Skyscraper Height

Jason Barr

Rutgers University, Newark

jmbarr@rutgers.edu
Rutgers University Newark Working Paper #2008-002
Abstract
This paper investigates the determinants of skyscraper height. First
a simple model is provided where potential developers desire not only
prots but also status, as measured by their rank in the height hi-
erarchy. The optimal height in equilibrium is a function of the cost
and benets of building as well as the height of surrounding buildings.
Using data from New York City, I empirically estimate skyscraper
height over the 20th century. The results show that the quest for
status has increased building height by about 15 oors above the non-
status prot maximizing height. In addition, I provide estimates of
which buildings are too tall and by how many oors.
JEL Classication: D24, D44, N62, R33
Key words: Skyscrapers, building height, status, New York City

I would like to thank Alexander Peterhansl, Howard Bodenhorn, Sara Markowitz and
seminar participants at Lafayette College for their helpful comments. I would like to
acknowledge the New York City Hall Library, the New York City Department of City
Planning and the Real Estate Board of New York for the provision of data. This work was
partially funded from a Rutgers University, Newark Research Council Grant. Any errors
are mine.
1
1 Introduction
Skyscrapers are not simply tall buildings. They are symbols and works of
art. Collectively they generate a separate entitythe skylinewhich has its
own symbolic and aesthetic importance.
Despite the initial fears that the attacks of September 11, 2001 would cur-
tail construction, skyscrapers continue to be built in large numbers around
the globe (Economist, 2006). The current cycle of therace to the sky is in
full swing (Ramstack, 2007). The Burj Dubai, still under construction as of
February 2008, may top out at nearly half a mile tall. This building will re-
place the Taipei 101 as the world record holder. With increased globalization
and international development, cities world over seek to develop skyscrapers
as a way to announce their newly created economic strength and to put their
cities on the map (Gluckman, 2003).
Skyscrapers are used to advertise and signal economic strength for their
builders, be they speculative developers, major international corporations or
government entities. As such, they not only provide prots but also status.
For this reason, height has a strategic component. If a developer prefers to
have his building stand out in the skyline or to be taller than others, then
he must consider the height of surrounding buildings.
Despite their importance for both local and national economies, skyscrap-
ers have drawn little attention from economists. Beyond the many journal-
istic and popular accounts, the last time economists discussed skyscrapers
in any detail was during the building boom of the late 1920s. During that
time, especially in New York City, the debate centered around whether these
buildings were somehow freak buildings, built not on sound economic prin-
ciples, but rather as expressions of personal ego or for their ability to advertise
(Clark and Kingston, 1930).
Buildings such as the Bank of Manhattan (now 40 Wall Street) (1929),
the Chrysler (1929) and the Empire State (1930) illustrate the symbolic and
strategic importance of skyscrapers. At the time of their completions each
was the worlds tallest building, and the developers were explicit about their
intention to be the world record holder, despite each being taller than the
prot maximizing height (Tauranac, 1995).
Skyscraper height can be thought of as a good that brings value to both
builders and height consumers, who desire dramatic views, and, as such,
there are gains to trade in the height market. The developer must make both
an economic and strategic decision about how tall to build; this height is de-
2
Figure 1: New York City Skyline, lower Manhttan.
termined both by the builders desire for prots and status and consumers
utility derived from height, which also provides status (such as executives
placing their oces on top oors, and the wealthy living in penthouse apart-
ments) as well as the enjoyment of the fantastic views (of the skyline itself).
Figure 1 presents a photograph of part of the New York City skyline. As
can be seen from the picture, collectively, the skyline is an entity unto itself
due to the density and height of the buildings. Within this skyline there is
a great deal of variation in building height. Not every building can be the
tallest and not every builder cares to build the tallest. Rather we can infer
that building height is a function of economics, land use regulations and also
the desire for status.
One also notices that within the skyline there are distinct waves of
building heights, with height rising toward the center. These waves re-
ect the endogenous relationship between strategic height, land values and
agglomeration economies. Corporations need to be near each other to lower
their business costs and increase demand, yet they also desire to stand out in
the skyline. Being close is valuable, which is reected in property values in
the center; large land costs, in turn, drives developers to build even higher if
they are to get a return on their investment, as well as have their buildings
stand out.
This paper is an investigation into the determinants of skyscraper height.
To the best of my knowledge, it is the rst work that investigates the theo-
3
retical and empirical determinants of height for any city skyline over such a
long time period. Here, I use the example of New York City, since it is one
of the most important and active skyscraper cities in the world. I investigate
the relative eects of status, economics and regulation using a data set of
458 skyscrapers completed in Manhattan from 1895 to 2004, which includes
a mix of residential, oce and other building types.
First, I provide a simple model of building height. The model has two
parts. First developers bid for the right to develop a plot of land. Next the
winning developer chooses a height that will maximize his utility, which is a
combination of the economic returns from the project plus a benet derived
from the relative ranking of the buildings height. That is, the developer also
includes his desire for status when choosing a height.
To simplify matters, I assume that the land market for developable plots
is a type of rst-price sealed-bid auction, where developers submit bids, and
the highest bidder wins the right to develop the land and pays his bid. This
is a relatively simple variation of the standard zero-prot condition for land
allocation. Typical models assume that land is allocated to its most valu-
able use, which is based on, in part, transportation costs and agglomeration
economies (see DiPasquale and Wheaton (1995), for example). While these
models can demonstrate what factors generate land use, they do not gener-
ally demonstrate who gets to build on the land. By introducing heterogeneity
in builder preferences, the model gives an equilibrium for a type of status
game, where the developer who gains access to the land has the largest rel-
ative preference for status among the bidders.
Here skyscraper status is meant to encompass a few dierent factors.
First, major corporations seek status to advertize their corporations. Also
speculative developers desire status because, presumably, this status will
increase rents or just bring more respect to the developers themselves, who
often have enormous egos (Helsley and Strange, 2007; Betsky, 2002).
In dense real estate markets, developers are forced to act in secrecy, since
any information about their intentions can lead to hold outs (see Strange
(1995), for example) and the speculative bidding up of land prices before
their nal use is determined. Often shell corporations do the bidding on
behalf of the developer (see Samuels (1997) for example). As such, it is a
reasonable assumption that builders preferences for a development project
are unknown by the others during the time that the plot is on the market.
Next, I use the optimal height equation as a guide to estimate the ef-
fects of economics, land use regulations and status on skyscraper height in
4
New York City. Since it is virtually impossible to collect data on actual con-
struction costs and income ows, I use several economic variables that can
measure the costs and benets of construction. On the costs side, I show
that building materials costs and interests rates negatively impact building
height. On the benets side, I show that population, oce employment and
land value growth are positively related to height. In addition, I am able to
quantify the eects of zoning regulations on height by showing how they alter
the incentives to build taller. Furthermore, some skyscraper historians have
argued that Manhattans bedrock formation has helped to contribute to New
York Citys skyline. I nd mixed support for this theory; specically I nd
a small negative relationship between the depth to bedrock and a buildings
height in midtown and no direct eect for buildings downtown (though a
larger indirect negative eect for all buildings downtown).
Lastly, I am able to measure the desire for status for building height.
Corporations who build their own headquarters add only a modest amount
of height; on average adding about two additional oors. Using the lagged
average height of all completed skyscrapers, I am able to measure the impor-
tance of standing out in the skyline. I estimate that builders responded
by adding about one foot to their own buildings for each one foot growth
in the skyline itself. By the end of the 20th century, the eect of this was
that builders were adding about 15 extra oors, on average, above the prot
maximizing level so their buildings can be seen.
The rest of this papers is as follows. The next section gives a review of the
relevant literature. Then, section 3 presents the land allocation game and
the optimal height decision. Next, section 4 discusses the functional form
for skyscraper construction; this function is used as a guide for empirical
estimation. Then section 5 discusses the relevant issues for New York City.
Discussion of the data and the empirical results follow in section 6. Section
7 uses the estimates to make some predictions about which buildings are
too tall as compared to the estimated optimal economic height; as well,
time series for the optimal economic height and status height are given
for the 20th century. Section 8 oers some concluding remarks. Finally two
appendices provide additional information.
5
2 Related Literature
Despite the attention given to skyscrapers by the popular media, there have
been only very few recent studies directly addressing their economics. The
last time that economists have looked at skyscrapers in any detail was during
the great building boom of late 1920s. Then the debate focused on whether
tall buildings like the Chrysler and Empire State were built to be monuments
rather than money makers.
Perhaps the most cited work from that time is that of Clark and Kingston
(1930), who estimate the costs and income ows from a hypothetical building
of various heights. They placed their building across the street from Grand
Central Station, the center of the midtown business district; using land prices,
construction costs and rent data from 1929, they conclude that a 63 story
building would provide the highest return.
1
Their aim was to demonstrate
that skyscrapers, at their heart, were economically rational investments.
2
In
fact, they also estimated that a 100 story building would provide a net return
of 7.08%.
More recently, two papers deal directly with the economics of skyscrap-
ers.
3
Barr (2007) looks at the market for height in Manhattan over the
period 1895 to 2004 by investigating the time series of the number of sky-
scraper completions and the average height of these completions. The paper
nds that though the costs and benets that have determined the decision
about whether to build and how tall to build have varied over the course
the twentieth century, there has been no fundamental change in skyscraper
building patterns over the 20th century. Though the 1920s represented an
1
Their ctional plot size of 81,000 square feet is in the 87th percentile for plot size in
my data set; its in the 91st percentile for plot size for buildings completed before 1950.
To give a sense of comparison, the Empire State Building has a plot area of 91,351 square
feet (with 102 oors), and the Chrysler Buildings plot area is 37,525 square feet (with 77
oors).
2
Their buildings estimated return on investment was rational given 1929 rent values.
However, if Clark and Kingston had forecasted that rents would soon turn down from their
1929 peaks (and vacancy rates were to go up), 63 stories would most likely not have been
the optimal height. As is famously noted, after the opening of the Empire State Building
in 1930, it soon become known as the Empty State Building because of the Great
Depression (Tauranac, 1995).
3
There are also two related strands of real estate literature that I do not address here:
real estate cycles (such as Wheaton (1999) and Case and Shiller (1989)), and the decision
about when to build (such as Titman (1985) and Bar-Ilan and Strange (1996)).
6
aberration in terms of the number of skyscraper completions at its peak, the
forces driving the average heights of buildings, however, have not changed
signicantly over the 20th century. Rather the average height is based on the
supply and demand for this height, which is determined by factors related to
both the New York City and national economies, regulations on land usage,
and taxation. The work here is dierent in that I look directly at the de-
terminants of building height, at the building level, asking what fraction of
building height can be accounted for by economics, land use regulation and
the quest for status.
Another paper is by Helsley and Strange (2007). They investigate a two-
person game, where each player aims to building the worlds tallest building.
They demonstrate that when players value being the tallest for its own sake
(as a desire for status), the contest can dissipate prots from tall buildings.
My study is related to that of Helsley and Strange in that I investigate the
degree to which competition among builders can aect the skyline, as well as
cause non-prot maximizing building. But unlike their paper, my objective
is broader, investigating the determinants of skyscraper height within a city
and over time.
The paper here also draws from recent work on the economics of status.
Most notable is the paper of Hopkins and Kornienko (2004), who consider
a consumption game that includes status. Their paper assumes that status
enters into agents utility function by way of a status ranking function (cdf),
which determines their relative position in the consumption hierarchy. Their
model provides a symmetric equilibrium that maps income to consumption.
They nd that, in equilibrium, spending on the status good increases relative
to the nonstatus good, but everyones ranking is determined simply by their
location in the income distribution. This paper is similar in that I assume
builders value status for its own sake, and that the height decision has a
strategic component, especially in the bidding process. Developers look at
the mean heights of completed buildings when deciding how tall to build; as
the means rise so does the extra height.
3 The Model
Here I provide a simple model for the optimal height decision. The model
is a type of auction game. I assume that each plot of land is sold to the
highest bidder, who pays his bid to the seller. Each bidders valuation of
7
the plot comes from the prot and relative status that can be earned from
developing the land; this valuation is private information because it is a
function of an i.i.d. private signal about how much the developer values
status. The winning bidder then chooses a height that maximizes his utility.
I also assume that each time a plot comes up there is a new auction and
a new realization of N bidders.
4
The equilibrium is symmetric, with each
bidder using the same bid function.
3.1 The Height Decision
First, we begin with the optimal height decision, then show that, given this
height decision by each potential builder, there is an equilibrium in the auc-
tion game. Here agent i, i = 1, ..., N, has a utility function given by
u
i
(h) = (h) +
i
F (h) l
i
, (1)
where (h) is the developers prot that can be earned from building a
skyscraper of height h. For simplicity, assume that lots are xed in size
and normalized to one (we relax this assumption below). Assume that the
prot function is continuous in h, concave, single-peaked, and for h
h
0,

