negative
values inferred for land. We aim to overcome these challenges using a large national data set
of market transactions of land from the CoStar COMPS database, using an econometric model
informed by urban theory. Our estimates of urban land values prove to be slightly higher and more
stable than values implied by the FOF.
Our indexes of central and average land values appear intuitive and reliable. Both increase with
the city area but are highly variable, providing nuanced support for the monocentric city model of
Alonso (1964), Mills (1967), and Muth (1969). The highest central land values are found in New
York, Washington, Honolulu, San Francisco and Chicago. In these cities, central values are 6.8
times higher than values 10 miles away, although across all cities the unweighted average ratio is
only 1.7. Over their entire urban areas, Honolulu, New York/Jersey City, and San Francisco/San
Jose have the highest average values. Their average values are roughly 35 times higher than those
in the lowest five cities. Values in 2009 averaged $257,000 per acre, down from $427,000 in 2006,
as the total value of urban land fell from $20.8 to $12.5 trillion, or from 1.5 to 0.9 times GDP.
2
Description of Transactions Data and Urban Land Area
Our primary data source is the CoStar COMPS database, which recorded land transaction prices
recorded between 2005 and 2010.1 CoStar provides fields containing the price, lot size, address,
and a “proposed use” for each property. We exclude transactions CoStar has marked as non-arms
1
The CoStar Group claims to have the commercial real estate industry’s largest research organization. The COMPS
database provided by CoStar University, free for academics, includes transaction details for all types of commercial
real estate. We use every sale CoStar considers “land.” Recently, a small literature has used this data for analyses within
metro areas. Haughwout et al. (2008) demonstrate the data’s extensive coverage and construct a land price index for
1999 to 2006 within the New York metro area. Kok et al. (2014) document land sales within the San Francisco Bay
Area, and relate sales prices to topographical, demographic, and regulatory features. Nichols et al. (2013) construct
a panel of land-value indexes for 23 metros from the 1990s to 2009. These indexes are for use over time and are not
comparable across metros.
1
length, without complete information, that feature a structure, are over 60 miles from the city
center, or are less than $100 per acre. The remaining dataset contains 68,756 observed land sales.2
The “cities” we examine correspond to 1999 OMB definitions of Metropolitan Statistical Areas
(MSAs). Some MSAs, known as Consolidated MSAs (CMSAs) are divided into constituent “Primary” MSAs, which we treat separately. In 2000, all MSAs accounted for 80 percent of the U.S.
population. Because MSAs consist of counties, which often contain a large amount of agricultural
land, we consider only land that is part of an urban area by 2000 Census definitions. The main
requirement is that the area consists of contiguous block groups with a population density of over
1,000 per square mile (1.56 per acre), with a total population of over 2,500.
We take city centers to be the City Hall or Mayor’s office of each city. Many MSA names
contain multiple cities, e.g., Minneapolis-St. Paul. We address this by considering each named
city as having its own center. Land parcels within the MSA are assigned to the city center closest
in Euclidean distance. In such cases, reported central values are for averages across centers.
Appendix Figure A.1 displays the geographic pattern of land sales for four CMSAs: New
York, Los Angeles, Chicago, and Houston. The figure shows that land sales are well-dispersed
throughout the metro areas, with sales activity more frequent near city centers.3
3
Econometric Methods
There are two major obstacles to constructing a cross-metropolitan land value index from observed
transactions data. First, observed transactions are not a random sample of all parcels in a city.
Second, we observe few sales in many smaller metro areas, reducing the reliability of the estimates.
Our econometric methods try to overcome both of these obstacles.
3.1
Regression Model of Land Values over Space and Time
Following the monocentric city model, we take each city j as having a fixed center, with coordinates zcj . Land values vary according to a city-specific polynomial in the distance metric,
2
The appendix provides information on the treatment of the data as well as some descriptive statistics.
Land transactions are not randomly distributed over space. Yet, as Haughwout et al. (2008) comment on the New
York data, “ Overall, vacant land transactions occurred throughout the region, with a heavy concentration in the most
densely developed areas ...”. As Nichols et al. (2013) discuss, it is impossible to correct for all types of selection bias
without observing transaction prices for unsold lots, a logical contradiction. Fortunately, the literature has generally
found selection bias to be minor for land and commercial real estate prices. Colwell and Munneke (1997), studying
land prices in Cook County, IL, report, “The estimates with the selection variable and those without are surprisingly
consistent for each land use.” Studying the office market in Phoenix, Munneke and Barrett (2000) find, “the price
indices generated after correcting for sample-selection bias do not appear significantly different from those that do
not consider selectivity bias.” In their construction of metro price indices, Munneke and Barrett (2001) report, “Little
selection bias is found in the estimates.” Finally, Fisher et al. (2007), in their study of commercial real estate properties,
state “sample selection bias does not appear to be an issue with our annual model specification.”
3
2
D(zij , zcj ), between plot i’s coordinates zij and the center, with coefficients δjk determining the
shape of the value-distance gradient.
ln rijt =
2010
X
αjt +
t=2005
K
X
k
δjk D(zij , zcj ) + Xijt β + eijt ,
eijt ∼ i.i.d.N (0, σe2 ),
(1)
k=1
The specification allows for city-center values αjt to vary by year, t, and for controls Xijt , such
as proposed use and lot size.4 Figure 1a shows estimated first-order and fourth-order polynomials
for the Houston MSA, along with the underlying land prices. Both polynomials generally slope
downward with distance, but the fourth-order polynomial allows for a richer distance function.
3.2
Shrinkage Estimation and Its Target
To deal with limited sample sizes we develop a hierarchical model, which “shrinks” estimates
towards national average functions. We allow the shrinkage target to depend on each city’s urban
?
?
is normalized
, where αj2005
area, Aj . We begin by decomposing αjt into two parts, αjt = αj + αjt
0
to zero. Vectorizing δj = [δj1 δj2 · · · δjK ] , we assume a model with with a time-varying prior
?
αjt
∼ N (τt , σt2 ) and cross-sectional priors:
"
αj
δj
#
"
a0 a1
=
d0 d1
#"
1
ln Aj
#
"
+
eα,j
eδ,j
# "
,
eα,j
eδ,j
#
"
∼ i.i.d.N
0
0
# "
Σαα Σαδ
,
Σδα Σδδ
#!
(2)
This technique essentially constructs a “metacity” with a land rent gradient typical of a city with
area Aj . Cities with lower central land values may see their land values dovetail with agricultural
or other non-urban values much closer to the center than cities with higher land values. The model
allows for a full covariance matrix between eα,j and the δjk parameters. When all other parameters
?
are known and αjt
= 0, the best linear unbiased predictor (BLUP) for [αj , δj0 ]0 is a weighted average
between their prior mean and fixed effect estimates,
"
α
ej
δej
#
"
= Wj
a0 a1
d0 d1
#"
1
ln Aj
#
"
+ (I − Wj )
α
bj
δbj
#
.
(3)
where the weighting matrix Wj varies by city and the amount of shrinkage depends on the number
of observations in city j and the relative size of uncertainty in the prior (Σαα , Σδα , Σδδ ) and id
We define D(zij , zcj ) = ln 1 + ||zij − zcj || , using Euclidean distances in miles. Adding one mile to the distance
has two desirable features: it dampens the effect of small changes in distance very close to the city center, and it allows
interpretation of the αjt coefficients as (finite) log land values at the city center. Since the true gradient may vary along
rays with different angles from the center, this serves largely as an averaging technique, used for comparisons across
cities. Some cities have land rent gradients that decline monotonically from the center all the way to their agricultural
fringe. Others see a dip in central-city values that rise again for the inner suburbs, before declining again at the fringe.
4
3
∗
iosyncratic component (σe2 ). The second component in the intercept, αjt
, captures the city-specific
time trend. By similar logic, we shrink the MSA-level time trend toward the national level time
trend, τt , where the degree of the heterogeneity in MSA-level time trends is allowed to change over
time through σt2 .
In practice, to estimate the BLUP for the random intercept and gradient parameters, unknown
fixed parameters (β, a0 , a1 , d0 , d1 ) and variance parameters (σ 2 , Σαα , Σαδ , Σδδ ) must be estimated
as well. In this paper, we adopt an empirical Bayes-type approach in which these parameters are
found by maximizing the marginal likelihood with a flat improper prior. Then, we obtain estimates
for [αjt , δj0 ]0 by substituting these estimates into the posterior mean formula as if the fixed and
variance parameters were known. Appendix B describes the shrinkage procedure in more detail.
Figure 1: Example of Land Value Gradient Estimates for the Houston, TX Metro Area
(b) Estimated Land Value Surface with Census Tract
Centroids
(a) Estimated Distance Polynomial with
D = ln(1 + mileage)
3.3
Integrating Land Values Over the Urban Area
We use the estimated land value functions to compute average land values over each city’s urban
area in each year. For each census tract l in city j in year t, we calculate the predicted land value rjlt
at the tract centroid and assign that average value to the entire tract.5 This value is then multiplied
by the area of each tract Ajl , excluding any non-urban block groups. The total value of land in
P
city j is then Rjt = l Ajl rjlt , and the average value is rjt = Rjt /Aj . In other words, total land
values in city j are the volume of the estimated land value “cone,” while the average land value is
the cone’s average height. Figure 1b displays the estimated cone for the Houston MSA, with the
5
We use only αjt and δj , as well as the distance to the coast, to predict the land values, setting all other covariates
equal to the sample averages. We include only tract centers within 60 miles of the city center.
4
small dots representing Census tract centers. Very high land values at the city center are clearly
visible in the figure, which also shows high values for the Census tracts near the coast.
The estimated “meta-city” allows us to impute land values for metros with no observations, in
which case Wj = I. Tract values are imputed based on typical intercepts and gradients for cities
of size Aj in year t, based on their position relative to the closest city center and coastlines.
