T HE E CONOMICS OF D ENSITY: E VIDENCE FROM THE
B ERLIN WALL∗
Gabriel M. Ahlfeldt
London School of Economics
Stephen J. Redding
London School of Economics and CEPR
Daniel M. Sturm
London School of Economics and CEPR
Nikolaus Wolf
Warwick University and CEPR
January 13, 2010
Abstract
A fundamental question for understanding the distribution of economic activity across space
is the strength of agglomeration and dispersion forces. This paper uses the division and reunification of Berlin as a natural experiment to provide evidence on the strength of these forces. We find
that division led to a re-orientation of Berlin’s rent gradient away from the pre-war city center
East of the Berlin Wall and towards the center of West Berlin. In the aftermath of re-unification,
this change in the rent gradient is reversed with the re-emergence of the pre-war city center. We
show that a canonical model of economic geography can account for these findings.
∗
This paper was produced as part of the Globalization Programme of the ESRC-funded Centre for Economic Performance. We are grateful to the German Science Foundation (DFG) for financial support. Claudia Steinwender and Nicolai
Wendland provided excellent research assistance. We would like to thank Dave Donaldson and Esteban Rossi-Hansberg
for helpful comments and suggestions. The usual disclaimer applies.
1
1
Introduction
The uneven distribution of economic activity perhaps receives its most vivid manifestation in the
formation of cities. Equally striking is the concentration of economic functions in particular localities within cities, such as the collection of advertising agencies found in mid-town Manhattan.1 The
strength of the agglomeration and dispersion forces that support these extreme concentrations of economic activity and shape their internal structure is therefore a fundamental question in economics.
Achieving a firm understanding of these forces is central to a host of economic and policy issues. To
take just one example, cost-benefit analyzes of transport infrastructure improvements depend crucially on the balance of agglomeration and dispersion forces. More generally, these forces are central
to determining the productivity advantages of cities, the spatial concentration of economic growth
and appropriate tax and spending policies.2
While there is a long literature on economic geography dating back to at least Marshall (1920), determining the strength of agglomeration and dispersion forces remains a challenging problem. High
commercial and residential rents in a location are consistent with strong agglomeration forces, but
could also reflect locational fundamentals that make a residential area an attractive place to live (e.g.
leafy suburbs and scenic views) or that make a commercial area a productive site (e.g. access to water or other favorable natural advantages). This problem of distinguishing spillovers from correlated
individual effects is all the more difficult for location choices, because natural advantages of purely
historical significance can have long-lived effects as a result of sunk costs, durable structures and coordination problems (e.g. the ports of London and New York). To convincingly distinguish spillovers
from correlated individual effects, one requires a source of exogenous variation in the surrounding
mass of economic activity, which is typically difficult to find.
In this paper, we use the division and reunification of Berlin as a natural experiment that provides
exogenous variation in the surrounding concentration of economic activity. Map 1 illustrates how the
Berlin Wall ran through the heart of the pre-war city and severed West Berlin from the pre-war city’s
commercial center in the district (“Bezirk”) of Mitte. The drawing of the border between East and
West Berlin was motivated by military considerations of allocating areas of approximately equal size
and population to the American, British and Soviet armies. Following the collapse of the wartime
alliance between the Western Powers and the Soviet Union, and to stem the flow of civilians leaving
East Germany, West Berlin’s boundaries with East Berlin and its surrounding hinterland were completely sealed by East Germany’s construction of the Berlin Wall in 1961. While the division of Berlin
appeared to be permanent, Soviet policies of “Glasnost” and “Perestroika” in the late 1980s started
1
2
See, for example, Arzaghi and Henderson (2008).
See, for example, the discussion in Desmet and Rossi-Hansberg (2009).
2
a process of opening up in East Europe. This opening-up in turn stimulated large-scale demonstrations in East Germany, which culminated in the fall of the Berlin Wall on 9 November 1989. Only
eleven months later East and West Germany were formally reunified on 3 October 1990 and the two
halves of Berlin became again part of the same city.