h
i
,
(h) 0 and (0) =

= 0. The prot function represents the net value

of the building less the construction costs. l
i
is the cost of the plot of land.
Assume that all potential developers know the prot function and that it is
the same for all developers.
5
The value of the building can be determined
in part from site-specic factors, such as its access to public transportation,
zoning regulations, and proximity to the business district core as well as
economy-wide or regional factors such as interest rates and building costs.

i
is the developers private value that is placed on status or his ranking in
the height hierarchy; it is i.i.d. across agents and the cumulative distribution
function, G() , has a closed, bounded and continuous support,

0,

, with
G(0) = 0 and G

= 1. To simplify the analysis, assume that is small

enough if that if an agent with value was to bid and win, he would still
4
Though not addressed here, both the number of bidders as well as the number of
auctions could be made endogenous.
5
Also, I make the simplication that expectations about future income streams dont
play a role. Clearly whether expectations are myopic or rational, for example, can impact
building height, but this is not investigated here.
8
have utility at or above his reservation level. In addition, all agents know
G() .
F (h) is a continuous, strictly monotonic ranking function (or cumulative
distribution function) such that there are values h, h, with 0 h < h, with
F (h) = 0 for 0 h h, and F

h

= 1 for h h. That is, if F (h) is

the rank of a developers building in terms of its height, h is the minimum
size necessary to achieve any status all. h is the current record-holder for
the tallest building in the city or region. These minimum and maximum
values can change over time, but builders take then as given when deciding
on a possible height for their building. In short
i
F (h) is the contribution
of status to a builders utility; F (h) is his possible rank, given the current
skyline, when deciding how tall to build. Building a skyscraper is no small
feat, and not all builders have the skills, knowledge or access to capital;
as such, status is only conferred upon those can succeed in constructing a
building of certain height.
Lets say that agent i with status parameter
i
wins the right the develop
the land. He would then choose a height, h

, such that h

= arg max
hR
+
u
i
(h) :
u
0
i
(h

) =
0
(h

) +
i
f (h

) = 0,
where f (h) = F
0
(h) > 0 for h

h, h

. (For the remainder of this section

the subscripts are dropped to simplify notation.)
In the case where status does not matter, the developer would simply
choose an optimal height that maximized the net return from height. We
can dene this optimal height as the competitive outcome, which is given
as h
c
= arg max
hR
+
(h) . Given this utility function and maximization
problem it is straightforward to show that (1) the optimal height with status
exists and is unique for each ; (2) the height chosen by the developer is
larger than if status were not relevant; and (3) that the optimal skyscraper
height is monotonically increasing in . These are presented formally, and
proofs are given in Appendix A.
Lemma 1 For

0,

, there is a unique value of h, h

, such that h

=
arg max
hR
+
u
i
(h) .
Lemma 2 h

> h
c
for

0,

; h

= h
c
for = 0.
Lemma 3 h

is strictly increasing with .

9
In summary, this section has shown that for a given developer who is
going to develop a plot of land, he will build taller than the competitive
height due to the desire to have a place in the height hierarchy. Furthermore,
this height is increasing in the value he places on status.
3.2 The Land Allocation Game
Above, I discussed the height decision of a builder, conditional on having the
right to develop a plot of land. Now I demonstrate how land is allocated.
Building on the standard assumption in urban economics that land is allo-
cated to its most valuable use, here, it is shown that if developers value the
land for both prots and status, and that land is auctioned o to the high-
est bidder, then there exists a symmetric equilibrium, where each agents bid
is a function of his own private valuation, which is a function of the common,
publicly known prot and the private, randomly-determined value for status.
Given that potential buyers know there is private variation in the valu-
ation, they need to strategically consider their bids. No rational developer
would bid more than the plot is worth to him (assuming no purely spec-
ulative land purchases). Any bid below his maximum value introduces a
tradeo: an increase in the bid will increase the probability of winning, but
will also reduce the possible gains from the project. This introduces the
familiar rst-price sealed-bid auction mechanism for allocating the plot.
Assume a common reservation value, r 0, which is the lowest value of
utility a developer is willing to accept from a skyscraper project. Let l

i
be
developer i
0
s land valuation from choosing an optimal height:
l

i
= (h

) +
i
F (h

) r.
Further, denote l
c
= (h
c
) r as the value that developers would place on
the land if status were not an issue. Without status, we could simply assume
that the plot would sell for l
c
, since in a competitive market land values
would provide the builder with zero economic utility (or prots, if we assume
r = 0). Further assume that F (h) is common knowledge, the builders know
there are N builders interested in the property and that they all know the
distribution of .
It is straightforward to show that l
c
is the lower bound on income that
any seller would receive for the plot. Further, it can be shown that land
values are strictly rising in . I assume for the sake of simplicity that has
10
a uniform distribution with support