3.4
Model Selection and Cross-Validation
The cross-validation exercise summarized in table 1 assesses the performance of several econometric specifications, as detailed in appendix B. The exercise fixes a number of MSAs, and retains
a few observations per year. It then uses those few observations and the model estimates from
other MSAs to predict the values of the non-retained observations. The mean squared error (MSE)
between the predicted price and the actual price of these non-retained observations is used to assess
the model. Results in Panel A retain 3 observations per city-year; panel B retains 30.
TABLE 1: ECONOMETRIC MODEL CROSS-VALIDATION RESULTS
(1)
Model Specification
(3)
(4)
(5)
(2)
(6)
(7)
Panel A: 3 observations per city-year
Mean Squared Error
1.640
Bias
-0.004
Variance
1.586
1.143
0.013
1.105
0.939
0.016
0.910
0.938
0.013
0.909
0.936
0.013
0.907
0.936
0.013
0.906
0.935
0.013
0.905
Panel B: 30 observations per city-year
Mean Squared Error
1.449
Bias
-0.004
Variance
1.441
0.912
-0.003
0.907
0.904
0.001
0.899
0.902
0.000
0.898
0.898
0.001
0.893
0.897
0.001
0.892
0.896
0.000
0.891
No
1
1
Yes
1
1
Yes
2
1
Yes
3
1
Yes
4
1
Yes
4
3
Shrunken?
Polynomial Order - Distance
Polynomial Order - Lot Size
No
0
0
Out-of-sample cross-validation exercise described in detail in the appendix. Column 1 shows results of a
naive model that is the simple average of values per acre. Columns 2 through 7 contain controls for all
covariates in Appendix Table A.1. Panel A shows results for exercise in which 3 observations per cityyear are combined with all out-of-city data to predict remaining land values in city. Panel B shows
results for exercise in which 30 observations per city-year are combined with all out-of-city data to
predict remaining land values in city. Out of sample predictions in btoh panels were conducted in 58
cities that had at least 50 observations per year for at least two years.
The first specification, in column 1, is of a “naive” model that takes the (geometric) average
value per acre of all sales by metro. It establishes a baseline for other models to improve upon. The
5
second column shows the results from a simple version of model (1), with only linear city-specific
terms in distance (K = 1), as well as city-time specific intercepts, measures of coastal proximity,
controls for proposed use, and a linear term in log lot size. This basic econometric model lowers
the mean squared error (MSE) over the naive model substantially by reducing the variance of the
estimates. The third specification applies the empirical Bayes shrinkage technique according to
the prior (2), allowing both intercepts and gradients to be random. As expected, this produces a
substantial improvement by further reducing the variance. Thus, both the monocentric regression
model and shrinkage help overcome the obstacles of small samples and non-random locations, as
seen by lower prediction errors.
The rest of the table considers what are minor improvements. The fourth through sixth columns
contain add additional distance polynomials to the model in 3. Allowing for a more flexible distance gradient reduces the MSE only moderately. The final column includes a cubic polynomial in
log lot size, which also slightly improves the prediction. As further terms produce no noticeable
improvement, we take the model with Bayesian shrinkage, a quartic polynomial in distance, and a
cubic polynomial in log lot size as our preferred specification.
4
Cross-Sectional Results
4.1
Patterns in the Data
Figures 2a-2c plot estimated central land values, the ratio of those values to values 10 miles from
downtown, and average land values, each against the urban area of the metro area.6 The grey dots
represent the unshrunken estimates; the dark dots, the shrunken estimates: the vertical distances
between the two display how much the Bayesian approach shrinks the estimates. Larger cities,
which feature more observations, experience less shrinkage, as the additional observations make
the prior less important.
The dashed upward-sloping line of best fit in Figure 2a reflects the tendency of larger cities
to have more expensive central land. A ten-percent increase in a city’s footprint implies an 8percent increase in the central land value. The upward-sloping fitted line in Figure 2b reveals that
in larger cities central land values tend to be much higher relative to values 10 miles away. For
the smallest cities the gradient is typically flat. In large cities, the ratio is much larger, but highly
variable, even after shrinkage. Together, these two patterns lead to the weaker, but still positive,
correlation between city size and average urban land values in Figure 2c. These empirical results
are generally supportive of a monocentric city with convex rent gradients. These gradients steepen
with centrality due to sorting and substitution effects among households and firms, who increase
6
We take land values one-half mile from the point defined at the center as our measure of central land values.
6
their bid per-acre to pay for proximity, as they use diminishing amounts of land.7
Figure 2: Estimation Results - All Metro Areas
(a) Central Land Values
(b) Ratio of Central to 10-mile
Distant Land Values
(c) Average Land Values
While our main interest is estimating land values and their cross-sectional differences by MSA,
it is worth briefly describing the estimated coefficients on the model covariates, presented in Table
A1 of appendix A.1. The most important predictor of log value per acre is log lot size, which
enters the regression model as a cubic polynomial. The estimates implying that price per acre
is declining in lot size over the size range. This is a standard result called the “plattage effect”,
described by Colwell and Sirmans (1993) in this Review as “a well-known empirical regularity.” It
is often ascribed to costs from subdividing land parcels, both from infrastructure and zoning.8
Many of the planned uses have statistically and economically significant associations with land
values: retail, apartment, mixed use, and medical proposed uses have substantially higher values,
while commercial, industrial, and multifamily uses have lower values. Lots with no planned use,
or a planned use of “hold for development” or “hold for investment” also have lower values. Not
surprisingly, within-metro land values rise with coastal proximity. Unlike lot sizes and planned
uses, which are averaged out of the predicted value of tract centroids, the effect of coastal proximity
is factored back into those predictions in the construction of the land value index.
4.2
A Cross-Metropolitan Land Value Index
Table 2 presents urban land value estimates for selected metro areas.9 The first two columns show
the name of each MSA, and its rank out of 324 according to the estimated land value in our
7
Combes et al. (2016) also find that land-rent gradients are steeper in large French cities than in small ones.
With such costs, large lots may contain more land than is optimal for their intended use. For instance, a lot may
have more land than is need to build an apartment building, but cannot be subdivided into two lots on which to build
two apartment buildings. In that case, the price per acre of the large lot will be lower than if it contained the optimal
amount of land for its intended use.
9
Estimates for all MSAs in the sample are available in table A2 of the online appendix.
8
7
preferred model specification in Table 1, column 7. Next are the urban (not total) areas of each
metro, and the number of observed land sales. The fifth column presents average values from
the naive model. Column 6 reports estimated central land values using the preferred model, and
column 7 presents estimated average values across the urban area. Column 8 reports the estimated
ratio of central values to those 10 miles away. The last column provides the total value of urban
land by metro, which is totalled at the bottom.
The numbers in columns 5 and 7 contrast the role of the model-based estimator over the naive
one. While the two are positively correlated with a coefficient of 0.63, the standard deviation of the
naive estimates is nearly seven times higher than that of the model-based estimates. For instance,
New York has the highest naively estimated value per acre, $26.1 million. Pittsfield, MA has the
lowest naively estimated values, $17,000 per acre. In general, MSAs with high naively estimated
values benefit from favorable covariates, such as small lot sizes.
Overall, our estimates cover 76,095 square miles of urban land. The total estimated value of
this land is $17,030 billion on average over the sample period. The average value of urban land
was $350,000 per acre, with an unweighted standard deviation of $238,000 across metro areas.
This average implies a cost of roughly $70,000 for a typical fifth-acre residential lot, or $1,400 for
a typical parking spot.
The highest central land values are found in New York, at $16 million per acre. The remaining
top 5 are, Washington, Honolulu, San Francisco, and Chicago, with values between $6 and 4
million. The relative size of the skyscrapers in these cities — with the smallest in Washington
and the tallest in Chicago — suggest that their skylines have as much to do with building-height
restrictions, as with economic fundamentals.
The highest average values are found in quality locations with smaller land areas. This is
not surprising for Honolulu, at $1.9 million per acre, as it is loaded with densely populated scenic
locations, with miles of coastline and a desirable climate. Second is Jersey City, which is a valuable
strip of 47 square miles worth $1.5 million per acre, with great views of Manhattan. The latter
forms part of the New York PMSA, which averaging in several other counties is worth only $1.3
million per acre. San Jose and San Francisco have average values within this range, which seems
plausible given their natural amenities and economic opportunities.
The top ten cities in terms of average values are all on salt water coasts. In the Midwest,
the values are more moderate: Chicago has an average value of $494 thousand per acre, while
Pittsburgh has an average of $143K. Detroit’s is $240K, despite values of near zero just a few
miles from its center, underscoring the value of its suburbs. In the South, Atlanta has an average
value of $315K per acre, while Houston and Dallas both have average values of around $240K.
The lowest values are found in small cities such as Saginaw-Bay City-Midland, MI, Jackson, MI,
and Jamestown, NY, at less than $50K per acre.
8
TABLE 2: SELECTED METROPOLITAN LAND VALUE INDICES, 2005-2010
Land Values - $000s/Acre
Rank Metropolitan Area Name
(1)
(2)
Total Urban No. of
Area (Sq. Land
Miles)
Sales
(3)
(4)
1
2
3
4
5
6
7
8
9
10
Honolulu, HI
Jersey City, NJ
San Jose, CA
San Francisco, CA
New York, NY
Orange County, CA
Los Angeles-Long Beach, CA
Miami, FL
Oakland, CA
Stamford-Norwalk, CT
198
56
47
43
305
217
300
152
749 1,603
494
233
1,359 1,760
372 1,233
495
132
179
19
13
17
24
30
31
59
65
101
107
109
208
Las Vegas, NV-AZ
Washington, DC-MD-VA-WV
Phoenix-Mesa, AZ
Boston, MA-NH
Chicago, IL
Denver, CO
Atlanta, GA
Houston, TX
Detroit, MI
Dallas, TX
Pittsburgh, PA
317
1,458
897
1,295
2,035
536
2,105
1,341
1,426
1,057
1,003
146
57
46
322 Saginaw-Bay City-Midland, MI
323 Jackson, MI
324 Jamestown, NY
Total U.S.
Simple Average U.S.