The key idea behind our empirical approach is that localities within West Berlin were differentially affected by the city’s division. Neighborhoods close to the pre-war commercial center East of
the Berlin Wall lost access to nearby sources of production externalities and employment opportunities. In contrast, the effect on neighborhoods further from the pre-war commercial center was more
muted, because they were more remote from these sources of production externalities and employment opportunities. We examine how the internal structure of economic activity within West Berlin
was re-organized in response to the loss of the pre-war commercial center. We also examine the
extent to which any change in the internal structure of West Berlin persisted after re-unification or
whether economic activity began to re-orientate itself along pre-war lines.
Our approach has a number of attractive features. First, the division of the city was driven by
exogenous military considerations that are unlikely to be correlated with pre-war characteristics of
localities. Second, we are able to examine the impact of division over a long period of time and can
compare the effects of division and reunification. Third, the Berlin Wall severed pre-war transport
arteries such as suburban (“S-Bahn”) and underground (“U-Bahn”) railway lines, which provides an
additional source of variation across areas of West Berlin. Fourth, we have a rich source of historical and contemporary data on the organization of economic activity within Berlin, including land
rents, employment by workplace and residence, and transportation infrastructure among a number
of other variables. Fifth, these data are available at a fine level of spatial disaggregation comprising
around 5,000 city blocks within West Berlin. Sixth, we combine the exogenous variation provided
by the construction and dissolution of the Berlin Wall with the structural estimation of a canonical
model of economic geography. We use the structure of the model to estimate the strength of agglomeration and dispersion forces and show that the model can account quantitatively for the impact of
both division and re-unification.
To guide our empirical analysis, we develop an extension of the model of city structure of Lucas
and Rossi-Hansberg (2002, henceforth LRH). To allow for rich asymmetries in city structure and
facilitate the quantitative analysis of the model, we introduce heterogeneities in agents’ commuting
choices following Eaton and Kortum (2002). City structure is determined in the model by the tension
between the agglomeration forces of knowledge spillovers and the dispersion forces of an inelastic
supply of land and commuting costs. The model allows for variation in natural advantage across
localities within Berlin and it is determined endogenously within the model whether city structure
3
is mono- or poly-centric.
Given data on employment by workplace, employment by residence and the transport network
(which determines commuting costs), the model can be solved for equilibrium wages in each place
of employment. Having solved for wages, productivity can be determined from employment by
workplace and land rents, while consumption amenities can be determined from employment by
residence and land rents. Having pinned down productivity and consumption amenities, agglomeration forces and commuting costs can be identified using the structure imposed by the model on
the relationship between productivity and the surrounding concentration of economic activity and
on the relationship between commuting costs and distance traveled.
Depending on the strength of agglomeration and dispersion forces, the model can exhibit a
unique equilibrium or multiple equilibria. As a result, while reunification makes areas of West Berlin
close to the Berlin Wall more attractive locations for employment and residence, it need not necessarily reverse the impact of division. Examining the extent to which the organization of economic activity within Berlin reverts towards its pre-war pattern therefore contains information on the strength
of agglomeration and dispersion forces within the model.
Our paper is related to a number of literatures. First we build on a long line of theoretical research
on the formation and structure of cities. Following Alonso (1964), Mills (1967) and Muth (1969),
traditional theories of cities assume a mono-centric structure and derive the equilibrium rent gradient
and pattern of urban land use as a function of distance from the Central Business District (CBD).
More recently, Fujita and Ogawa (1982) and Lucas and Rossi-Hansberg (2002, henceforth LRH) have
developed richer models of city structure, in which it is endogenous whether the city takes a monoor poly-centric form.3
Second our findings relate to the large empirical literature on the nature and sources of agglomeration economies, as reviewed in Rosenthal and Strange (2004). A key insight underlying this
literature is that population mobility requires the higher land rents of urban areas to be offset by
higher wages, which in turn requires higher productivity in urban areas. Therefore a long line of
research has used wage data to estimate the relationship between productivity and population density, including Glaeser and Mare (2001), Rauch (1993) and Sveikauskas (1975) among many others.4
A somewhat smaller literature has used rent data to estimate this relationship, including in particular Deckle and Eaton (1999) and Roback (1982). While there is a strong empirical relationship
between productivity and population density, determining how much of this relationship is causal
is more challenging. One approach to is to use instruments for density, but finding variables that
3
For a survey of the theoretical literature on the micro-foundations of urban agglomeration economies, see Duranton
and Puga (2004).