0,

, where, again,

is assumed to be
not so large that a developer with =

who takes ownership of the plot
would still have a utility at least as large as r. The properties of the land
values are stated formally and their proofs are given in Appendix A.
Lemma 4 l

is monotonically increasing in ; l

= l
c
when = 0.
Lemma 5 Given the monotonicity of l

, the minimum and maximum land

values for a plot is l
c
and

l

= (h

) + F (h

) , respectively.
Lemma 6 Given that U

0,

, the probability distribution function for

land valuations is given by
k (l

) =

1
F(h

)
, l

l
c
,

0, otherwise

,
with a cdf of
K (l

) =
l

l
c
F (h

)
, l

l
c
,

.
It is straightforward to show that there exists a symmetric equilibrium,
where each agent uses the same bid function = (l

) .
Proposition 1 Given each agents land valuation function, and N bid-
ders, there exists a unique, symmetric equilibrium of the land auction game
such that (l

) =
(N1)l

+l
c
N
.
Notice that (l

) =
(N1)l

+l
c
N
is a weighted average of the land values
with and without status. For example, if N = 2, then (l

) =
1
2
(l

+ l
c
) ; as
N , (l

) l

.
4 Functional Form and Optimal Height
For the developer with the highest valuation, utility is given by
u(h) = V (a, h; ) C (h, a; ) + F (h) l,
where h is the building height (in feet), a is the land area of the plot (in
square feet). V is the expected present discounted value of the net rent
11
ows from a newly constructed building. is the contribution to the value
from city-wide and site-specic factors. City-wide factors might include city
employment and population; while site-specic factors might include zoning
regulations, which might limit use or height, and neighborhood eects (such
as localization economies).
Again, F (h) is the ranking function. V is a function of height since there
is demand for height fromheight consumers who enjoy the great views from
up above, as well as those to whom being on the top oors is a demonstration
of conspicuous consumption and/or status. C (h, a; ) is the cost function,
which is determined by the plot area, the height, and , which measures the
factors that can aect the costs of construction, such as the depth to bedrock,
the regularity or shape of the plot, input costs and interest rates.
To make matters more concrete lets assume that total building value, V,
is per oor value times the height of the building, that is V = vh, where
v =

a +
1
2
h

. a is the per oor value for the site (assuming that a builder
builds on the whole plot) and
1
2
h (whose coecient is normalized to one-half)
is the additional per oor value that comes from the increased height due to
the extra value for height by consumers.
6
In terms of building costs, assume that height has increasing marginal
costs due to the fact that as buildings go taller there are additional ex-
penses, including increased foundation preparation, wind bracing materials,
more elevator shafts, and larger heating and cooling systems (Clark and
Kingston, 1930; Shabbagh, 1989). For simplicity, assume a cost per oor of
c =

a +

2
h

, where a is the cost per oor and /2 is the marginal height

cost for each additional oor. Total construction cost is given by C = ch.
Further assume that the ranking function is a uniform cumulative dis-
tribution function, F (h) =
hh
hh
, where h is the record, h is the minimum
building height to have any status. Putting the value and the cost functions
together, and rearranging terms gives
u(h) =

( ) a +

h h

h
1
2
(1 ) h
2

h
h h
l. (2)
Assume that (1 ) > 0 and

( ) a +

hh

is large relative to
1
2
( 1).
The optimal height with status is derived from the rst order condition for
6
Note that h could be converted to the number of oors, where, for example floors =
12h, if we assume that oors are 12 feet high.
12
equation (2). Denote 1/ (1 ), and note that for a uniform distri-
bution, the mean ranking is =

h h

/2. The optimal height is given

by
h

( ) a +

2

. (3)
In summary, optimal height is given by the prot from development plus
the interaction of the average building height and desire for status. Further
note that h
c
= ( ) a, which means that h

= h
c
+

2
, where

2
is the
status component of height.
5 New York City
Given the discussion of skyscraper height above, we now turn to empirically
investigate the determinants of skyscraper height as given by the model,
using New York as an example. The aim is to estimate the role of economics
versus status factors. In this section, I rst present a brief history of New
York and discuss the relevant issues for estimating skyscraper height. Then
I discuss the empirical model, the data and the results.
After the completion of the Erie Canal in 1825, NewYork City became the
nations most important city, and the center of nance and commerce.
7
The
citys economic and population growth created a great demand for land on
which to house both residents and businesses. At its widest point, Manhattan
island is about three miles wide, and about thirteen miles long, comprising
a total of about 23 square miles. The citys initial development was on the
lower, southern tip of the island; over the 19th century development generally
proceeded up the island, northward.
In 1811, in an eort to rationalize its street pattern, the city implemented
its now-famous gridplan. The plan standardized street patterns and lot sizes.
Standard blocks measured 200 feet wide (north-south) and ranged from 400
to 920 feet long (east-west). Lots were generally 25 feet wide and 100 feet
long. These small lot sizes were deemed, at the time, suitable for individual
homes or shops. Because lot sizes were relatively small, as the city became
built up, acquiring larger lots for tall buildings became more dicult. By
7
Up until 1874, New York City was just the island of Manhattan, when it annexed
parts of the Bronx. In 1898, New York City merged with the city of Brooklyn and other
surrounding towns to become what is today the ve boroughs of the city.
13
the late-19th century, this created an incentive for developers to use each lot
more intensely by building higher due to the articially produced scarcity of
large plots (Willis, 1995).
Building Technology Up until the mid part of the 19th century, build-
ing height was limited primarily by technological considerations. A buildings
load bearing was done by masonry walls. To build taller required ever thicker
walls, which then cut into the usable space of the plot. Secondly, without
elevators, people were forced to ascend to the upper oors via the stairwells,
which they were reluctant to climb beyond ve or six stories. As a result,
the top oors generally housed the least valuable economic activities (Shultz
and Simmons, 1959).
Over the course of the 19th century, a series of innovations eliminated the
technological barriers to height. First, was the development of steel beams,
which removed the need for load-bearing masonry walls. Instead, a steel
skeleton cage could be constructed, with a thin brick or stone facade.
In addition, new methods of transporting people upwards had to be de-
veloped. The original elevators were powered by steam, and they were soon
replaced by hydraulic lifts. But it was the invention of electric elevators in
the 1880s that eliminated passenger height constraints altogether. Perhaps
most important was the creation by Elisha Otis in 1853 of the safety break,
which eliminated the fear that the elevator might violently plummet to the
ground (Landau and Condit, 1996).
Other important innovations include the development of caissons for dig-
ging through lower Manhattans quicksand to reach bedrock. Engineers also
had to learn how to brace skyscrapers against the erce winds. New building
machines, such as cranes and derricks, had to be built. In addition, new
methods of heating, cooling, lighting and plumbing were created (Landau
and Condit, 1996). In sum, by around 1890, technological issues were no
longer the major determinant of building height; rather skyscraper height
was primarily one of economics and status.
Bedrock As Landau and Condit (1996) write, In theory, the geology
of Manhattan Island is ideal for skyscrapers. The islands sunken, glaciated
bedrock system, made up of metamorphic rock that constitutes the Manhat-
tan prong of the New England Province, is for the most part good bearing
rock (p. 24). On the southern tip bedrock lies below a bed of quicksand and
14
clay, and is, on average, 208 feet below street level (with a standard deviation
of 104 feet), based on the sample here. In midtown, the bedrock lies quite
close to the surface, and on some parts of the island one can see outcroppings,
such as in Central Park (the midtown mean depth is 56 feet, with a standard
deviation of 34 feet). Between downtown and midtown, however, there is a
steep drop in the bedrock levels (such as in Greenwich Village).
Though the depth to this bedrock varies greatly from north to south,
the placement of this rock relatively near the surface, it has been argued,
has aected, if not determined, the location of skyscrapers throughout out
the island (Landau and Condit,1996). Here, I implicitly test this theory by
including the depth to bedrock for each plot. If bedrock was an important
determinant of height, I would expect to see a negative relationship between
the two, since the depth needed to dig to bedrock would presumably aect
the cost of building and therefore the optimal height.
Zoning By the turn of the 20th century, many New Yorkers were con-
cerned that the unregulated growth of skyscrapers, and the metropolis in
general, were causing several urban problems, such as excessive congestion
and the casting of shadows onto existing structures. As a response, in 1916,
New York City approved a comprehensive zoning plan, which was the rst
of its kind in the nation. The plan established three types of use zones or
districtsresidential, business and unrestrictedto promote the separation of
these economic activities.
In regard to building height, the zoning plan created dierent height dis-
tricts, which did not limit height per se, but rather established rules governing
how tall a building could go before it had to be setback. For example, parts
of midtown were designated as a two times district; a building could rise
to a height of two times the width of the street before it had to set back.
8
In addition there were no height restrictions on any portion of the building
that occupied 25% or less of the plot area. These regulations promoted the
so-called wedding cake style of architecture. For Manhattan, the setback
multiples ranged from 1.25 to 2.5, and the number was positively related to
the density that already existed as of 1916.
8
In a two times district no building shall be erected to a height in excess of twice the
width of the street, but for each one foot that the building or a portion of it sets back
from the street line four feet shall be added to the height limit of such building or such
portion thereof (Building Zoning Resolution, 1916, Section 8(d)).
15
In 1961, New York City enacted a new comprehensive zoning plan, which
was designed to correct some of the perceived mistakes of the original plan,
as well as to regulate new developments, such as the rise of the automobile
throughout the rst half of the 20th century.
9
The 1961 plan, like its prede-
cessor, did not limit height per se, but rather placed limits on the so-called
oor area ratio (FAR).
The maximum allowable FAR is a limit on the total buildable space, and
is given as a multiple of the plot area. AFARof 10, for example, means that a
developer can build 10,000 square feet of usable space for every 100 square feet
of plot area. In the downtown and midtown oce districts, maximum FARs
were set at 15. In addition, to promote the development of public amenities,
such as plazas, the zoning regulations allowed for a 20% FAR bonus if the
developer provided an amenity. Starting in the late-1960s, negotiated FAR
bonuses also became common, as developers sought additional bonuses by
negotiating with the mayor and the Department of City Planning to provide
additional amenities, such as rehabilitated subway stations or extra park
space.
By creating a maximum FAR for each building, the new zoning code
promoted the market for air rights. Owners of buildings, such as landmarks
and theaters, that needed funds, could sell the unused oor areas that, in
theory, existed above the building. Purchasers of the air rights could then
gain additional buildable space for a nearby development. Air rights were
initially instituted to protect landmarks such as Grand Central Station (after
the demolition of the original Pennsylvania Railroad Station in 1963). Today,
however, the air rights market is not limited to just a select few buildings,
but rather is applicable to many buildings in Manhattan (with the caveat
that the rights be sold to adjacent or nearby properties).
Status and Ego After the Civil War, the American economy became
increasingly national in scope due to the reduction of production, transporta-
tion and communication costs. The new industrial economy required a class
of oce workers to organize and process the large amounts of information
(Chandler, 1977). Because New York was the center of much of this economic
9
One intention of 1961 plan was to reduce the maximum allowable population den-
sity. Under the 1916 zoning rules, the city would have been able to house a maximum
population of 55.6 million. The 1961 zoning code was designed to house a maximum of
12.3 million (Bennet, 1960). In 2006, the population of New York City was 8.21 million
(http://www.census.gov).
16
activity, major corporations established their headquarters there. With the
vast amounts of wealth being generated by the new economic activity, cor-
porations sought to project this wealth onto the skyline itself.
Corporate executives were keenly aware of the role that skyscrapers could
play for them and their companys image. Newspaper publishing, congre-
gated downtown, near City Hall, was perhaps the rst industry, in the 1870s,
to engage in the strategic use of height. As Wallace (2006) writes, In early
newspaper buildings, architecture reasserted itself in monumental tributes to
the power of the printing press and its most assertive masters, New York
City newspaper publishers. Newspaper buildings attempted to communicate
the supremacy of the press generally and their own paper specically (p.
178).
10
In 1910, for example, F. W. Woolworth told the New York Times about
his soon-to-be-built eponymous tower, I do not want a mere building, I want
something that will be an ornament to the city (NY Times, 1910). In 1928,
Darwin P. Kinglsey, president of the New York Life Insurance, said at the
opening of his companys new tower, The skyline of New York is singularly
beautiful because it expresses power; it strikes a new note of power....[O]ur
objective has been to express in this building the power that makes the New
York skyline beautiful... (NY Times, 1928).
Even today, developers, be they speculative or corporate, are still inter-
ested in projecting their ambitions onto the skyline. Noted builder, Donald
Trump, in 1998, told the New York Times when announcing his new resi-
dential building Trump World Tower, Ive always thought that New York
should have the tallest building in the world....It doesnt. But now, it has the
tallest and most luxurious residential building in the world (Bagli, 1998).
6 Empirical Analysis
The optimal height equation (3) demonstrates that the building height on a
specic site will be a function of the income that can be generated, the cost
10
In regard to the 1875 New York Tribune building, Wallace (2006) writes, The nine-
story height insured that the tower would be taller than any existing New York oce
building and was thus neither an arbitrary choice of height nor one based on the functional
space requirements of the newspaper. The design of the Tribune building was primarily
governed by the enhanced public image that would be garnered for the newspaper and
only tangentially by the potential economic benets of building tall (p. 179).
17
of building, and the utility value of status. To estimate the model, however,
requires the collection of unobservable data, such as on the value of , and on
building-specic costs and income ows. Unfortunately, this type of specic
building data do not readily exist.
Rather I use a combination of site-specic and economy-wide variables to
measure the costs and benets of skyscraper construction. For this reason,
I estimate the following linear econometric model for the optimal building
height:
h