Simple Std. Dev. across Metros
Weighted Average U.S.
Wtd. Std. Dev. across Metros
Naïve
Model
(5)
Central
(6)
Urban
Avg.
(7)
Ratio of
Central to Total Urban
10-Mile Land Value
Values ($ billions)
(8)
(9)
4,357
7,667
2,580
8,722
26,139
3,163
3,709
3,052
2,648
2,753
5,067
2,615
1,805
4,652
16,342
1,209
2,579
1,139
1,310
1,040
1,939
1,475
1,419
1,334
1,271
1,249
1,153
944
865
848
3.0
3.1
1.3
3.7
11.9
1.0
1.9
1.3
1.3
1.6
245.4
44.1
277.0
256.5
609.5
394.8
1,002.6
224.8
273.7
97.2
2,553
1,840
5,946
122
3,511
2,015
5,229
1,143
679
811
240
1,193
3,548
370
1,243
1,455
828
402
423
456
454
433
1,000
6,274
1,045
2,088
4,384
1,661
735
690
603
811
496
768
737
533
506
494
350
315
248
240
238
143
1.3
8.0
1.8
3.1
7.1
5.1
2.1
2.8
2.4
4.1
3.2
156.0
687.9
305.7
419.3
643.9
120.1
424.8
213.0
219.3
161.2
91.9
41
8
10
92
49
43
61
48
39
45
38
30
1.2
1.4
1.0
4.3
1.4
0.9
76,581 68,756
235
212
304
592
739
- 1,214
591
1,660
1,052
2,701
487
1,121
1,076
1,944
251
239
350
281
1.7
1.2
2.4
1.9
17,085.1
52.6
109.9
166
208.0
MSAs are ranked by average urban land values. Land-value data from CoStar COMPS database for years 2005 to
2010. Naïve model is simple average of observed prices per acre. Estimated allows land values to depend on quartic
polynomial in log distance from city center plus one mile, with random coefficients. City center land values are for
one-half mile from downtown, and mile 10 land values are for 10 miles from downtown. Weighted statisitcs for U.S.
are weighted by total metropolitan urban area. Standard deviations are unweighted. See online appendix table A2 for
complete list of MSAs. Averages and standard deviations for the U.S. do not include MSAs for which there were no
observed land sales.
Although the estimated rank correlation between central and average land values is 0.88, the
ratio of central values, the ratio of central values to those 10 miles away varies considerably. The
weighted (unweighted) average is 2.4 (1.7), with a standard deviation of 1.9 (1.2). New York has
9
the highest ratio, 11.9; Washington, DC’s is 8.0. Fifty-eight metros have land values that are higher
10 miles from downtown than at their main city center: of these, Orange County, CA is probably
the most prominent.
The Los Angeles PMSA has the greatest total land value of any metro, at just over $1 trillion. This is followed by the Washington PMSA, Chicago PMSA, New York PMSA, and Atlanta
MSA.When the data are aggregated to the higher CMSA level, the New York CMSA has the highest total urban land value, of approximately $1.9 trillion, followed by the Los Angeles CMSA,
with $1.7 trillion: these two urban agglomerations account for over 20% of the value of all urban
land.
4.3
Comparing Transaction- and Residual-based Estimates
A common approach to measure land values is to treat them as a residual, whereby land values are
taken as the difference between a property’s entire value and the estimated value of its structure.
Case (2007) explains how to use FOF data to impute land values in this way, using the replacement
cost of housing structures. A caveat of this method is that it equates the market value of a structure
with its replacement value, neglecting adjustment costs in building and irreversibilities in investment (Glaeser and Gyourko, 2005). When the market value of structures falls below replacement
costs, the residual method assigns the entire decrease to land values. The residual method can even
infer negative value to land, as Davis and Heathcote (2007) do for residential land in 1940, and
as Larson et al. (2015) shows that the Flow of Funds approach implied for the corporate business
sector in 2009, to the extent of $178 billion (Bureau of Economic Analysis, 2013). This appears
illogical as there is certainly a “buyer” who would have been pleased to accept far less than $178
billion to receive all corporate land in the U.S.
Davis and Palumbo (2008) use the residual method to estimate an index of land values across
46 metros that is purely residential. While our transaction-based measure covers all urban land,
theirs is the only other cross-sectional index of its kind. Cross-sectionally, our transaction-based
methodology produces higher values. Comparing the same MSAs, the average value per acre
across the same MSAs (unweighted) in our measure nearly $444K, while theirs is $161K. As
seen in Appendix Table A.3, a fifth acre in our data is $63.1K in Atlanta and $70.0K in Denver,
while in theirs it is $5.3K and $18.2K. While this may have something to do with placement,
methodology likely accounts for much of these differences. In contrast, our index implies smaller
price movements within cities over the sample period. The average coefficient of variation of land
values within the 46 cities according to our index was 0.24, while theirs was 0.43.10
10
Appendix A.2 details the comparison procedure, which it illustrates in Figure A.2. As Davis and Heathcote (2007)
note, the residual method attaches “the label ‘land’ to anything that makes a house worth more than the cost of putting
up a new structure of similar size and quality on a vacant lot.” Thus, the residual method will attribute higher costs
10
5
Aggregate Urban Land Values over Time
In this section, we sum our urban land values across metros to calculate annual aggregate urban
land values for the United States.11 Table 3 presents these totals.
Year
2005
2006
2007
2008
2009
2010
TABLE 3: URBAN LAND VALUES IN THE UNITED STATES, 2005-2010
S&P CoreLogic Total Urban
Avg. Urban
Case-Shiller U.S. Land Value
Avg. Urban Total Urban Land Value per
Ratio of Total National HPI
- Residual
Land Value per Land Value Acre (Index, Nominal GDP Urban Land (Normalized to
Method
Acre ($)
($ billions)
2005 = 100)
($ billions) Value to GDP
2005=100)
($ billions)
402,214
426,529
397,777
352,925
256,533
262,150
19,651
20,838
19,433
17,244
12,536
12,808
100.0
106.0
98.9
87.7
63.8
65.2
13,094
13,856
14,478
14,719
14,419
14,964
1.50
1.50
1.34
1.17
0.87
0.86
100.0
106.8
104.8
95.5
86.5
84.2
16,758
16,931
16,001
9,569
5,767
6,234
Land-value data from CoStar COMPS database for years 2005 to 2010. Residual method calculates total real estate
holdings at market value of nonfinancial businesses, households, and nonprofit organizations from Financial Accounts of
the United States (formerly known as the Flow of Funds) and subtracts current-cost net stock of private structures from
National Income and Product Accounts.
Over our sample period, average values peaked in 2006 at $427,000 per acre, an increase of
6.0% from 2005. Average values then fell slightly in 2007, before declining precipitously. By 2009
the average value was roughly $257,000 per acre, 64% of its 2005 level. The ratio of aggregate
urban land values to gross domestic product declined considerably as well. The ratio was 1.50 in
2005 and 2006 before declining to reach a value 0.86 by 2010.
For comparison, we construct a series for aggregate U.S. land values using the residual method
based on FOF data (now the Financial Accounts of the U.S.). We sum the total value of real estate
at market value held by non-financial non-corporate businesses, non-financial corporate businesses,
and households and nonprofit organizations to arrive at the total market value of privately held real
estate. We then subtract the current-cost net stock of private structures to arrive at a residual-based
value for land. In 2006, the estimated value of real estate was $43.3 trillion, while structures were
valued at $26.3 trillion, implying that the total value of land was $16.9 trillion. Our transactionsbased estimate, in contrast, is $20.8 trillion, roughly 23% higher.
In addition to the methodological differences, the totals may differ because they cover different
stemming from inefficiencies in factor usage – e.g., geographic and regulatory constraints that hinder building – to
higher land values. In a follow-up paper, Albouy and Ehrlich (2016) use differences in the value of housing prices
from land and structure costs to measure the costs imposed by such constraints. See Glaeser and Gyourko (2017) for
a related, but more reduced-form approach that assumes land is a fixed fraction of housing costs.
11
Our sample includes observations from 324 out of the 331 MSAs and PMSAs in the 1999 OMB definitions. The
combined imputed land value for the seven metros with no data is $63 billion for 2005, less than half a percent of our
aggregate number.
11
land. Our estimates are based on total metro urban areas, including public lands for roads, parks,
and civic buildings. Assuming that the public owns urban land worth 40% of the total value, $12.5
trillion of land would be owned privately. On the other hand, the FOF numbers include land outside
of metro-urban areas, which we exclude.
Land values calculated from the FOF fell even more dramatically than our series, down to
only $5.8 trillion in 2009. The peak-to-trough decline in the transactions-based index was 39%,
substantially less than the 63% decline in the FOF.
Last, we consider how land values compare with housing prices. The final column of table 3 reports the S&P CoreLogic Case-Shiller U.S. National House Price Index, normalized to have value
100 in 2005. Overall, land values appear to have led house prices slightly, and were substantially
more volatile than house prices over the sample period. This result is consistent with the Bostic
et al. (2007) land leverage hypothesis implying that housing should have less volatile values than
land.
6
Conclusion
The analysis here produces novel and plausible estimates of land values, even in metros with relatively thin data, by combining insights from the monocentric city model with empirical Bayesian
methods. These methods might easily be applied to estimate other city-wide phenomena, such as
on wages or property prices. In addressing the question of how much land is worth, it offers advantages over residual approaches, suggesting that urban land values may be higher, less volatile,
and unlikely to be negative. Furthermore, the model sheds light on the enormous differences in
land values both across and within cities, with high central values providing indirect support for
monocentric cities, albeit with heterogeneous value gradients.
We hope that the measures we provide may form the basis of reliable estimates of aggregate
land wealth. With additional data, future modeling could be enriched to incorporate greater spatial
structure and modifications for observed land uses. The cross-sectional index we provide should
also prove useful to researchers examining differences in amenities and costs across metro areas.
References
David Albouy. What are cities worth? land rents, local productivity, and the total value of amenities. Review of Economics and Statistics, 98(3):477–487, 2016.