4
Other related approaches use data on employment density (as in Ciccone and Hall 1996) or employment growth (as in
Glaeser et al. 1992 and Henderson, Kuncoro and Turner 1995).
4
plausibly satisfy the exclusion restriction of only affecting productivity through population density
is difficult.5 Furthermore, while this line of research has generated a large number of estimates of
the reduced-form relationship between productivity and population density, there have been few
attempts to structurally estimate the separate contributions of agglomeration and dispersion forces.
Indeed, there is relatively little evidence in general on the magnitude of dispersion forces such as
commuting costs.
Third another literature has to tried to isolate the causal impact of agglomeration on employment
wages, land rents and productivity by exploiting changes in transport infrastructure. For example,
Crafts and Leunig (2009) consider the expansion of the railway network in nineteenth-century England and Wales, Gibbons and Machin (2005) examine the Jubilee Underground Line extension in
London, while McDonald and Osuji (1995) consider the Chicago Midway Rapid Transit Line. The
chief challenge facing this line of research is that the construction of transportation infrastructure is
typically endogenous and it is often difficult to find valid instruments. In contrast, the construction
and fall of the Berlin Wall provide a plausibly exogenous source of variation in network connections
for transport infrastructure within West Berlin.
Fourth our paper is related to the broader empirical literature on economic geography.6 A line of
research within this literature has examined the impact of natural experiments on industry location,
including war-time bombing (as in Davis and Weinstein 2002 and Brakman et al. 2004) and the
division and re-unification of East and West Germany (as in Redding and Sturm 2008 and Redding
et al. 2009). Here we exploit data on a much finer spatial scale (thousands of blocks within a city)
and for a wider range of economic outcomes than usually available (rents and employment by both
workplace and residence). Together with the model, this richness of the data permits a structural
estimation of agglomeration and dispersion forces and a quantitative analysis of the model’s ability
to explain the change in the organization of economic activity within West Berlin.
The remainder of the paper is structured as follows. Section 2 discusses the historical background.
Section 3 outlines the model. Section 4 reports some reduced-form empirical results on the impact
of Berlin’s division and reunification. Section 5 undertakes a quantitative analysis of the model’s
predictions. Section 6 concludes.
5
For example, lagged population density is often used as an instrument for current population density, but as natural
advantages are strongly persistent over time, it is unlikely that lagged population density only affects productivity through
current population density. To address such concerns, Rosenthal and Strange (2008) and Combes et al. (2009) use geology
as an instrument, exploiting the limitations imposed on building heights by geological sub-structure.
6
For surveys of this empirical literature, see Head and Mayer (2004), Overman et al. (2003) and Redding (2009).
5
2
Historical Background
The political process that ultimately led to the construction of the Berlin Wall has its origins in the
Second World War. While a number of proposals to divide Germany after its eventual defeat were
discussed during the early phase of the Second World War, the United States and the Soviet Union
backed off such plans towards the end of the war (see for example Franklin 1963 and Kettenacker
1989). Instead it was decided to allocate separate “occupation zones” to the American, British and
Soviet armies. Although Berlin was located around 200 kilometers East of the border of the Soviet
occupation zone, it was decided that the capital of pre-war Germany should be jointly occupied. For
this purpose, Berlin was itself divided into separate “occupation sectors.” The planning process for
these zones and sectors began in Spring 1943, negotiations continued during 1944, and the protocol
formalizing their boundaries was signed in London in September 1944. The protocol was modified
in 1945 to create a small French occupation zone and sector by reducing the size of the British and
American zones and sectors.