i
=
0
+
1
a
i
+
0
2
x
i
+
0
3
z
i
+
0
3
w
i
+
i
,
where
i
=
0
2
x
i
,
i
=
0
3
z
i
,and
i
=
0
5
w
i
.
11
That is,
i
can be decomposed into a weighted sum of factors that aect
the net income fromthe site.
i
is a weighted sumof factors that contribute to
construction costs,
i
is decomposed into a weighted sum of status factors.
a
i
is the plot area.
i
is the random, unmeasured, i.i.d. component of height,
assumed to be normally distributed. We now turn to the specic variables.
6.1 Data
Here I give a general description of the variables. Appendix B contains more
information on the sources of the data. As the model above illustrates, height
is a function of the costs of building, the buildings expected revenues, the
plot size and the degree to which a building will provide relative status.
Table 1 provides descriptive statistics for the sample of 458 skyscrapers in
Manhattan, completed between 1895 and 2004; all are located in Manhattans
densest area (south of 96th street).
To simplify the analysis, I limit the sample to buildings that are 322
feet (100 meters) in height, as determined by the international real estate
consulting rm Emporis. The height measured is structural height, and
therefore excludes antennae or decorative elements. Limiting a skyscraper
to 100 meters or taller means that I only include buildings that have about
30 or more oors.
12
As discussed above, since the late 1880s, the problem
of engineering height was essentially solved, and thus the issue of how tall
11
Note that I assume that is constant over time, and thus and are the important
parameters with respect to the costs and benets of height.
12
Note that there is not a one-to-one relationship between the number of oors and
height. In this sample, the average number of feet per oor 12.62, with a standard deviation
of 1.81, with a min. of 8.98, and a max. of 19.96.
18
to build was one of economics. Clearly, what makes a building a skyscraper
is relative, especially in the early period of their construction. But since
100-meter buildings have regularly been built since 1895, I use this value as
the minimum height to be considered a skyscraper (see Barr (2007) for more
information on skyscraper time series).
Building Information The dependent variable, as mentioned above,
is building height in feet. For each building, I include the log of plot size.
The log form is used since the distribution of plot size is highly skewed to
the right. A few plots like those of the World Trade Center and United
Nations are extremely large. In these instances, the respective government
agencies created a master plan of superblocks. To take into consideration
that sometimes plot sizes can be irregular (due to holdouts or odd-shaped
blocks) and this irregularity may add an extra cost to construction, I have
included a dummy variable that takes is one if the block is not perfectly
rectangular or square, zero otherwise.
In addition, I include dummy variables for the buildings main use at
the time of completion. Use is determined by both zoning, which can limit
building types in specic neighborhoods, as well as land values and the rel-
ative rents available from oce or residential space, which can depend, in
part on taxes and subsidies. I assume that exogenously determined histor-
ical land use patterns and zoning rules are the primary driver of building
uses. As can be seen from Table 1, oces comprise roughly 64% of building
types, while 21% are residential. The remaining 15% are divided between ho-
tels, mixed use (which include oce-residential, oce-hotel, and residential-
hotels), government-use buildings (e.g., courthouses), hospitals, and a nal
category, utility, which comprises highrise buildings that house telephone
and communications equipment. One would assume oces would be higher,
cet. par., due to their greater income ows.
For each plot, I also have a measure of the approximate average depth to
bedrock to investigate the degree to which this depth has aected building
height. To account for the possibility that downtown and midtown may have
diering soil conditions, I created two separate variables, one is the depth of
the bedrock for downtown buildings and the other is the depth for midtown
buildings (i.e., each variable is the depth to bedrock interacted with an area
dummy variable). Presumably, the further down a builder needs to go the
greater the costs and the lower the height.
19
Variable Mean Std. Dev. Min. Max.
Building Information
Height (feet) 490.24 141.22 328.08 1368.11
Plot (000 sq. feet) 44.67 63.66 4.00 681.60
Plot Irregular Dummy 0.570
Depth to Bedrock (feet) 88.56 85.20 0.26 524.43
Distance to District Core (Miles) 0.612 0.448 0.016 2.55
Downtown Dummy 0.212
Use Dummy Variables
Residential Condominum 0.081
Residential Rental 0.131
Government 0.013
Hospital 0.002
Hotel 0.055
Mixed-use 0.072
Oce 0.642
Utility 0.004
Status Variables
Avg. Height of All Completions
t1
465.23 25.93 311.70 490.66
Corporate HQ Dummy 0.164
World Record Dummy 0.022
NYC Record within Use Category Dummy 0.083
Zoning Variables
Built Under 1916 Zoning Laws Dummy 0.360
1916 Setback Multiple

1.927 0.374 1.25 2.5

Built Under 1961 Zoning Laws Dummy 0.590
Plaza Bonus Dummy 0.362
Max. Floor Area Ratio

12.71 2.58 3.44 15.0

Purchased Air Rights Dummy 0.153
Special Zoning District Dummy 0.037
Economic Variables (# obs.=86, for each year only)
Real Interest Rate (%) 2.75 4.50 -13.86 23.92
Real Construction Cost Index 1.34 0.266 0.872 1.67
NYC Area Population (Millions) 9.16 2.45 3.21 11.90
National F.I.R.E./Employment (%) 4.79 1.38 1.76 6.57
ln(Equalized Land Assessed Value) (%) 4.88 7.98 -18.4 33.0
Table 1: Descriptive Statistics for skyscrapers completed from 1895 to 2004.
# obs.= 458, unless otherwise noted. Sources: See Appendix B.