David Albouy and Gabriel Ehrlich. Housing productivity and the social cost of land use restrictions. NBER Working Paper, 18110, 2016.
12
William Alonso. Location and land use. Toward a general theory of land rent. Harvard University
Press, 1964.
Raphael W Bostic, Stanley D Longhofer, and Christian L Redfearn. Land leverage: decomposing
home price dynamics. Real Estate Economics, 35(2):183–208, 2007.
Jan K. Brueckner. Property value maximization and public-sector efficiency. Journal of Urban
Economics, 14:1–15, 1983.
Bureau of Economic Analysis. Relation of BEA’s Current-Cost Net Stock of Private Structures to
the corresponding items in the Federal Reserve Board’s Financial Accounts of the United States.
Technical report, 2013. URL https://www.bea.gov/national/pdf/st13.pdf.
Karl Case. The value of land in the united states: 1975-2005. In GK Ingram and Y.-H. Hong,
editors, Land Policies and Their Outcomes. Lincoln Institute of Land Policy, 2007.
Peter F Colwell and Henry J Munneke. The structure of urban land prices. Journal of Urban
Economics, 41(3):321–336, 1997.
Peter F Colwell and Clemon F Sirmans. A comment on zoning, returns to scale, and the value of
undeveloped land. The Review of Economics and Statistics, pages 783–786, 1993.
Pierre-Philippe Combes, Gilles Duranton, and Duranton Gobillon. The costs of agglomeration:
House and land prices in french cities. mimeo, 2016.
Morris A Davis and Jonathan Heathcote. The price and quantity of residential land in the united
states. Journal of Monetary Economics, 54(8):2595–2620, 2007.
Morris A Davis and Michael G Palumbo. The price of residential land in large us cities. Journal
of Urban Economics, 63(1):352–384, 2008.
Jeff Fisher, David Geltner, and Henry Pollakowski. A quarterly transactions-based index of institutional real estate investment performance and movements in supply and demand. Journal of
Real Estate Finance and Economics, 34:5–33, 2007.
Henry George. Progress and poverty: An inquiry into the cause of industrial depressions, and of
increase of want with increase of wealth, the remedy. W. Reeves, 1884.
Edward L Glaeser and Joseph Gyourko. Urban decline and durable housing. Journal of political
economy, 113(2):345–375, 2005.
Edward L Glaeser and Joseph Gyourko. The economic implications of housing supply. Zell/Lurie
Wroking Paper, 802, 2017.
D. A. Harville. Maximum likelihood approaches to variance component estimation and to related
problems. Journal of the American Statistical Association, 72(358):320–338, 1977.
Andrew Haughwout, James Orr, and David Bedoll. The price of land in the new york metropolitan
area. Federal Reserve Bank of New York Current Issues in Economics and Finance, 2008.
Nils Kok, Paavo Monkkonen, and John M Quigley. Land use regulations and the value of land and
housing: An intra-metropolitan analysis. Journal of Urban Economics, 81:136–148, 2014.
13
Nan M. Laird and Thomas A. Louis. Empirical bayes ranking methods. Journal of Education
Statistics, 14(1):29–46, 1989.
William Larson et al. New estimates of value of land of the united states. Technical report, Bureau
of Economic Analysis, 2015.
Edwin S. Mills. An aggregative model of resource allocation in a metropolitan area. American
Economic Review, 57:197–210, 1967.
Henry Munneke and Slade Barrett. An empirical study of sample-selection bias in indices of
commercial real estate. Journal of Real Estate Finance and Economics, 21:45–64, 2000.
Henry Munneke and Slade Barrett. A metropolitan transaction-based commercial price index: A
time-varying parameter approach. Real Estate Economics, 29:55–84, 2001.
Richard F Muth. Cities and Housing: the Spatial Pattern of Urban Residential Land Use. University of Chicago Press, 1969.
Joseph B Nichols, Stephen D Oliner, and Michael R Mulhall. Swings in commercial and residential
land prices in the united states. Journal of Urban Economics, 73(1):57–76, 2013.
Adam Ozimek, Daniel Miles, et al. Stata utilities for geocoding and generating travel time and
travel distance information. Stata Journal, 11(1):106, 2011.
David Ricardo. On the principles of political economy, and taxation. John Murray, 1821.
Jennifer Roback. Wages, rents, and the quality of life. Journal of Political Economy, 90:1257–
1258, 1982.
S.L. Zeger and M.R. Karim. Generalized linear models with random effects: A gibbs sampling
approach. Journal of the American Statistical Association, 86, 1991.
14
A
A.1
Appendix for Online Publication
Additional Data Notes
When a CMSA contains multiple PMSAs, we treat each PMSA as its own MSA for purposes of
estimation and reporting. For instance, we treat the Washington, DC-MD-VA-WV and Baltimore,
MD PMSAs as separate MSAs, although they are both parts of the Washington-Baltimore DCMD-VA-WV CMSA. For New York City we use the Empire State Building as the city center
rather than City Hall, following Haughwout et al. (2008). We treat each named city in an MSA
with a hyphenated city name as having its own city center. For instance, we treat Minneapolis-St.
Paul, MN-WI as containing two distinct cities, Minneapolis, MN, and St. Paul, MN. However, we
treat such cities as belonging to one MSA for purposes of estimation and reporting. We exclude
MSAs located in Puerto Rico from the analysis.
In the CoStar data, we consider 12 of the most common proposed uses, which are neither
mutually exclusive nor collectively exhaustive. We consider an observation to feature a structure
when the transaction record includes the fields for “ Bldg Type”, “ Year Built”, “ Age”, or the
phrase “ Business Value Included” in the field “ Sale Conditions.” We geocoded the lot sales using
the Stata “geocode” module of Ozimek et al. (2011). In addition to the exclusions discussed in
the main text, we also exclude outlier observations with a listed price of less than $100 per acre
or a lot size over 5,000 acres, or further than 60 miles away from the city center. We also exclude
locations we could not geocode successfully.
Median lot size is 3.5 acres versus a mean of 26 acres. Land sales occur more frequently in
the beginning of our sample period, with 21.7% of our sample from 2005 and 11.4% from 2010.
Residential uses are common but by no means predominant in the sample: 17.6% of properties have
a proposed use of single-family, multifamily, or apartments. 23.4% is being held for development
or investment, and 16% of the sample had no listed proposed use.
A.2
Comparison to Davis and Palumbo Index
The Davis and Palumbo (2008) index is quarterly; we take geometric averages to arrive at annual
and whole-sample values. Matching our MSAs to their cities is typically straightforward using
the name of the principal city; we match their estimates for Santa Ana to the Orange County, CA
MSA.
Figure A.2 displays graphical comparisons between the transactions-based estimates developed
in this paper and the residual-based estimates in Davis and Palumbo (2008). Figure A.2a plots
the estimated average land values for each city. The transactions-based estimates are uniformly
higher than the residual-based estimates. Figure A.2b plots the estimated difference between the
minimum and maximum annual estimated average land values within each city, expressed as a
percentage of the maximum value. The estimates from the residual-based method typically show
greater variation than the estimates from the transactions-based method, as also seen in the time
series for aggregate U.S. land values.
i
B
Computation
Estimation of αjt and δj
B.1
For notational convenience we rewrite the model in (1) as
0
ln rijt = Zijt
γj + Xijt β + eijt ,
eijt ∼ N (0, σe2 ),
(A.1)
2007 2008 2009 2010
4
3
2
0
, with Dij = D(zij , zcj ), and
, 12006
, Dij
, Dij
= 1, Dij , Dij
where Zijt
ijt , 1ijt , 1ijt , 1ijt , 1ijt
where 1sijt is an indicator variable that takes value 1 if s = t and 0 otherwise. The parameter vector
γj collects city and time specific parameters with a multivariate normal prior distribution
γj = [αj , δj1 , δj2 , δj3 , δj4 , αj,2006 , αj,2007 , αj,2008 , αj,2009 , αj,2010 ]0 ∼ N (mγ,j , Vγ,0 )
where

mγ,j

a0 + a1 A j
= b0 + b1 Aj 
τ

and
Vγ,0

Σαα
Σαδ 0(1×5)
Σδδ 0(4×5) 
=  Σδα
0(5×1) 0(5×4) Στ τ
(A.2)
(A.3)
2
2
2
2
2
2
]0 ).
, σ2010
, σ2009
, σ2008
, σ2007
, σ2006
with τ = [τ2006 , τ2007 , τ2008 , τ2009 , τ2010 ]0 and Στ τ = diag([σ2005
Conditional on fixed and variance parameters (θ = [β, a0 , a1 , b0 , b1 , τ, σe2 , Σαα , Σδα , Στ τ ]) and
observed data for city j, the posterior distribution of γj follows the multivariate normal distribution
e
γj |θ, Data ∼ N m
e γ,j (θ), Vγ,j (θ)
(A.4)
with a posterior mean asthe weighted average
between the prior mean (mγ,0 ) and the fixed effect
−1
0
0
Zj (ln rj − Xj β) :
estimate γ
bj = (Zj Zj )
m
e γ,j (θ) = Wj (θ)mγ,j + [I − Wj (θ)] γ
bj (θ)
−1 −1
where Wj (θ) = V0−1 + σe−2 (Zj0 Zj )
V0 . (A.5)
Here we write Zj , ln rj , and Xj as matrices that stack elements only relevant for the city j. The
weight depends on the number of observations in the city j (nj ), the relative size of the prior
variance (Vj ) and the idiosyncratic shock variance (σ 2 ). The posterior variance is
−1
.