The proposed boundaries of the occupation sectors in Berlin underwent several changes over
time. The key principles underlying the drawing of these boundaries were that the sectors should
be geographically-orientated to correspond with the occupation zones (with the Soviets in the East
and the Western Allies in the West); the boundaries of the occupation sectors should respect the
boundaries of the existing administrative districts (“Bezirke”) of Berlin; and the sectors should be
approximately equal in population (prior to the creation of the French sector from part of the British
sector). The final agreement in July 1945 allocated six districts to the American sector (31 percent of
the 1939 population and 24 percent of the area), four districts to the British sector (21 percent of the
1939 population and 19 percent of the area), two districts to the French sector (12 percent of the 1939
population and 12 percent of the area), and eight districts to the Soviet sector (37 percent of the 1939
population and 46 percent of the area).7
The protocol envisioned an Allied Control Council to facilitate co-operation between the occupation zones of Germany and a Kommandatura to fulfil the same function for the occupation sectors
of Berlin. While Berlin was located around 200 kilometers within the Soviet zone, the protocol made
no separate provision for access to Berlin from the Western zones. Instead access by road, rail and
air to Berlin was negotiated between the military commanders of the Allied armies to coincide with
the withdrawal in July 1945 of the Western armies to the boundaries of the occupation zones agreed
in the protocol. With the onset of the Cold War, the relationship between the Western allies and the
7
In delineating the sectors, the boundaries of Berlin as a whole were based on the 1920 definition of Greater Berlin
(“Gross Berlin”), while the boundaries of the individual districts were based on the amendments to the 1920 district boundaries that came into effect in March 1938. For further discussion of the various proposals for sectors that were considered,
see Sharp (1973).
6
Soviet Union began to deteriorate. Growing tensions about in particular plans for closer economic
integration in the Western zones and sectors led to the collapse of the Allied Control Council and
Kommandatura and the creation of separate city governments for the Western and Soviet sectors of
Berlin. Following a currency reform initiated by the Western allies in June 1948, the Soviet Union
blockaded road and rail access to Berlin. As a result, for nearly a year the Western sectors of the city
were supplied by air until the restrictions on road and rail access were lifted in May 1949.8
This sequence of events led in 1949 to the foundation of West Germany on the American, British
and French occupation zones and the foundation of East Germany on the Soviet occupation zone.
Berlin remained formally occupied by the war-time allies, although West and East Berlin functioned
in many respects as de facto parts of East and West Germany respectively. From 1952 onward, a
sophisticated system of border fences and other barriers was constructed on the eastern side of the
border between East and West Germany to prevent civilians escaping from East Germany. While this
border between the two Germanies was completely sealed, travel between East and West Berlin remained feasible, which led to a continuous stream of refugees passing through West Berlin on route
from East to West Germany. To stem this flow of refugees, the East German authorities constructed
the Berlin Wall in August 1961, which separated West Berlin from its surrounding economic hinterland in East Germany and from East Berlin.9 The construction of the Berlin Wall ended all local
economic interaction between East and West Berlin and severed twelve underground and suburban
rail lines that were blocked off at the sector boundaries.10
The Berlin Wall cut through the heart of the pre-war city and divided the Western sectors from the
city’s pre-war CBD in the district of “Mitte” in the Soviet sector.11 In the aftermath of its construction,
a new CBD in West Berlin began to emerge that was centered on the Kudamm (“Kurfürstendamm”)
near the Berlin Zoo and the Kaiser Wilhelm Memorial Church in Charlottenburg, which had developed into a fashionable shopping area in the years leading up to the Second World War. This change
in city structure was reflected in a re-orientation of the transport network, with the opening of new
suburban rail lines and the construction of major North-South highways either side of the new CBD
to connect the Northern and Southern portions of West Berlin.
8
A formal agreement on access routes from West Germany to West Berlin was only reached in 1971, with the signing of
the Four Power Agreement of September 1971 and the subsequent Transit Agreement (“Transitabkommen”) of December
1971. This Transit Agreement designated a small number of road, rail and air corridors from West Germany to West Berlin.
9
The Statistical Yearbook of West Germany reports that 257,308 East German refugees left West Berlin by plane in 1953
following the violent uprisings in June of that year. During 1954-60 this stream of East German refugees departing from
West Berlin by plane continues at a rate of approximately 95,000 people per year and ceases with the construction of the
Berlin Wall in 1961.
10
In the immediate post-war years, it was estimated that around 400,000 people crossed the sector boundaries daily
(Robinson 1953), and both underground and suburban rail ran across the sector boundaries. With the construction of the
Berlin Wall, this traffic completely ceased. The only infrastructure link that remained between the two halves of the city
was the sewage system, which became a source of foreign currency revenue for the East German economy.
11
As well as being the site of pre-war Berlin’s main administrative, cultural and educational institutions, Mitte also
contained two-thirds of its insurance firms and three-quarters of its banks (Merritt 1973).