Stats. are
for buildings completed during relevant zoning rules.
20
To account for the fact that land values in the center are higher due to
agglomeration economies and the concentration of public transportation, I
use a measure of how close a building is to the central business core. In New
York, there are two cores, one centered on Wall Street and the other centered
in midtown at the Grand Central Station railway terminal. For downtown
buildings, I measure each buildings distance in miles to the corner of Wall
Street and Broadway. For midtown buildings, I measure each buildings dis-
tance in miles to Grand Central Station. Lastly, I include a dummy variable
for downtown buildings, to control for systematic dierences that may ex-
ist between downtown and midtown. Most notably, downtown has bedrock
quite far down and quicksand below the surface, while midtown has bedrock
quite near to the surface and relatively dry soil.
Zoning For the 1916 zoning rules, I include a dummy variable for the
buildings completed under this regime. In addition I include the setback
multiple (interacted with the 1916 dummy variable) to account for its aect
on height. I would expect a positive relationship with the multiple, but a
negative one with the dummy variable.
For the 1961 zoning rules, I also include a dummy variable for buildings
completed in this period (which is still in eect). I include the maximum
allowable FAR that each building had; this should have a positive coecient.
I also include dummy variables for buildings that have provided a public
amenity (as of right) and for buildings that have purchased air rights. I
would expect both of these dummy variables to have positive signs. Lastly I
also include a dummy variable for two special districts: Battery Park City
(which includes the World Trade Center) and the Times Square district. In
these cases, public agencies took the initiative in generating development in
these neighborhoods; as such developers were not bound by the same rules
in other districts. Because government agencies sought to promote business
development, zoning and FAR restrictions were relaxed. I would expect this
coecient to be positive.
Economy-wide Costs and Benets Because income and costs for
specic buildings are generally not available, I have used economy-wide time
series data to proxy for these variables (which are lagged one or two years
to account for the time between which height decisions are made and when
buildings are completed). For income, I include a measure of the population
21
of the New York City area (which, in this case, comprises the population of
New York City and three surrounding counties). Because NYC employment
data is not available back to 1893, I use national data to measure oce em-
ployment, which in this case is given by the ratio of national employment
in the Finance, Real Estate and Insurance (F.I.R.E.) industries to total em-
ployment. Finally to measure the health of the New York City real estate
market (and growth in land values), I include the lagged growth of equalized
assessed land values. Equalized assessed values are meant to adjust actual
assessed values to close to market values.
To measure costs I include the real interest rate on commercial paper and
an index of real building material costs (both lagged two years). I would
expect both of these coecients to be negative.
13
Status and Ego To measure the degree to which status aects height,
I include four status-related variables. First, I have a dummy variable to
account for buildings that are corporate headquarters. These buildings were
developed and owned by major corporations (this diers from speculative
developers who build skyscrapers and then lease them to corporations). If
a major corporation is also the developer then I would expect them to add
extra height to signal economic strength and advertise their building. Second,
I include the average building height of all buildings completed in the year
prior to the completion of each building. As equation (3) demonstrates,
average height of completions measures the relative ranking of a building.
Thus if status was important I would expect the coecient to be positive.
In addition, occasionally, a builder has the right economic conditions and
plot size to aim specically for the worlds tallest building (see Tauranac
(1995) for the case of the Empire State Building). To control for this fact, I
include a dummy variable for world record buildings. Lastly, I also include
a dummy variable to measure the additional height added by a developer
who aims to have the tallest building within its use class. Presumably status
is conferred upon the builder who, say, builds the citys largest residential
building (as did Donald Trump in 1998) or the citys tallest hotel.
13
Note that I dont not have any direct measures of technological change in building
materials and methods over the 20th century. However, the cost index used here implicitly
captures this change. For example, the index peaked in 1979. Between 1979 and 2003,
real constructions costs fell 15%. In 2003, the index was roughly the same value as it was
in 1947.
22
6.2 Results
Table 2 presents the results of the regressions. The rst regression is a
measure of economic height, without any of the status measures. Equation
(2) includes the measures of status and ego. The status measures increase
the explanatory power by about 15 percentage points. Three out of four of
the status measures are statistically signicant at the 90% level and all four
have signs consistent with the theory.
In regard to plot size, there is the potential issue of endogeneity with
respect to height. It could be that developers who would like to build tall
seek out the largest available plot sizes; thus there might be an issue of plot
size selection bias. To investigate this issue further, I collected three possi-
ble instruments related to the particular city blocks on which the buildings
reside: (1) the size of the block in feet squared (2) a dummy variable tak-
ing on the value of one if the block is a superblock, 0 otherwise, and (3) a
dummy variable that takes on the value of one if the block was planned by
a government agency according to some master plan.
Block sizes appear to be a suitable instrument given that they were de-
termined in 1811 with the implementation of the grid plan; thus they would
appear to be related to plot size but not skyscraper height. In a very few
cases, block sizes were combined to be superblocks. These are generally rare
(about 9% of the blocks). In the case of Battery Park City, new blocks were
created from landll. Superblocks tend to be clustered near the rivers and
not in the center of the island, since port activity used to dominate these
areas. Finally, a few blocks, generally near the rivers, were the product of
master plans, such as the United Nations, and the World Trade Center. The
master plans are generally produced in areas that were formally related to
port activities and as a result emerged due to exogenous historical conditions.
To consider using block size and related variables as an instruments, I
rst performed a test of the overidentifying restrictions, which produced a

2
-statistic=2.58. With 2 degrees of freedom, I can not reject the null hypoth-
esis of strictly exogenous instruments. In addition, the rst-stage regression
with instruments had an R
2
= 0.39, while the rst-stage regression without
the instruments had an R
2
= 0.22, indicating that these instruments had a
substantial impact on lot size. Also, as an additional test, when the three
potential instruments were included in equation (2), table 2, an F-test did
not reject the null hypothesis that they were jointly zero. Finally the Haus-
man test could not reject the null hypothesis of exogeneity for plot size. In
23
sum, the instrumental variable tests show that the block size variables are
valid instruments but that they are not needed since plot size appears to be
exogenous.
14
In regard to bedrock, the downtown depth to bedrock coecient is posi-
tive and statistically signicant in equation (1) and statistically insignicant
for equation (2). The positive eect in equation (1) might have been due to
an omitted variable bias eliminated in equation (2). However, this positive
eect comes because of the inclusion of the downtown dummy variable, which
appears to measure (at least in part) the diering subsoil conditions between
lower Manhattan and midtown. The downtown subsoil is often composed
of quicksand, which necessitates the use of caissons and expensive retaining
walls to ensure that surrounding buildings dont cave in during construction.
In midtown, the bedrock is relatively close to the surface and the soil is dryer.
In many cases, the main issue with foundation preparation is blasting away
the bedrock that is close to the surface. For midtown, the depth to bedrock
coecient shows a small negative eect; for example, for every extra 10 feet
a builder has to dig, I estimate a reduction in height of about 1.6 feet. Note
that when equation (2) was run without the downtown dummy variable (re-
sults not shown), the depth to bedrock coecients for both midtown and
downtown were negative but statistically insignicant.
The negative sign on the downtown dummy variable may also be picking
up something related the year of completion, since most of the early skyscrap-
ers were built downtown. But another regression (not shown) that includes
the year on the right hand side, shows the year variable to be statistically in-
signicant. In sum, the evidence does not strongly support that the depth to
bedrock is a key factor in skyscraper height. However, a more direct analysis
of the depth to bedrock and the placement of early skyscrapers is needed;
this is left for future research.
15
In general, all of the zoning-related coecients give the correct signs. The
estimates show, for example, that under the 1916 zoning rules, a building in a
two times district, would have reduced its height by 190.6+58.8(2) = 73
feet (about 6 oors) compared to the non-zoning era. Under the 1961 rules,
14
Descriptive statistics for the instruments as well as IV test results are available upon
request.
15
The failure to nd statistical signicance for the bedrock coecient may also be due
to measurement error, given the estimated averages from the maps, or possibly due to
the fact that several foundations may have been prepared for buildings that preceded the
skyscrapers in the data set.
24
Variable (1) (2)
ln(Plot Size) 74.4
(7.57)