Veγ,j (θ) = V0−1 + σe−2 (Zj0 Zj )
(A.6)
It is well known that the posterior mean m
e γ,j (θ) is the best linear unbiased predictor for γj given
θ and the observed data. In our application, we do not know θ. Instead of taking a full Bayesian
approach and putting a prior on θ, we take the empirical Bayesian approach in which θ is calibrated
by maximizing the following marginal likelihood (Laird and Louis, 1989):
Z
b
θ ∈ argminθ L(data|θ) = p(ln rijt |Z, X, θ, γ)dγ
(A.7)
where the γ is integrated out from the conditional posterior distribution an improper prior, p(γ) ∝
1, viz. Harville (1977). Then, we treat θb as a known and fixed quantity and use the following
ii
posterior distribution for the computation of land values and the prediction,
b
b
e
γj |Data ∼ N m
e γ,j (θ), Vγ,j (θ) .
(A.8)
One of the potential shortcomings of this approach is that it neglects uncertainty coming from the
estimation of θ, and the resulting posterior distribution for γj underestimates uncertainty. Fortunately, we have a relatively large amount of data about θ (about 67,000 observations in total).
Second, the practicality of our shrinkage estimator is evaluated by the out-of-sample forecasting
evaluation. However, we note that a full Bayesian approach is possible (Zeger and Karim, 1991)
at the cost of longer computation time. We choose to take the empirical Bayes approach because
of the out-of-sample evaluation of our shrinkage procedure.
B.2
Point prediction
Once we obtain the posterior distribution of γj , we can generate land value predictions. For the
cross-validation exercise, we generate and evaluate point predictions for the log-price of the land
∗
∗
as the mean of the posterior predictive
and Zijt
parcels in the city j at time t with characteristic Xijt
distribution. When the land parcel the value of which we predict is in a city j that has observed
data, this point prediction is
Z
∗
∗
[
ln rijt = ln rijt p(ln rijt |data, Xijt
, Zijt
)d ln rijt
Z Z
(A.9)
∗
∗
=
ln rijt p(ln rijt |data, Xijt
, Zijt
, γj )p(γj |data)dγj d ln rijt
∗0
b + X ∗0 β.
b
= Zijt
m
e γ,j (θ)
ijt
We can also generate predictions for the land parcels that are in a “missing” city where we do not
have observed transaction prices. In this case, our prediction is just
∗0
b + X ∗0 β,
b
[
ln
rijt = Zijt
mγ,j (θ)
ijt
(A.10)
b
where our prediction is based on the prior with estimated hyperparameters, θ.
B.3
Computation of Land values
For each census tract l in city j in year t, we calculate the predicted land value rilt at the tract
centroid and assign that average value to the entire tract.
Rjt =
L
X
l=1
iii
rbilt Al
(A.11)
where Al is the tract area we use the mean of the predictive distribution for rilt as the predicted
land value. That is,
Z
∗∗
∗∗
rbilt = exp(rilt )p(rilt |data, Xilt
, Zilt
)drilt
Z Z
∗∗ ∗∗
(A.12)
Zilt , γj )p(γj |data)drilt dγj
=
exp(rilt )p(rilt |data, Xilt
0
0
b + X ∗∗0 βb + 1/2b
b ∗∗0
= exp Z ∗∗ mγ,j (θ)
σ 2 + 1/2Z ∗∗ Vγ,j (θ)Z
ilt
ilt
e
ilt
ilt
where the last two terms are due to the log-normal correction. We can also estimate values for
cities with no observed land sales using only the prior.
B.4
Cross-validation
The cross-validation technique helps to evaluate the effectiveness of the econometric specification
and shrinkage model. We design a pseudo out-of-sample prediction exercise that quantifies the
potential gains or losses from different models. For this exercise we take cities that have at least 50
observations per year for at least two years This leads to 58 cities with 55,155 total observations.
Then, for each city j,
1. Randomly choose nhold observations out of njt observations for each time t = 2005, 2006, ..., 2010
in city j. We keep those 6 ∗ nhold observations as well as the remaining sample of data from
other cities.
2. Estimate each of models using the method described in subsection B.1
3. Generate predictions for sample held out in step 1 for city j based on the method in subsection B.2
4. Compute and store the prediction error for this hold-out samples. {ej,r,1 , ej,r,2 , ..., ej,r,(njt −nhold ) }
(forecast errors are defined as predicted minus actual).
5. Repeat Step 1 – Step 4 for r = 1, 2, ..., R.
6. Repeat Step 1 – Step 5 for each city j = 1, , J.
7. Compute aggregated out-of-sample prediction evaluation statistics. For example, the MSE
for the city j is computed as
M SE(j) =
1
R × (nj,t − nhold )
−nhold )
R (njtX
X
r=1
e2r,j,i
(A.13)
i=1
where we set R = 30. We perform for nhold = 3 (small sample size) and nhold = 30 (moderate
sample size) for each city. About 35% of MSAs in our sample have observations less than equal to
18 observations (which is approximately 3 per year in our data set) and about 81% of MSAs in our
sample have observations less than equal to 180 (which is approximately 30 per year in our data
set). We report average M SE(j) over j = 1, 2, ...58.
iv
Unshrunken Estimator. The unshrunken estimates are based on the fixed effect estimation. The
estimator is defined as γ
bj = (Zj0 Zj )−1 (Zj0 (ln rj − Xj β) and used in Equation A.5.
v
TABLE A1: ESTIMATED COEFFICIENTS ON COVARIATES IN
PREFERRED SPECIFICATION
Estimated Standard
Covariate
Coefficient Error
t-statistic p-value
Log Lot Size
(Log Lot Size Squared)/100
(Log Lot Size Cubed)/1000
Log Distance to Coast
-0.543
-3.053
3.601
-0.052
0.0037
0.1592
0.2498
0.0043
-146.134
-19.176
14.415
-12.196
0.000
0.000
0.000
0.000
Planned Use:
None Listed
-0.182
0.0112
-16.193 0.000
Commercial
-0.380
0.0599
-6.354 0.000
Industrial
-0.346
0.0141
-24.578 0.000
Retail
0.255
0.0134
18.963 0.000
Single Family
0.003
0.0133
0.202 0.840
Multifamily
-0.139
0.0198
-7.055 0.000
Office
0.046
0.0148
3.129 0.002
Apartment
0.288
0.0196
14.713 0.000
Hold for Development
-0.073
0.0118
-6.171 0.000
Hold for Investment
-0.283
0.0195
-14.523 0.000
Mixed Use
0.250
0.0265
9.438 0.000
Medical
0.171
0.0355
4.810 0.000
Parking
0.076
0.0373
2.044 0.041
This table reports the coefficients on the covariates from the preferred
specification in table 1 from the main body of the text, which applies
shrinakge to a model with a quartic polynomial in log distance to the city
center plus one mile.
vi
TABLE A2: METROPOLITAN LAND VALUE INDICES RANKED BY AVERAGE URBAN LAND VALUE PER ACRE, 2005-2010
Land Values - $000s/Acre
Ratio of
Total Est.
Total Metro
Urban
Central to Urban Land
Urban Area Observed
Central Average - Mile 10 Land Value ($
Rank Metropolitan Area Name
(Sq. Miles) Land Sales Naïve Model Estimated
Estimated
Values
billions)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
Honolulu, HI
Jersey City, NJ
San Jose, CA
San Francisco, CA
New York, NY
Orange County, CA
Los Angeles-Long Beach, CA
Miami, FL
Oakland, CA
Stamford-Norwalk, CT
Fort Lauderdale, FL
West Palm Beach-Boca Raton, FL
Las Vegas, NV-AZ
Seattle-Bellevue-Everett, WA
San Diego, CA
Bergen-Passaic, NJ
Washington, DC-MD-VA-WV
Ventura, CA
Nassau-Suffolk, NY
Santa Barbara-Santa Maria-Lompoc, CA
Santa Cruz-Watsonville, CA
Naples, FL
Santa Rosa, CA
Phoenix-Mesa, AZ
Sarasota-Bradenton, FL
Newark, NJ
Portland-Vancouver, OR-WA
Vallejo-Fairfield-Napa, CA
San Luis Obispo-Atascadero-Paso Robles, CA
Boston, MA-NH
Chicago, IL
Tacoma, WA
Orlando, FL
198
47
305
300
749
494
1,359
372
495
179
372
398
317
782
803
316
1,458
202
850
159
72
145
144
897
260
567
527
129
91
1,295
2,035
308
666
56
43
217
152
1,603
233
1,760
1,233
132
19
741
321
2,553
1,626
957
79
1,840
131
396
29
12
78
153
5,946
601
142
1,191
146
43
122
3,511
539
1,612
4,357
7,667
2,580
8,722
26,139
3,163
3,709
3,052
2,648
2,753
2,417
2,188
1,193
2,741
2,488
1,957
3,548
1,048
1,540
2,345
2,007
791
1,034
370
893
2,059
777
786
1,416
1,243
1,455
570
739
5,067
2,615
1,805
4,652
16,342
1,209
2,579
1,139
1,310
1,040
1,166
1,852
1,000
3,268
2,925
1,089
6,274
1,097
595
864
944
627
679
1,045
700
1,304
1,769
705
577
2,088
4,384
933
1,324
1,939
1,475
1,419
1,334
1,271
1,249
1,153
944
865
848
836
780
768
759
754
746
737
715
709
689
644
641
631
533
529
519
512
510
509
506
494
494
492
3.0
3.1
1.3
3.7
11.9
1.0
1.9
1.3
1.3
1.6
1.4
2.4
1.3
4.8
3.4
1.4
8.0
1.7
0.8
1.3
2.3
1.0
1.2
1.8
1.3
1.9
3.5
1.5
1.1
3.1
7.1
1.9
2.4
245.4
44.1
277.0
256.5
609.5
394.8
1,002.6
224.8
273.7
97.2
199.1
199.0
156.0
379.5
387.1
150.9
687.9
92.7
385.7
70.3
29.5
59.4
58.2
305.7
88.1
188.4
172.9
41.9
29.6
419.3
643.9
97.5
209.7
TABLE A2: METROPOLITAN LAND VALUE INDICES RANKED BY AVERAGE URBAN LAND VALUE PER ACRE, 2005-2010
Land Values - $000s/Acre
Ratio of
Total Est.