7
During the post-war period, the West German government introduced a variety of policies designed to offset West Berlin’s peripheral status as an enclave surrounded by East German territory.
Industries located in West Berlin benefited from substantial West German subsidies for goods transported to and from West Berlin. Incentives were provided for residence in West Berlin and public
administration was located there (see for example Ellger 1992). Nonetheless, West Berlin’s total population declined during the period of division, with the result that even today the population of
Berlin as a whole is around one million less than its 1939 population of 4.3 million.
While the division of Germany and Berlin appeared to be permanent, the Soviet policies of “Glasnost” and “Perestroika ” introduced by Mikhail Gorbachev in 1985 started a process of opening up of
Eastern Europe.12 As part of this wider transformation, large-scale demonstrations in East Germany
in 1989 led to the fall of the Berlin Wall on 9 November 1989. In the aftermath of these events, the
East German system rapidly began to disintegrate. Only eleven months later East and West Germany
were formally reunified on 3 October 1990. In June 1991 the German parliament voted to relocate the
seat of the parliament and the majority of the federal ministries back to Berlin. As the integration of
West and East Berlin proceeded, transport arteries between East and West were re-connected. Areas
such as Potsdamer Platz, which was located close to the pre-war CBD just West of the Berlin Wall,
have once again become major centers of commercial development.
3
Theoretical Model
To guide our empirical analysis, we use a canonical model of cities that emphasizes the tension between production externalities (which favor the concentration of economic activity) and commuting
costs and an inelastic supply of land (which favor the dispersion of economic activity). We extend
the model of LRH, which has the key advantages for our empirical analysis of modeling the city in
two-dimensional space and of not imposing mono-centricity but instead allowing for a rich structure
of economic activity within the city. While LRH develop a model of a circular city and focus on symmetric equilibria in which the radius perfectly characterizes city structure, we allow for asymmetries
in consumption amenities and natural advantages across locations as observed in our data. To ensure that the model remains tractable despite these asymmetries, we incorporate heterogeneity in
workers’ commuting choices following Eaton and Kortum (2002). The resulting framework is highly
tractable and amenable to quantitative analysis. While we focus on Berlin’s division and reunification, this framework provides a natural platform for evaluating other interventions in different
12
After the signing of the Basic Treaty (“Grundlagenvertrag ”) in December 1972, which recognized “two German states
in one German Nation”, East and West Germany were accepted as full members of the United Nations. West German
opinion polls in the 1980s show that less than 10 percent of the respondents expected a re-unification to occur during their
lifetime (Herdegen and Schultz 1993).
8
contexts, including for example the construction of new transport infrastructure.13
3.1
Preferences and Endowments
We consider a city embedded within a larger economy, which corresponds to pre-war Germany
before division, West Germany after division, and modern-day Germany after reunification. The city
produces a single final good, which is sold to (or purchased from) the larger economy at a competitive
price. The final good is traded at zero cost and is chosen as the numeraire, so that p = 1.14 The city is
comprised of a set of discrete blocks indexed by i = 1, ..., S. The effective supply of land within each
block, Li , depends on geographic area and building density and can vary across blocks, with the total
P
effective supply of land for the city as a whole equal to L̄ = Si=1 Li . Land within each block can be
used either commercially or residentially, and we denote the fractions of land allocated to commercial
and residential use by θi and 1 − θi respectively. The land market is perfectly competitive and rent is
accrued by absentee landlords and not spent within the city.15 The city is populated with a mass of H̄
workers, who are mobile across blocks within the city and between the city and the larger economy.
Workers choose their blocks of residence and employment to maximize their utility taking as given
goods and factor prices and the location decisions of other workers and firms.
Each worker is endowed with one unit of labor, which is supplied inelastically with zero disutility.
Workers are risk neutral such that preferences are linear in an aggregate consumption index: U =
C. This aggregate consumption index is defined over consumption of a single final good (c) and
residential land (`), and is assumed for simplicity to take the Cobb-Douglas form:16
C (ci , `i ) = Bi cβi `1−β
,
i
0 < β < 1.
(1)
where Bi captures exogenous differences in block-specific consumption amenities (such as green
spaces and leafy suburbs) that make the block a more or less pleasant place to live.