59.1
(8.67)

Irregular Plot 22.3

(1.89)

18.2
(1.80)

Rental Apartment 25.1

(1.56)
19.7
(1.32)
Oce Building 13.1
(0.87)
26.2
(1.84)

Distance to Core (miles) 45.0

(3.25)

41.4
(3.27)

Depth to Bedrock Downtown (feet) 0.234

(2.21)

0.116
(1.14)
Depth to Bedrock Midtown (feet) 0.100
(0.81)
0.160
(1.64)

Downtown 79.8
(2.72)

61.6
(2.22)

Zoning 1916 251.8

(3.75)

190.6
(3.00)

Setback Multiple 66.4

(2.67)

58.8
(2.49)

Zoning 1961 312.6

(4.44)

268.0
(4.01)

Max. FAR 12.1

(3.77)

10.6
(3.68)

Plaza Bonus 38.8

(2.65)

51.08
(4.12)

Air Rights 82.6

(5.65)

65.0
(4.83)

Special District 132.2

(2.93)

95.8
(3.24)

Pop. NYC Area

t2
(millions) 19.8
(2.30)

20.8
(2.40)

% F.I.R.E./Employment
t2
41.6
(2.75)

39.3
(2.61)

% ln(Eq. Assessed Land Value)

t1
1.25
(1.62)
1.14
(1.71)

% Real Interest Rate

t2
4.52
(2.00)

4.99
(2.22)

Real Construction Costs

t2
241.3
(5.20)

255.1
(5.76)

Corporate HQ 24.50
(1.53)
World Record 337.0
(5.40)

Category Record 77.9

(3.19)

Avg. Height
t1
(feet) 1.10
(1.68)

Constant 198.2
(2.12)

589.6
(2.42)

R
2
0.41 0.57

R
2
0.38 0.54
Table 2: Dependent Variable: Skyscraper Height (in feet). Number of obser-
vation is 458. Absolute value of robust t-statistics below coecient estimates.

Stat. sig. at 10% level,

Stat. sig. at 5% level.

Stat. sig. at 1% level.
25
assuming a building was in a FAR district of 15, took advantage of a plaza
bonus, and purchased air rights from a neighboring building, the net eect
of zoning on height would be a modest 7.4 feet or less than one oor as
compared to the no-zoning regime before 1916.
All of the economic time series variables all have correct signs and are sta-
tistically signicant. Over the course of the 20th century, I estimate that for
each additional million people living in New York and surrounding counties,
height has increased by about one and a half oors. Interestingly, the interest
rate has a modest eect on height. For example, a doubling of interest rates
from 5% to 10% is associated with only a two oor drop in heights.
The role of status, as measured in the regression has had both a statis-
tically and economically signicant impact on the skyline. Most interesting
is perhaps the coecient for the average height of buildings. The coecient
shows that, on average, building height increases with the total completions
average a little more than equally (i.e., a one foot increase in the average of
all buildings increases building height by 1.1 feet). Over the twentieth cen-
tury the total completed average increased by about 177 feet, which means
that, on average, building height is now increased by about 15 oors so that
developers can keep up with the Trumps.
Being a corporate headquarters seems to add a relatively modest amount
of height, adding, on average about two oors. Those who aim for the
world record appear to shoot for the moon. Perhaps, on average, the record-
breaking developers aim to hold onto their world record for as long as time
as possible, since all else equal, a world record breaking developer adds over
400 feet (32 oors) to his building (the sum of the record break dummy and
the category record dummy). The developer aiming for a category record ap-
pears to be more modest in his aims, adding only about six oors to achieve
that record. This makes sense given that the categories include apartments,
hotels and hospitals, where income ows are not as great as oce buildings,
which have, to date, been the world record buildings.
7 Optimal Height and Beyond
In this section, I use the coecient estimates to perform two exercises. First,
by comparing the predicted economic height to the actual height, we can
get a measure of the degree to which a building is too tall in terms of
the competitive prot maximization sense. Second, we can look at how the
26
optimal height has changed over time, since this optimal height is a function
of the costs and benets of building, which generally vary from year to year.
Table 3 investigates the top too tall buildings in New York City, by us-
ing the predicted height values given by equation (1) in table 2. The table also
includes the actual number of oors and the predicted number of oors. Pre-
dicted oors were calculated via the formula
d
floors
i
=

h
c
i
(floors
i
/height
i
).
As can be seen from the table, the number one building is the Empire State,
with an estimated 54 oors more than economic height. Interestingly, too
tall buildings are common throughout the century; and most of them were
either world record holders and/or corporate headquarters. Only the second
to last buildings on the list was built as a pure speculative oce project; the
last building on the list is a hotel.
Rank Building Year h

h
c
Di. Floors
d
Fls.
c
Di.
1 Empire State 1931 1250 589 661 102 48 54
2 One World Trade Ctr. 1972 1368 865 503 110 70 40
3 Chrysler 1930 1047 547 500 77 40 37
4 Two World Trade Ctr. 1973 1362 890 472 110 72 38
5 AIG/Cities Services 1932 951 537 414 66 37 29
6 40 Wall St 1930 928 544 384 70 41 29
7 Citigroup Center 1977 915 574 342 59 37 22
8 JP Morgan Chase HQ 1960 705 371 335 52 27 25
9 Woolworth 1913 791 511 280 57 37 20
10 GE/RCA 1933 850 589 260 69 48 21
11 One Chase Man. Plz. 1961 814 563 251 60 42 18
12 20 Exchange Pl. 1931 741 503 239 57 39 18
13 Singer Building 1908 614 381 233 47 29 18
14 CitySpire Center 1987 814 593 221 75 55 20
15 Ritz Hotel Tower 1926 541 332 209 41 25 16
Table 3: Rank of top 15 (out of 458) too tall buildings in New York City.
Predicted values come from equation (1), table 2. h is the actual height;

h
c
is the predicted optimal economic height.
Figure 2 presents the predicted economic height and status height over
time for the same hypothetical plot discussed in Clark and Kingston (1930)
(here in CK). As did CK, I assume that a speculative oce would be con-
structed on the 81,000 square foot, regular-shaped plot, with a distance to
Grand Central Station of 0.1 miles, and a depth to bedrock of 55 feet. Next
27
I assumed that no zoning regulations were in eect till 1918 (assuming a two
year lag between ground breaking and completion). 1916 zoning rules were
in eect till 1963, and 1961 rules thereafter. For each year, I generated a
predicted value using annual time series data for the New York City area
population, oce employment, land value growth, real interest rates and
construction costs. Economic height was calculated by holding the lagged
average height of completed buildings constant at the initial value in 1894 of
311.7 feet (95 meters). The predicted height in feet was then converted to
the number of oors by dividing height by 12, the number of feet per oor
in the CK building.
Note that the predicted values are generated from a regression run on of-
ce buildings only (results available upon request). Also note that the graph
includes predicted values for years in which no skyscrapers were completed,
including during the depression and World War II. (See Barr 2007 for actual
time series of average heights and number of completions over the same time
period.)
20
30
40
50
60
70
1904 1914 1924 1934 1944 1954 1964 1974 1984 1994 2004
Economic Height Height withStatus
F
l
o
o
r
s
Figure 2: Predicted economic height and status height from 1904 to 2004 for
the Clark and Kington (1930) plot.
The graph shows a few things. First, on average, both economic and
status height are increasing over the century. Economic height starts at about
37 oors, and by 2004 is approximately 52 oors; similarly status height goes
from about 40 oors to 64 oors. This would be expected given the increasing
economic activity in New York. However, optimal height moves in waves,
28
with approximately four cycles over the twentieth century. Second, there is a
substantial dierence between economic and status height; by the centurys
end, the status dierential on this plot is about 12 oors by the centurys
end. Lastly, the regressions under-predict CKs optimal height of 63 oors
in 1929. For that year, the regressions give an optimal height of 41 and 49
for economic and status height, respectively. As discussed above, CK used
actual rent and cost numbers for a specic plot, whereas here the estimates
come from proxy variables and include coecients estimated from data for
110 years.
8 Conclusion
This paper has investigated the determinants of skyscraper height. First I
provide a simple model of builders, who must consider the prots from con-
struction in addition to their relative status ranking when deciding how tall
to build. Empirical estimates of this model for New York City show that eco-
nomics, land use regulation and status are important determinants of build-
ing height over the 20th century. In general, as would be expected, as the
costs and benets to construction have changed, so has the optimal height.
Height has responds positively to population and oce job growth, and neg-
atively to interest rates and building costs. As expected, I also nd evidence
that height is positively related to land values. The estimated height gra-
dient shows that height drops at a rate of about 3.3 oors per mile vis a
vis the business core. Zoning regulations, while not limiting height per se,
have negatively impacted height by placing restrictions on the shape of the
building or the total amount of building volume. But amenity bonuses and
air rights purchases, which have been common over the last 30 years, have
allowed builders to go taller than otherwise. I nd no strong evidence that
the depth to bedrock has aected skyscraper height, though diering subsoil
conditions and bedrock depths in lower Manhattan and midtown appear to
have aected height in this two dierent urban cores.
In addition, I include four variables to account for builders desire to ob-
tain status and recognition. I nd that corporations who build headquarters
add only a modest amount of height, as compared to speculative oce de-
velopers (about two extra oors). I also nd that builders who aim for world
records strive for the moon by adding, on average, about 32 extra oors
beyond the prot maximizing amount. While those developers aiming for
29
a record within their own use group (such as apartments or hotels) add a
modest six oors. Finally, by looking at the lagged average height of comple-
tions, I estimate that the desire to stand out in the skyline has increased
building height by about 15 oors by the end of the 20th century.
As discussed above, surprisingly little work has been done on the eco-
nomics of skyscrapers. Despite the continued fascination by the public,
journalists and scholars within other disciplines, the eld of skynomics
remains relatively unexplored. Future work might consider measuring the
rate at which technological improvements in skyscrapers have occurred since
the 1890s. More work is needed to directly measure of the supply and de-
mand for height. Lastly, one could measure the degree to which skylines, as
goods unto themselves, improve the well-being of urban residents.
30
A Proofs
Lemma 1 For