Total Metro
Urban
Central to Urban Land
Urban Area Observed
Central Average - Mile 10 Land Value ($
Rank Metropolitan Area Name
(Sq. Miles) Land Sales Naïve Model Estimated
Estimated
Values
billions)
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
Middlesex-Somerset-Hunterdon, NJ
Yolo, CA
Sacramento, CA
Baltimore, MD
Tampa-St. Petersburg-Clearwater, FL
Provo-Orem, UT
Fort Myers-Cape Coral, FL
Anchorage, AK
Melbourne-Titusville-Palm Bay, FL
Riverside-San Bernardino, CA
Trenton, NJ
Olympia, WA
Lowell, MA-NH
Stockton-Lodi, CA
Philadelphia, PA-NJ
Jacksonville, FL
Reno, NV
Charlottesville, VA
Salt Lake City-Ogden, UT
Minneapolis-St. Paul, MN-WI
Salinas, CA
Fort Pierce-Port St. Lucie, FL
Monmouth-Ocean, NJ
Wilmington-Newark, DE-MD
Boulder-Longmont, CO
Denver, CO
New Bedford, MA
Reading, PA
Modesto, CA
Punta Gorda, FL
Fresno, CA
Atlanta, GA
Grand Junction, CO
424
34
412
858
957
105
240
94
255
971
127
105
161
130
1,725
497
125
47
411
1,026
101
180
524
215
91
536
72
124
120
98
215
2,105
60
101
50
448
802
1,220
47
294
21
420
2,452
35
250
12
163
859
793
57
4
145
846
12
71
124
107
183
2,015
14
36
142
63
137
5,229
21
828
624
602
969
1,144
499
593
851
688
433
432
455
544
531
939
559
530
728
482
613
814
475
642
445
758
828
503
324
407
648
247
402
343
665
508
588
1,003
1,111
566
258
234
432
514
438
431
590
331
2,201
568
513
372
517
980
360
324
839
462
299
1,661
424
287
434
433
324
735
365
468
446
433
431
421
415
414
410
398
397
395
394
392
391
383
381
379
373
371
368
365
358
357
355
353
350
347
347
342
331
318
315
310
1.4
1.2
1.3
2.2
2.6
1.5
0.6
0.7
1.2
1.0
1.0
1.3
1.7
0.7
5.5
1.4
1.6
1.2
1.3
2.3
1.1
0.9
2.7
1.5
1.0
5.1
1.5
0.9
1.3
1.2
1.0
2.1
1.3
127.0
9.6
114.1
236.5
257.9
27.9
63.5
24.6
64.9
247.1
32.2
26.5
40.3
32.5
422.7
121.1
30.4
11.2
97.6
241.7
23.5
41.3
119.7
48.8
20.5
120.1
16.1
27.5
26.3
20.8
43.8
424.8
11.9
TABLE A2: METROPOLITAN LAND VALUE INDICES RANKED BY AVERAGE URBAN LAND VALUE PER ACRE, 2005-2010
Land Values - $000s/Acre
Ratio of
Total Est.
Total Metro
Urban
Central to Urban Land
Urban Area Observed
Central Average - Mile 10 Land Value ($
Rank Metropolitan Area Name
(Sq. Miles) Land Sales Naïve Model Estimated
Estimated
Values
billions)
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
Merced, CA
Norfolk-Virginia Beach-Newport News, VA-NC
Madison, WI
Danbury, CT
Providence-Fall River-Warwick, RI-MA
Allentown-Bethlehem-Easton, PA
Albuquerque, NM
Raleigh-Durham-Chapel Hill, NC
Fort Walton Beach, FL
Lakeland-Winter Haven, FL
Kenosha, WI
Daytona Beach, FL
Lawrence, MA-NH
Austin-San Marcos, TX
Flagstaff, AZ-UT
New Haven-Meriden, CT
Barnstable-Yarmouth, MA
Gary, IN
Colorado Springs, CO
Bremerton, WA
Tucson, AZ
Hagerstown, MD
Bridgeport, CT
Atlantic-Cape May, NJ
Charleston-North Charleston, SC
Newburgh, NY-PA
Brockton, MA
Huntsville, AL
Salem, OR
Hamilton-Middletown, OH
Savannah, GA
Wilmington, NC
Cleveland-Lorain-Elyria, OH
61
554
130
160
439
264
281
532
86
247
64
225
236
423
35
270
188
285
199
130
325
49
200
174
238
169
148
179
109
124
128
139
745
64
392
239
23
62
85
114
782
14
561
58
93
29
384
7
43
3
111
892
18
1,749
28
26
37
214
54
22
29
54
151
64
50
416
319
377
468
426
1,194
281
413
497
300
324
604
539
410
434
294
658
387
468
409
465
320
348
837
538
498
206
362
190
356
372
337
420
545
258
541
1,337
476
781
420
256
397
444
314
168
201
570
1,846
296
731
411
806
332
453
392
403
420
476
618
314
391
320
498
167
270
353
367
307
307
306
306
303
302
299
296
295
293
293
291
288
283
283
283
279
276
273
272
271
268
268
267
267
265
262
260
258
256
253
252
252
0.9
1.7
5.9
1.9
3.3
1.8
0.8
1.3
1.6
1.1
0.5
0.7
2.4
6.9
1.5
2.8
1.6
3.5
1.2
1.7
1.4
3.0
1.5
1.9
2.8
1.3
1.6
1.1
2.2
0.5
1.0
1.4
1.2
11.9
108.8
25.5
31.4
85.0
51.0
53.7
100.8
16.2
46.3
12.1
42.0
43.4
76.8
6.4
48.9
33.5
50.4
34.8
22.6
56.4
8.4
34.2
29.8
40.6
28.6
24.9
29.7
17.9
20.3
20.7
22.5
120.0
TABLE A2: METROPOLITAN LAND VALUE INDICES RANKED BY AVERAGE URBAN LAND VALUE PER ACRE, 2005-2010
Land Values - $000s/Acre
Ratio of
Total Est.
Total Metro
Urban
Central to Urban Land
Urban Area Observed
Central Average - Mile 10 Land Value ($
Rank Metropolitan Area Name
(Sq. Miles) Land Sales Naïve Model Estimated
Estimated
Values
billions)
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
Nashville, TN
Houston, TX
Cincinnati, OH-KY-IN
Fort Collins-Loveland, CO
Iowa City, IA
Columbus, OH
Visalia-Tulare-Porterville, CA
Detroit, MI
Boise City, ID
Dallas, TX
Worcester, MA-CT
Panama City, FL
Greeley, CO
Dutchess County, NY
Springfield, MA
Sioux Falls, SD
Eugene-Springfield, OR
Janesville-Beloit, WI
Bakersfield, CA
Fayetteville-Springdale-Rogers, AR
Monroe, LA
Burlington, VT
New Orleans, LA
Columbus, GA-AL
Milwaukee-Waukesha, WI
Asheville, NC
Chattanooga, TN-GA
Tallahassee, FL
Myrtle Beach, SC
Lexington, KY
Richmond-Petersburg, VA
Bellingham, WA
Medford-Ashland, OR
580
1,341
645
91
36
512
104
1,426
158
1,057
248
101
48
158
252
49
92
56
161
135
78
68
364
114
542
126
303
125
97
138
441
57
66
455
1,143
637
344
9
671
32
679
106
811
56
41
320
33
28
17
36
15
64
43
7
5
66
11
399
41
51
52
84
29
399
19
12
499
423
441
417
423
614
614
456
294
454
454
815
302
588
523
306
413
277
250
356
360
790
672
250
313
318
387
474
507
345
293
286
379
761
690
518
260
216
527
308
603
361
811
902
124
187
418
497
231
354
269
189
218
272
291
391
294
472
348
437
341
471
224
560
261
241
249
248
248
246
245
244
243
240
239
238
238
234
229
229
228
227
227
226
224
223
221
221
221
220
219
218
216
215
215
214
212
212
211
2.9
2.8
2.1
1.0
1.1
2.0
2.1
2.4
1.6
4.1
4.4
0.4
0.8
1.9
2.5
1.4
2.0
2.4
0.5
1.3
1.6
1.6
1.9
1.6
2.1
1.9
2.0
2.0
2.5
1.0
2.4
1.6
1.3
92.3
213.0
102.3
14.4
5.6
80.0
16.2
219.3
24.2
161.2
37.7
15.2
7.0
23.1
36.7
7.2
13.4
8.1
23.1
19.3
11.1
9.7
51.5
16.1
76.1
17.5
41.8
17.3
13.4
18.9
59.9
7.8
8.9
TABLE A2: METROPOLITAN LAND VALUE INDICES RANKED BY AVERAGE URBAN LAND VALUE PER ACRE, 2005-2010
Land Values - $000s/Acre
Ratio of
Total Est.