City blocks are connected by a bilateral transport network, which workers can use to commute
to work. Income net of commuting costs for a worker residing at block i and working at block j, vij ,
depends on the worker’s wage, wj , and effective units of labor at her block of employment. Effective
units of labor depend on the loss of labor time in commuting between blocks i and j, dij , and an
idiosyncratic stochastic shock, z, which varies for each worker residing in block i across possible
13
A more detailed exposition of the model is contained in a web-based technical appendix.
Even during division, there was substantial trade between West Berlin and West Germany, which was facilitated by
West German subsidies of transportation to and from West Berlin. In 1963, the ratio of exports to GDP in West Berlin was
around 70 percent, and industry accounted for around 50 percent of West Berlin’s GDP (American Embassy 1965).
15
While the assumption of absentee landlords follows LRH and is standard, we could alternatively assume that land
rent is redistributed lump sum to workers.
16
For empirical evidence using US data in support of the constant housing expenditure share implied by the CobbDouglas functional form, see Davis and Ortalo-Magne (2008).
14
9
blocks of employment j:
vij (zωij ) =
zωij wj
,
dij
(2)
where ω indexes individual workers. Commuting costs are therefore modeled as forgone labor earnings and the wage is measured per unit of effective labor. The labor time lost in commuting between
blocks i and j depends on transport connections: dij = eκτ ij ≥ 1, where τ ij denotes travel time and
κ > 0 parameterizes commuting costs.
The stochastic shock zωij captures idiosyncratic variation in the suitability of each worker ω residing in a block i for production at each possible block of employment j. One interpretation is that
the production technology for the final good varies across blocks of employment and z captures heterogeneous worker characteristics that influence worker productivity with the different final goods
production technologies.17 The stochastic shock z introduces heterogeneity in workers’ commuting
decisions, which prevents corner solutions where small differences in wages net of commuting costs
induce all workers residing in one block to commute to another block. As a result, bilateral commuting flows exhibit a gravity relationship, which has been found empirically to provide a close
approximation to observed commuting flows.18
Workers make location decisions before the uncertainty about their effective units of labor is
realized. Specifically, they choose their block of residence before observing the realizations of effective units of labor for each possible block of employment.19 Having made her location decision, each
worker draws her effective units of labor for each block of employment from an independent Fréchet
distribution with scale parameter T and shape parameter . The cumulative distribution function for
effective units of labor is therefore:
−
Fij (z) = e−T z ,
T > 0, > 1,
(3)
where for simplicity we assume that the scale parameter T is the same for all workers and blocks of
residence and employment.
Once a block of residence has been chosen and the realizations of effective units of labor for each
block of employment are observed, each worker chooses her block of employment, residential land
use and goods consumption to maximize her utility. In general, workers residing within a given
17
While we interpret z as an idiosyncratic shock to a worker’s suitability for production at a particular employment
location, it has an equivalent interpretation in (2) as an idiosyncratic shock to a worker’s commuting costs between a given
block of residence and a given block employment.
18
For empirical evidence that bilateral commuting flows are well approximated by a gravity equation, see for example
Erlander and Stewart (1990) and Sen and Smith (1995). For empirical evidence for pre-war Berlin on the role of distance in
dampening commuting flows, see Feder (1939).
19
In this specification, workers choose their block of residence before their block of employment. It is straightforward
to consider the reverse case where the block of employment is chosen first, as long as when the block of employment is
chosen there is some uncertainty about commuting costs to each block of residence.
10
block experience different ex post realizations of effective units of labor and hence choose different
blocks of employment and receive different incomes net of commuting costs.
Population mobility implies that workers must have the same ex ante expected utility across all
blocks of residence that are populated in equilibrium. Combining population mobility with the firstorder conditions for utility maximization, we obtain the following expression for equilibrium residential rents:
Q1−β
= ū−1 β β (1 − β)1−β Bi E [vi ] ,
i
(4)
where E [vi ] is expected worker income net of commuting costs and ū denotes workers’ reservation
level of expected utility in the larger economy. From this population mobility condition, variation in
equilibrium residential rents across blocks reflects differences in exogenous consumption amenities,
Bi and differences in endogenous expected worker income, E [vi ].