0,

, there is a unique value of h, h

, such that h

=
arg max
hR
+
u
i
(h) .
Proof. For = 0, h

= arg max
hR
+
(h

) , which is unique by denition.

For > 0, at h = 0, (0) +F (0) = 0. Since F (h) is a cdf, its maximum value is
one. Further, since prot is single-peaked and strictly concave, there exits and

h,
such that,

+

= 0. Thus for h
h
0,

h
i
, there must be a global maximum,
since the contribution of F (h) to utility is bounded and adding F (h) to (h)
preserves the single peaked nature of the utility function when h
h
0,

h
i
.
Lemma 2 h

> h
c
for

0,

; h

= h
c
for = 0.
Proof. If = 0, then h

= h
c
will be equal since the status developer is
maximizing the same function as the competitive developer. If 0 < , then
given Lemma (1), there exists a unique h

,such that u
0
i
(h

) =
0
(h

)+f (h

) = 0,
or
0
(h

) = f (h

) , where f (h

) > 0. Given that (h) is single-peaked, the

optimal building height is therefore taller than a building where
0
(h) = 0; that
is, h

will be chosen along the negatively sloped portion of the prot function.
Lemma 3 h

is strictly increasing with .

Proof. For a given h

, the utility function is at a global maximum and

therefore u
0
(h

) = 0, and u
00
(h

) =
00
(h

) +f
0
(h

) < 0. The rst order condition

gives
0
(h

) +f (h

) = 0. Via the envelope theorem: [

00
(h

) + f
0
(h

)] h

/+
f (h

) = 0, which gives h

/ = f (h

) / [
00
(h

) + f
0
(h

)] > 0.
Lemma 4 l

is monotonically increasing in ; l

= l
c
, when = 0.
Proof. Given the optimal height for a plot, h

, land value is given by l

(
i
) =
(h

())+
i
F (h

(
i
))r. By the envelope theorem, dl

/d = [
0
(h

()) + f (h

())] h

/+
F (h

) = F (h

) > 0 for all > 0 since

0
(h

) + f (h

) = 0. If = 0, land value
is simply given by l

(
i
) = (h
c
) r, since h
c
maximizes prot.
Lemma 5 Given the monotonicity of l

, the minimum and maximum land

values for a plot of land is l
c
and

l

= (h

) + F (h

) r, respectively.
Proof. If = 0, l

= l
c
= l (h
c
) = (h
c
) r, where l
0
(h
c
) =
0
(h
c
) = 0. As
discussed in lemma (4), for > 0, since l

is strictly increasing in , and has a

maximum of

, no developer would be willing to pay more than

l

Lemma 6 Given that U

0,

, the pdf for land valuations is given by

k (l

) =
(
1
F(h

)
, l

l
c
,

0, otherwise
)
;with a cdf of K(l

) =
l

l
c
F(h

)
, l

l
c
,

.
Proof. First note that land valuations are linear functions of : l

() =
[ (h

) r] + F (h

) . Since U

0,

, g () =
1

. The pdf of l

() follows
31
simply from the formula for the distribution of a random variable that is a linear
function of a uniformly distributed variable (DeGroot, 1989). Note that l
c
=
(h

) r. The cdf follows from K(l

) =
1
F(h

)
R
l

l
c
dx =
l

l
c
F(h

)
.
Proposition 1 Given each agents land valuation function, and that the sta-
tus parameter has a Uniform

0,

distribution, there exists a unique, symmet-

ric equilibrium of the land auction game such that each agents bid is given by
(l

) =
(N1)l

+l
c
N
.
Proof. Note that this proof is adapted and condensed from Krishna (2002), to
which the reader is referred for more information. Lets say there are N bidders,
each with private valuation of l

i
(
i
) , where l

i
is strictly increasing in . Suppose
there exists a symmetric, increasing equilibrium strategy, (l

i
) . First, it would
never be optimal to bid b >

, since the agent would win the auction and

could have done better by slightly reducing his bid, as he could win and pay less.
Second, a bidder with = 0, would never submit a bid greater than l
c
, since
he would have negative utility if he were to win, thus (0) = l
c
. Bidder i wins
the auction when he submits the highest bid; that is when max
j6=i

< b.
Dene
N1
as the value of for the second highest bidder out of N bidders.
Since (l

) is increasing, bidder i wins if he has the highest value of l

i
(i.e.,
if
i
>
N1
) or if

l
N1

< b, or equivalently if l
N1
<
1
(b) . Agent
i
0
s expected payo is therefore K

1
(b)

N1
(l

b) , where K (l

)
N1
is the
distribution of the second highest order statistic for land values. Taking the rst
order condition, replacing b = (l

) (at the symmetric equilibrium), and solving for

the dierential equation given by the FOC, yields the equilibrium function (l

) =
h
1/K (l

)
N1
i
(N 1)
R
l

l
c
ydK (y)
N2
dy. Given that land values are distributed
U

l
c
,

, this bid function is (l

) =
(N1)l

+l
c
N
.
This result is a necessary condition for the optimal strategy, we now turn to
showing the sucient condition: that if the N 1 bidders follow (l

) , then it is
optimal for agent i to do so as well. Suppose that all agents but bidder i follow the
strategy (l

) . Given that the winner has the highest bid, it is never optimal for
agent i to bid more than

. Denote z =
1
(b) as the value for which b is the
equilibrium bid for agent i, that is (z) = b. The expected payo to agent i from
bidding (z) is ( (z) , l

) = K (z)
N1
(l

(z)) = K (z)
N1
(l

(z)) +
R
z
l

K (y)
N1
dy, where this equality is obtained via integration by parts. This
leads to the conclusion that ( (l

) , l

) ( (z) , l

) 0, regardless of whether
z l

or z l

. Thus for agent i, not using (l

) will make the agent no better

o, which implies that (l

) is a symmetric equilibrium strategy.

32
B Data Sources and Preparation
-Skyscraper Height, Number of Floors and Year of Completions: Emporis.com.
-Plot size: NYC Map Portal (http://gis.nyc.gov/doitt/mp/Portal.do); Ballard
(1978); http://www.mrocespace.com/; NYC Dept. of Buildings Building Infor-
mation System, (http://a810-bisweb.nyc.gov/bisweb/bsqpm01.jsp).
-Plot Regularity: Various editions of the Manhattan Land Book (see references)
and the NYC Map Portal.
-Use and Corporate HQ: For each building, one or more articles about the
building were obtained from the New York Times at the time of the buildings
construction or just after its completion. From this, I ascertained its primary use
and the developer. If the developer was a major corporation and the corporation
had an equity stake in the building, it was listed as a Corporate Headquarters.
-Distance from Core: For each building I obtained the latitude and longitude
from http://www.zonums.com/gmaps/digipoint.html. Then I calculate the dis-
tance for each building i = 1, ..., 458, from its respective core using the formula d
i
=
q
[69.1691 (latitude
i
latitude
core
)]
2
+ [52.5179 (longitude
i
longitude
core
)]
2
, where
latitude and longitude were initially measured in degrees. The degrees to miles con-
version is from http://jan.ucc.nau.edu/~cvm/latlongdist.html. Given New Yorks
location on the earth, 1