Total Metro
Urban
Central to Urban Land
Urban Area Observed
Central Average - Mile 10 Land Value ($
Rank Metropolitan Area Name
(Sq. Miles) Land Sales Naïve Model Estimated
Estimated
Values
billions)
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
Harrisburg-Lebanon-Carlisle, PA
Lincoln, NE
York, PA
Joplin, MO
Indianapolis, IN
Ocala, FL
Richland-Kennewick-Pasco, WA
Lancaster, PA
Santa Fe, NM
Waterbury, CT
Gainesville, FL
La Crosse, WI-MN
Las Cruces, NM
Nashua, NH
Portland, ME
McAllen-Edinburg-Mission, TX
Manchester, NH
Greenville, NC
Billings, MT
Baton Rouge, LA
Ann Arbor, MI
Fort Worth-Arlington, TX
Portsmouth-Rochester, NH-ME
Champaign-Urbana, IL
Fargo-Moorhead, ND-MN
Akron, OH
Steubenville-Weirton, OH-WV
Lafayette, IN
Green Bay, WI
Louisville, KY-IN
El Paso, TX
St. Louis, MO-IL
Vineland-Millville-Bridgeton, NJ
243
78
180
72
649
145
95
229
78
125
79
44
88
96
120
318
97
47
53
292
231
693
133
58
46
337
54
61
83
413
206
979
72
89
24
47
8
193
38
27
57
7
9
34
21
18
3
25
61
23
9
25
99
136
506
13
22
13
169
1
13
49
126
94
364
11
334
258
176
255
274
265
273
176
335
243
384
295
240
147
1,399
400
230
259
297
308
293
313
291
262
470
349
122
190
289
279
321
337
410
511
253
278
254
458
342
177
360
212
185
209
173
229
252
520
288
315
235
173
529
710
284
214
241
156
288
194
168
131
285
150
297
217
210
210
209
208
202
202
201
201
201
198
198
197
197
196
196
196
194
193
191
188
188
187
187
187
186
184
183
183
183
181
181
181
180
2.7
1.7
1.2
1.6
2.0
2.0
0.9
2.2
1.1
1.0
1.4
1.0
1.5
1.5
3.1
2.0
2.2
1.9
1.2
2.5
5.2
1.4
1.2
2.3
0.9
1.4
1.3
1.0
0.9
1.4
0.7
1.6
1.6
32.8
10.5
24.1
9.6
84.0
18.7
12.3
29.4
10.0
15.8
10.0
5.6
11.0
12.1
15.1
39.8
12.1
5.8
6.5
35.1
27.8
83.1
15.9
6.9
5.5
39.6
6.3
7.2
9.8
47.8
23.9
113.1
8.3
TABLE A2: METROPOLITAN LAND VALUE INDICES RANKED BY AVERAGE URBAN LAND VALUE PER ACRE, 2005-2010
Land Values - $000s/Acre
Ratio of
Total Est.
Total Metro
Urban
Central to Urban Land
Urban Area Observed
Central Average - Mile 10 Land Value ($
Rank Metropolitan Area Name
(Sq. Miles) Land Sales Naïve Model Estimated
Estimated
Values
billions)
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
Yuba City, CA
Greensboro--Winston Salem--High Point, NC
Lawrence, KS
South Bend, IN
Clarksville-Hopkinsville, TN-KY
Omaha, NE-IA
Laredo, TX
Kansas City, MO-KS
Rocky Mount, NC
Abilene, TX
Houma, LA
Spokane, WA
Columbia, MO
Galveston-Texas City, TX
Rockford, IL
Jacksonville, NC
Decatur, AL
Racine, WI
Tulsa, OK
Hartford, CT
Bloomington-Normal, IL
Dayton-Springfield, OH
Duluth-Superior, MN-WI
Auburn-Opelika, AL
Yakima, WA
Knoxville, TN
Missoula, MT
Rochester, MN
Dubuque, IA
Roanoke, VA
Springfield, MO
Decatur, IL
Canton-Massillon, OH
42
602
26
126
96
237
47
698
52
49
93
160
54
116
166
94
39
73
332
616
39
382
81
62
74
400
36
43
31
111
124
50
189
13
438
6
12
60
118
2
477
12
3
4
55
3
39
104
6
5
80
245
101
10
116
22
5
15
193
4
7
4
23
43
2
40
890
297
266
118
213
633
255
342
292
356
102
603
206
267
147
134
533
166
323
672
193
317
344
233
182
265
374
189
210
208
261
239
220
288
175
155
236
212
484
182
267
161
149
272
354
200
189
185
180
231
163
316
444
152
373
65
170
181
302
164
151
147
201
265
180
166
180
180
176
175
175
174
174
174
172
172
172
171
171
169
169
167
167
166
165
163
163
160
160
159
159
157
157
156
156
152
151
150
149
2.8
0.9
1.1
1.5
1.5
2.5
1.3
1.4
1.0
1.1
1.6
2.2
1.5
1.0
1.0
1.0
1.8
0.9
1.6
2.7
1.2
2.4
0.8
1.1
1.1
1.8
1.3
1.6
1.2
1.4
2.2
1.8
1.2
4.8
69.2
3.0
14.2
10.7
26.4
5.3
77.7
5.7
5.4
10.2
17.5
5.9
12.6
18.0
10.1
4.2
7.7
35.1
64.4
4.0
39.2
8.3
6.4
7.5
40.3
3.7
4.3
3.1
10.8
12.0
4.8
18.0
TABLE A2: METROPOLITAN LAND VALUE INDICES RANKED BY AVERAGE URBAN LAND VALUE PER ACRE, 2005-2010
Land Values - $000s/Acre
Ratio of
Total Est.
Total Metro
Urban
Central to Urban Land
Urban Area Observed
Central Average - Mile 10 Land Value ($
Rank Metropolitan Area Name
(Sq. Miles) Land Sales Naïve Model Estimated
Estimated
Values
billions)
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
Hickory-Morganton-Lenoir, NC
Biloxi-Gulfport-Pascagoula, MS
Little Rock-North Little Rock, AR
San Antonio, TX
Johnson City-Kingsport-Bristol, TN-VA
Pine Bluff, AR
New London-Norwich, CT-RI
Birmingham, AL
Bismarck, ND
Pittsburgh, PA
Cumberland, MD-WV
St. Cloud, MN
Brazoria, TX
Lake Charles, LA
Waco, TX
Memphis, TN-AR-MS
Athens, GA
Redding, CA
Jonesboro, AR
Dover, DE
Waterloo-Cedar Falls, IA
Springfield, IL
Pittsfield, MA
Columbia, SC
Bangor, ME
Fayetteville, NC
Williamsport, PA
Jackson, MS
Oklahoma City, OK
Sheboygan, WI
Amarillo, TX
Anniston, AL
Greenville-Spartanburg-Anderson, SC
217
172
244
468
273
35
166
421
34
1,003
41
45
110
99
73
423
79
78
41
54
53
94
49
270
38
156
39
192
391
33
85
77
542
88
30
110
348
28
2
31
148
22
240
6
17
62
14
14
173
15
8
8
7
12
11
3
139
5
25
9
43
395
15
27
4
507
184
163
305
224
169
232
299
238
155
433
145
331
225
206
185
328
189
150
194
151
229
430
17
237
339
476
49
191
285
112
173
214
294
204
267
359
263
208
138
178
131
114
496
171
144
101
191
145
156
171
130
200
154
145
172
149
346
215
88
140
248
107
139
125
145
145
149
149
146
146
146
145
144
144
143
143
143
143
142
142
141
141
141
141
140
140
138
138
138
137
137
134
134
133
130
129
128
128
127
1.7
2.2
2.8
1.7
1.5
1.2
1.1
0.8
1.0
3.2
1.3
1.2
1.2
1.6
1.3
1.4
1.4
0.9
2.1
1.3
1.2
1.8
1.3
2.9
2.0
0.6
1.3
1.8
0.7
1.2
1.2
1.3
1.2
20.7
16.3
22.8
43.7
25.5
3.3
15.4
38.7
3.1
91.9
3.7
4.1
10.0
9.0
6.6
38.2
7.2
7.0
3.7
4.8
4.7
8.3
4.3
23.7
3.4
13.4
3.3
16.3
32.5
2.7
7.0
6.3
44.2
TABLE A2: METROPOLITAN LAND VALUE INDICES RANKED BY AVERAGE URBAN LAND VALUE PER ACRE, 2005-2010
Land Values - $000s/Acre
Ratio of
Total Est.
Total Metro
Urban
Central to Urban Land
Urban Area Observed
Central Average - Mile 10 Land Value ($
Rank Metropolitan Area Name
(Sq. Miles) Land Sales Naïve Model Estimated
Estimated
Values
billions)
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
Pocatello, ID
Flint, MI
Yuma, AZ
Des Moines, IA
Lima, OH
Shreveport-Bossier City, LA
Scranton--Wilkes-Barre--Hazleton, PA
Texarkana, TX-Texarkana, AR
Pensacola, FL
Buffalo-Niagara Falls, NY
Wichita, KS
Cedar Rapids, IA
Tuscaloosa, AL
Elmira, NY
Hattiesburg, MS
Eau Claire, WI
Tyler, TX
Brownsville-Harlingen-San Benito, TX
Augusta-Aiken, GA-SC
Lynchburg, VA
Elkhart-Goshen, IN
Goldsboro, NC
Terre Haute, IN
Corvallis, OR
Davenport-Moline-Rock Island, IA-IL
Mobile, AL
Owensboro, KY
State College, PA
Wichita Falls, TX
Jackson, TN
Charlotte-Gastonia-Rock Hill, NC-SC
Syracuse, NY
Kankakee, IL
30
230
54
158
63
179
208
70
246
395
205
62
76
35
39
63
64
125
243
87
86
47
54
33
136
273
39
29
65
44
791
236
35
7
85
12
99
8
50
27
5
102
104
54
33
16
9
5
30
13
52
66
13
14
6
4
3
28
135
1
12
8
4
10
65
9
208
245
215
238
43
88
194
118
317
616
229
151
252
240
143
123
162
263
228
152
328
156
200
436
178
167
41
136
133
40
29
221
144
96
188
147
260
157
216
179
128
146
450
173
114
138
89
108
128
146
97
129
123
95
115
123
132
102
401
106
114
122
110
289
302
122
125
124
124
121
121
119
118
118
117
116
116
114
114
114
113
113
113
113
112
112
112
110
110
109
108
108
108
107
106
106
106
106
106
1.0
1.6
1.2
1.8
1.6
2.0
1.9
1.4
1.2
4.0
1.4
1.0
1.4
0.7
1.1
1.2
1.7
0.7
1.0
1.0
1.0
1.2
1.2
1.6
1.0
3.7
1.3
1.5
1.5
1.2
2.7
3.9
1.9
2.4
18.3
4.3
12.3
4.9
13.6
15.7
5.3
18.4
29.4
15.2
4.5
5.5
2.5
2.8
4.5
4.6
9.0
17.5
6.2
6.1
3.4
3.8
2.3
9.4
18.9
2.7
2.0
4.4
3.0
53.5
15.9
2.4
TABLE A2: METROPOLITAN LAND VALUE INDICES RANKED BY AVERAGE URBAN LAND VALUE PER ACRE, 2005-2010
Land Values - $000s/Acre
Ratio of
Total Est.