Using the Fréchet distribution for worker productivity, expected worker income net of commuting costs can be evaluated as:
E [vi ] =
1/
Φi Γ
−1
"
Φi ≡
,
S
X
#
T (ws /dis ) ,
(5)
s=1
where Γ is the Gamma function. Therefore expected worker income is higher for residential blocks
with good transport connections to employment locations that pay high wages.
As noted above, another implication of the Fréchet distribution for worker productivity is that
bilateral commuting flows satisfy a gravity equation. The probability that a worker residing in block
i works in block j, π ij , depends on wages at block j and bilateral transport connections between
i and j (“bilateral resistance”), but also depends on wages and transport connections for all other
possible blocks of employment (“multilateral resistance”):
(wj /dij )
π ij = PS
.
s=1 (ws /dis )
3.2
Production
The final good is produced under conditions of perfect competition and according to a constant
returns to scale technology. For simplicity, we assume that this production technology is CobbDouglas, and hence the aggregate amount of final goods output produced in block j is:
α
]
Xj = Aj H
(θj Lj )1−α ,
Mj
]
where Aj denotes final goods productivity; H
M j is effective employment, which depends on the
measure of workers employed and the effective units of labor of each worker.
11
Final goods productivity, Aj , varies across blocks depending on both exogenous natural advantages, aj , and endogenous agglomeration forces. Following LRH and a long literature in urban
economics, we assume that these agglomeration forces take the form of knowledge spillovers that
are increasing in the surrounding density of effective employment:20
" S
!#γ
X
]
H
Ms
e−δτ js
Aj = a j
,
0 < γ < 1,
Ls
(6)
s=1
where γ captures the relative importance of knowledge spillovers in influencing productivity; δ > 0
parameterizes the geographical localization of knowledge spillovers; and 0 ≤ e−δτ js ≤ 1. Thus dense
concentrations of employment in nearby blocks generate greater potential for knowledge spillovers
than sparse concentrations of employment in far-away blocks.
Firms choose their block of production, effective employment and commercial land use to maximize their profits taking as given goods and factor prices, productivity and the location decisions
of other firms and workers. From the first-order conditions for profit maximization and the requirement that zero profits are made if the final good is produced, equilibrium commercial rent for each
block with positive employment is determined as:
qj = (1 − α)
α
wj
α
1−α
1
Aj1−α .
(7)
Therefore higher productivity, Aj , and lower wages, wj , both make a block more attractive for production and are reflected in higher commercial rent (7).
Equilibrium wages are determined by labor market clearing, which requires that payments for
effective labor input equal income net of commuting costs. Using workers’ commuting choices, this
labor market clearing condition can be written as:
]
wj H
Mj =
S
X
i=1
(wj /dij )
hP
S
s=1 (ws /dis )
i1− 1 Γ
−1
HRi ,
where recall > 1. Finally, the equilibrium allocation of land to commercial and residential use
in each block is determined by land market clearing and no-arbitrage between alternative uses of
land. Therefore land is allocated to whichever use offers the higher rent, and in order for positive
fractions of land to be allocated to both commercial and residential use, we require that commercial
and residential rents are equalized.
We use the model to examine the impact of the construction of the Berlin Wall on the organization
of economic activity within West Berlin. We focus on West Berlin, since it remained a market-based
economy after division and we would therefore expect the mechanisms in the model to apply. As
20
While in focusing on knowledge spillovers we adopt the leading formulation of agglomeration forces in the urban
economics literature, there are other possible formulations, as discussed for example in Duranton and Puga (2004).
12
a result of Berlin’s division, firms in West Berlin cease to benefit from production externalities from
East Berlin and can no longer employ commuters from East Berlin. Similarly, residents in West Berlin
can no longer commute to East Berlin. Since both production externalities and commuting costs vary
with transport connections, these effects vary across areas within West Berlin with the change in
their transport connections as a result of division. First, areas of West Berlin close to employment
concentrations in East Berlin experience the largest reductions in productivity from lost production
externalities. Second, areas of West Berlin close to residential concentrations in East Berlin experience
the largest increases in the wages needed to attract commuters. Third, areas of West Berlin close
to employment concentrations in East Berlin experience the largest reductions in expected worker
income due to lost commuting possibilities.