latitude is about 69.1691 miles and 1

longitude is about
52.5179 miles. There are two cores: the intersection of Wall Street and Broadway
(downtown) and Grand Central Station (42nd Street and Park Ave.). All buildings
south of 14th street belong to the downtown core; all buildings on 14th street or
above belong to the midtown core.
-Depth to Bedrock: For each building, elevation from sea level (in feet) comes
from http://www.zonums.com/gmaps/digipoint.html. Depth of bedrock from sea
level (in feet) comes from maps provided by Dr. Klaus Jacob, Lamont-Doherty
Earth Observatory of Columbia University. The maps are based on hundreds of
borings throughout Manhattan. The depth to bedrock was calculated by subtract-
ing the depth of bedrock from sea level from the elevation from sea level.
-Zoning 1916 and 1961: The New York Times was consulted to determine the
rst buildings completed under the respective regimes.
-1916 Height Multiples: Original zoning maps in eect at the time of comple-
tion for each building. The maps were provided by the New York City Department
of City Planning.
-1961 Maximum Allowable FAR: Original zoning maps in eect at the time
of completion for each building. The maps were provided by the New York City
33
Department of City Planning.
-Special Districts: Zoning maps from NYC Department of City Panning, and
articles from the New York Times.
-Air Rights: Data about which buildings purchased air rights comes the New
York Times, Real Estate Weekly and http://beta.therealdeal.com/front.
-Plaza Bonus: Kayden (2000); www.nyc.gov/html/dcp/html/priv/priv.shtml
-Real Construction Cost Index (1893-2004): Index of construction material
costs: 1947-2004: Bureau of Labor Statistics Series Id: WPUSOP2200 Materials
and Components for Construction (1982=100). 1893-1947: Table E46 Building
Materials. Historical Statistics (1926=100). To join the two series, the earlier
series was multiplied by 0.12521, which is the ratio of the new series index to the
old index in 1947. The real index was create by dividing the construction cost
index by the GDP Deator for each year.
-GDP Deator (1893-2004): Johnston and Williamson (2007). (2000=100).
-Finance, Insurance and Real Estate Employment (F.I.R.E)/Total Employ-
ment (1893-2004): 1900-1970: F.I.R.E. data from Table D137, Historical Statis-
tics. Total (non farm) Employment: Table D127, Historical Statistics. 1971-2004:
F.I.R.E. data from BLS.gov Series Id: CEU5500000001 Financial Activities. To-
tal nonfarm employment 1971-2004 from BLS.gov Series Id:CEU0000000001. The
earlier and later employment tables were joined by regressing overlapping years
that were available from both sources of the new employment numbers on the old
employment numbers and then correcting the new number using the OLS equa-
tion; this process was also done with the F.I.R.E. data as well. 1893-1899: For both
the F.I.R.E. and total employment, values were extrapolated backwards using the
growth rates from the decade 1900 to 1909, which was 4.1% for F.I.R.E. and 3.1%
for employment.
-Real Interest Rate (nominal rate minus ination) (1893-2004): Nominal in-
terest rate: 1893-1970: Table X445 Prime Commercial Paper 4-6 months. His-
torical Statistics. 1971-1997 http://www.federalreserve.gov, 1998-2004: 6 month
CD rate. 6 month CD rate was adjusted to a CP rate by regressing 34 years of
overlapping data of the CP rate on the CD rate and then using the predicted
values for the CP rate for 1997-2004. Ination comes from the percentage change
in the GDP deator.
-Population NYC, Nassau, Suolk, and Westchester Counties (1893-2004):
1890-2004: Decennial Census on U.S. Population volumes. Annual data is gen-
erated by estimating the annual population via the formula pop
i,t
= pop
i,t1
e

i
,
where i is the census year, i.e., i {1890, 1900, ..., 2000}, t is the year, and
i
is
solved from the formula, pop
i
= pop
i1
e
10
i
. For the years 2001 - 2004, the same
34
growth rate from the 1990s is used.
-Equalized Assessed Land Value Manhattan (1893-2004): Assessed Land Val-
ues: 1893-1975: Various volumes of NYC Tax Commission Reports. 1975-2003
Real Estate Board of NY. Equalization Rates: 1893-1955: Various volumes of NYC
Tax Commission Reports. 1955-2004: NY State Oce of Real Property Services.
Equalization Rate: 1893-1955: Various reports NYC Tax Commission Reports.
1955-2004: NY State Oce of Real Property Services.
-World Records: Helsely and Strange (2007) and Emporis.com.
-Use category records: From the building height data, based on year of com-
pletion.
- Average Height: Taken from building height data, with demolitions removed
in year of demolition, except for the World Trade Center Buildings, which were
left in the averages. The rst value for this variable is 311.7 feet (95 meters).
35
References
[1] Bagli, C. V. (1998). Trump Starts A New Tower Near the U.N. New
York Times, October 16, A1.
[2] Ballard, Robert F. R. (1978). Directory of Manhattan Oce Buildings.
McGraw-Hill: New York.
[3] Bar-Ilan, A. and Strange, W. C. (1996). Investment Lags. American
Economic Review, 86(3), 610-622.
[4] Barr, J. (2007). Skyscrapers and the Skyline: Manhattan, 1895-2004.
Rutgers University Newark Working Paper #2007-002.
[5] Bennett, C. G. (1960). New Zoning Law Adopted by City. New York
Times, Dec. 16, p. 1.
[6] Betsky, A. (2002). Babylon Revisited. Architecture, 91(12), 42-52.
[7] Case, K. E. and Shiller, R. J. (1989). The Eciency of the Market for
Single-Family Homes. American Economic Review, 79(1), 125-137.
[8] Chandler, A. D., Jr. (1977). The Visible Hand: The Managerial Revolu-
tion in American Business. Belknap Press: Cambridge.
[9] Clark, W. C. and Kingston, J. L. (1930). The Skyscraper: A Study in
the Economic Height of Modern Oce Buildings. American Institute of
Steel Construction: New York.
[10] DeGroot, M. H. (1989). Probability and Statistics, 2nd Ed. Addison-
Wesley: Reading.
[11] DiPasquale D. and Wheaton, W. (1995). Urban Economics and Real
Estate Markets. Prentice-Hall: Englewood Clis.
[12] The Economist (2006). Better than Flying Alone, June 1, 69-71.
[13] Gluckman, R. (2003). How High Will They Build? Popular Science,
262(3), 60-70.
[14] Helsley, R. W. and Strange, W. C. (2007). A Game-
Theoretical Analysis of Skyscrapers. Journal of Urban Economics,
doi:10.1017.j.jue.2007.08.004.
36
[15] Historical Statistics of the United States: Colonial Times to 1970, Bicen-
tennial Edition (1976), vols. 1 & 2. Bureau of the Census: Washington
DC.
[16] Hopkins, E. and Kornienko, T. (2004). Running to Keep in the Same
Place: Consumer Choice as a Game of Status. American Economic
Review, 94(4), 1085-1107.
[17] Johnston, L. D. and Williamson, S. H. (2007) What Was the U.S. GDP
Then? MeasuringWorth.Com
[18] Kayden, J. S. (2000). Public Owned Space: The New York City Experi-
ence. John Wiley & Sons: New York.
[19] Krishna, V. (2002). Auction Theory. Academic Press: San Diego.
[20] Landau, S. B. and Condit, C. W. (1996). Rise of the New York Sky-
scraper: 1865-1913. Yale University Press: New Haven.
[21] Land Book of the Borough of Manhattan, City of New York. (1927).
G.W. Bromley: New York.
[22] Land Book of the Borough of Manhattan, City of New York. (1955).
G.W. Bromley: New York.
[23] Land Book of the Borough of Manhattan, City of New York (Sanborn).
(2002). TRW REDI Property Data: Weehawken.
[24] New York Times. (1910). New Woolworth Building on Broadway Will
Eclipse Singer Tower in Height. November 13, RE1.
[25] New York Times. (1928). Huge Home Opened by New York Life. De-
cember 13, 16.
[26] Ramstack, T. (2007). Sky No Longer Limit for Skyscrapers. Washing-
ton Times, September 2, A1.
[27] Samuels, D. (1997). The Real-Estate Royals: End of the Line? New
York Times, August 10, SM36.
[28] Sabbagh, K. (1989). Skyscraper: The Making of a Building. Penguin
Books: New York.
37
[29] Shultz, E. and Simmons, W. (1959). Oces in the Sky. Bobbs-Merrill
Company: Indianapolis.
[30] Strange, W. C. (1995). Information, Holdouts, and Land Assembly.
Journal of Urban Economics, 38, 317-332.
[31] Tauranac, J. (1995). The Empire State Building: the Making of a Land-
mark. Scribner: New York.
[32] Titman, S. (1985). Urban Land Prices under Uncertainty. American
Economic Review, 75(3), 505-514.
[33] Wallace, A. (2006). A Height Deemed Appalling: Nineteenth-Century
New York Newspaper Buildings. Journalism History, 31(4), 178-189.
[34] Wheaton, C. W. (1999). Real Estate Cycles: Some Fundamentals.
Real Estate Economics, 27(2), 209-230.
[35] Willis, C. (1995). Form Follows Finance: Skyscrapers and Skylines in
New York and Chicago. Princeton Architectural Press: New York.
38