Total Metro
Urban
Central to Urban Land
Urban Area Observed
Central Average - Mile 10 Land Value ($
Rank Metropolitan Area Name
(Sq. Miles) Land Sales Naïve Model Estimated
Estimated
Values
billions)
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
Grand Rapids-Muskegon-Holland, MI
Corpus Christi, TX
Fort Wayne, IN
Wausau, WI
Lawton, OK
Sumter, SC
Lafayette, LA
Montgomery, AL
Florence, SC
Macon, GA
Bryan-College Station, TX
Lansing-East Lansing, MI
Albany-Schenectady-Troy, NY
Dothan, AL
Beaumont-Port Arthur, TX
Kalamazoo-Battle Creek, MI
Alexandria, LA
Binghamton, NY
Grand Forks, ND-MN
Johnstown, PA
Fitchburg-Leominster, MA
Bloomington, IN
Victoria, TX
Killeen-Temple, TX
Lubbock, TX
Appleton-Oshkosh-Neenah, WI
Mansfield, OH
San Angelo, TX
Benton Harbor, MI
Toledo, OH
Erie, PA
Florence, AL
Odessa-Midland, TX
435
139
182
38
55
45
165
133
66
172
49
156
355
93
165
191
58
81
29
77
61
43
51
111
82
109
74
46
91
203
97
52
99
121
74
39
16
20
10
15
33
12
20
34
40
120
14
60
31
4
16
3
5
8
3
7
32
45
79
3
2
12
107
29
4
39
243
179
269
104
177
99
118
252
171
174
165
138
158
133
140
144
55
188
63
469
44
54
70
140
142
105
125
109
110
172
104
102
129
168
117
155
80
74
98
124
200
147
117
95
179
177
95
107
141
93
134
120
134
109
86
91
81
116
68
81
70
110
136
92
88
80
105
105
104
103
103
103
102
101
100
100
99
97
95
95
94
93
92
91
90
90
89
89
87
87
87
84
83
83
81
81
80
79
77
1.7
1.0
1.3
1.0
0.9
1.0
1.3
2.0
1.8
1.0
1.8
2.3
2.4
0.9
1.1
1.6
1.0
2.0
1.4
1.7
1.4
1.1
1.1
1.0
1.2
0.9
0.9
0.9
1.2
1.4
1.3
1.5
0.9
29.3
9.3
12.1
2.5
3.7
2.9
10.8
8.6
4.3
11.0
3.1
9.7
21.7
5.6
9.9
11.3
3.4
4.7
1.7
4.4
3.5
2.4
2.9
6.2
4.5
5.9
3.9
2.4
4.8
10.5
5.0
2.6
4.9
TABLE A2: METROPOLITAN LAND VALUE INDICES RANKED BY AVERAGE URBAN LAND VALUE PER ACRE, 2005-2010
Land Values - $000s/Acre
Ratio of
Total Est.
Total Metro
Urban
Central to Urban Land
Urban Area Observed
Central Average - Mile 10 Land Value ($
Rank Metropolitan Area Name
(Sq. Miles) Land Sales Naïve Model Estimated
Estimated
Values
billions)
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
Sherman-Denison, TX
Wheeling, WV-OH
Evansville-Henderson, IN-KY
Peoria-Pekin, IL
Topeka, KS
Rochester, NY
Lewiston-Auburn, ME
Parkersburg-Marietta, WV-OH
Fort Smith, AR-OK
Gadsden, AL
Great Falls, MT
St. Joseph, MO
Sharon, PA
Sioux City, IA-NE
Altoona, PA
Youngstown-Warren, OH
Muncie, IN
Rapid City, SD
Longview-Marshall, TX
Glens Falls, NY
Utica-Rome, NY
Danville, VA
Pueblo, CO
Enid, OK
Saginaw-Bay City-Midland, MI
Jackson, MI
Jamestown, NY
34
58
113
144
70
388
28
51
71
61
29
40
33
50
45
258
44
33
84
33
89
34
54
24
146
57
46
19
4
33
25
7
110
3
3
18
6
1
12
9
17
8
49
5
7
14
21
15
5
18
2
41
8
10
40
27
74
94
212
632
46
65
88
71
134
191
45
162
76
94
26
79
289
46
103
46
71
75
92
49
43
66
94
110
124
56
301
76
61
69
78
53
36
80
56
64
64
65
63
84
62
58
47
60
43
61
48
39
77
76
76
76
75
75
75
75
73
72
69
67
66
62
61
60
57
57
56
56
55
52
51
46
45
38
30
1.0
1.5
1.9
2.0
0.6
4.6
1.2
1.0
0.9
1.0
0.8
0.6
1.3
1.0
1.3
0.8
1.1
1.5
1.5
1.6
1.2
0.8
1.9
0.9
1.2
1.4
1.0
1.7
2.9
5.5
7.0
3.4
18.6
1.3
2.5
3.3
2.8
1.3
1.7
1.4
2.0
1.8
9.9
1.6
1.2
3.0
1.2
3.1
1.1
1.8
0.7
4.3
1.4
0.9
TABLE A2: METROPOLITAN LAND VALUE INDICES RANKED BY AVERAGE URBAN LAND VALUE PER ACRE, 2005-2010
Land Values - $000s/Acre
Ratio of
Total Est.
Total Metro
Urban
Central to Urban Land
Urban Area Observed
Central Average - Mile 10 Land Value ($
Rank Metropolitan Area Name
(Sq. Miles) Land Sales Naïve Model Estimated
Estimated
Values
billions)
N/A
N/A
N/A
N/A
N/A
N/A
N/A
Metropolitan Areas with no land sale observations:
Albany, GA
Casper, WY
Charleston, WV
Cheyenne, WY
Chico-Paradise, CA
Huntington-Ashland, WV-KY-OH
Kokomo, IN
66
26
115
34
89
115
41
0
0
0
0
0
0
0
-
194
96
268
114
247
269
142
175
111
187
124
189
200
143
1.4
1.1
1.5
1.2
1.5
1.5
1.2
7.4
1.9
13.8
2.7
10.8
14.7
3.7
TABLE A3: COMPARISON OF TRANSACTION-BASED AND RESIDUAL-BASED ESTIMATES OF
LAND VALUES, 2005-2010
Land Values - $000s/Acre
Davis and Palumbo (2008)
Residual-Based Estimate for COMPS Transaction-Based
Metropolitan Area Name
Residential Land
Estimates for All Urban Land
San Jose CA
666.1
1,418.7
San Francisco CA
844.5
1,333.8
New York NY
344.8
1,270.8
Orange County CA
608.4
1,248.7
Los Angeles-Long Beach CA
438.2
1,152.7
Miami FL
241.2
943.6
Oakland CA
474.6
864.6
Seattle-Bellevue-Everett WA
251.3
758.7
San Diego CA
413.5
753.5
Washington DC-MD-VA-WV
311.7
737.1
Phoenix-Mesa AZ
100.0
532.7
Portland-Vancouver OR-WA
183.1
512.4
Boston MA-NH
340.2
505.9
Chicago IL
107.5
494.3
Sacramento CA
160.0
432.7
Baltimore MD
241.3
430.6
Tampa-St. Petersburg-Clearwater FL
67.5
421.1
Riverside-San Bernardino CA
131.7
397.5
Philadelphia PA-NJ
122.9
382.8
Salt Lake City-Ogden UT
87.3
371.4
Minneapolis-St. Paul MN-WI
56.7
368.2
Denver CO
91.1
349.9
Atlanta GA
26.5
315.4
Norfolk-Virginia Beach-Newport News VA-NC
155.2
307.0
Providence-Fall River-Warwick RI-MA
175.5
302.8
Cleveland-Lorain-Elyria OH
32.6
251.6
Houston TX
31.3
248.1
Cincinnati OH-KY-IN
30.3
248.0
Columbus OH
44.2
244.1
Detroit MI
14.9
240.3
Dallas TX
55.5
238.2
New Orleans LA
72.3
220.7
Milwaukee-Waukesha WI
68.6
219.5
Indianapolis IN
10.6
202.1
Fort Worth-Arlington TX
25.9
187.4
St. Louis MO-IL
20.2
180.5
Kansas City MO-KS
17.7
173.8
Hartford CT
121.0
163.3
San Antonio TX
17.5
146.1
Birmingham AL
35.8
143.5
Pittsburgh PA
10.5
143.1
Memphis TN-AR-MS
14.2
141.3
Oklahoma City OK
12.2
129.7
Buffalo-Niagara Falls NY
24.6
116.2
Charlotte-Gastonia-Rock Hill NC-SC
88.1
105.7
Rochester NY
16.7
74.9
Davis and Palumbo (2008) estimates are geometric means of quarterly values from 2005q1 to 2010q4. MSAs
were matched based on city names, with the exception of SANTAANA, which was matched to Orange County,
CA. Davis and Palumbo estimates were downloaded from the Lincoln Land Institute website February 2017 at
http://datatoolkits.lincolninst.edu/subcenters/land-values/metro-area-land-prices.asp.
xviii
Figure A.1: Geographical Distribution of Land Sales in Four Consolidated MSAs
(a) New York Northern New Jersey, Long Island,
NY-NJ-CT-PA
(b) Los Angeles-Riverside-Orange County CA
(c) Chicago-Gary-Kenosha IL-IN-WI
(d) Houston-Galveston-Brazoria TX
The gray dots represent land sales. The black stars represent city centers.
xix
Figure A.2: Comparison of Transactions-Based Index to Residual-Based Index
(b) Within-City Time Series Variation
0
Transactions-Estimated Value per Acre ($000s)
500
1000
Transactions-Estimated Within-City Price Variation (%)
20
40
60
80
100
1500
(a) Average Land Values
0
200
400
600
800
20
Davis and Palumbo Value per Acre ($000s)
Estimated Values
40
60
80
Davis and Palumbo Within-City Price Variation (%)
45-degree line
Estimated Values
xx
45-degree line
100