The construction of the Berlin Wall therefore changes relative productivity, wages and expected
worker income across areas within West Berlin, which in turn lead to changes in the distributions of
employment and residents. The mechanism that restores equilibrium is changes in the rent gradient
until zero profits are made in each location with positive production, workers are again indifferent
across all locations populated in equilibrium and land is allocated to the highest return use. Therefore
the model has the following key prediction which we examine in our empirical analysis: areas of
West Berlin that previously had good transport connections to dense concentrations of economic
activity in East Berlin experience reductions in rents relative to other areas of West Berlin as a result
of division.
In examining the model’s predictions, one important issue is whether equilibrium city structure
is unique or there are instead multiple equilibria. In order for the model to have a non-degenerate
distribution of economic activity, we require that the agglomeration forces of knowledge spillovers
are not too large relative to the dispersion force of an inelastic supply of land, which corresponds
to the parameter inequality 0 < γ < 1 − α.21 As we observe a non-degenerate distribution of economic activity in the data, we focus on parameter values satisfying this inequality. For this range of
parameter values, the symmetric city model of LRH can exhibit multiple equilibria because of the
endogenous dependence of productivity on knowledge spillovers, but exhibits a unique equilibrium
when productivity is assumed to be exogenous. In the version of the model developed here, the
potential for multiple equilibria depends on both the importance of knowledge spillovers (as γ → 0
productivity becomes exogenous and the equilibrium is unique) and the extent of asymmetries in
for example natural advantage and consumption amenities (which can also render the equilibrium
unique).
Whether or not there are are multiple equilibria is crucial for the extent to which Berlin’s division
21
This parameter inequality corresponds to the “no black hole” condition in Fujita et al. (1999).
13
has a permanent impact on city structure. In the presence of a unique equilibrium, the same economic fundamentals should re-assert themselves after reunification as prior the Second World War.
Therefore we would expect a convergence in the organization of economic activity within Berlin after
reunification towards the pattern observed prior to the Second World War, though this convergence
may be incomplete to the extent that economic fundamentals changed between the two periods. In
contrast, in the presence of multiple equilibria, the division of Berlin could have a permanent effect
on the organization of economic activity within Berlin by shifting city structure between multiple
equilibria.
We use these qualitative predictions of the model to guide our empirical analysis of the data.
After establishing the main empirical features of the organization of economic activity within Berlin
before, during and after the city’s division, we return to the model to compare its quantitative predictions with the patterns observed in the data.
4
Data Description
The key advantages of our focus on Berlin are the exogenous variation provided by division and
reunification and the rich source of contemporary and historical data available. In particular, to estimate the strength of agglomeration and dispersion forces in the model, we require data on land
rents, employment by workplace and employment by residence at a fine level of spatial disaggregation. These data are available for Berlin at the level of thousands of statistical blocks over a period of
some seventy years.22 Our main dataset is a balanced panel containing 5,162 blocks in West Berlin
for the years 1936, 1986 and 2006.23 Blocks have an average area of 166 meters squared and an average population of 333 in 2006. We combine these data on rents and employment with Geographical
Information Systems (GIS) data on the boundaries of each block in the form of an ArcGIS shapefile.
The block boundaries are in turn combined with GIS data on transport connections, including the
suburban and underground rail network, the road network, the location of lakes, rivers and canals,
the location of forests and parks, the extent of destruction during the Second World War, and the
location of schools.24
The main source for the 1986 and 2006 data is the Office for City Development in Berlin (“Senatsverwaltung für Stadtentwicklung Berlin”), which imputes rents for each developed block and
22
Data are available at increasing levels of disaggregation for districts (“Bezirke”), statistical areas (“Gebiete”) and statistical blocks. West Berlin consists of twelve districts, with around eight areas per district and ninety one blocks per
area.
23
We abstract from East Berlin during the period of division, since the process of central planning is unlikely to mimic
market forces, and data at the same level of spatial disaggregation are not readily available.
24
Around twenty percent of the area of West Berlin is covered by forest and parks and around six percent is covered by
lakes and rivers (Friedensburg 1967). As in the model, blocks in the data can differ substantially in terms of their access to
these consumption amenities.
14
compiles data on population and employment in each block from the censuses of population and
occupation.25 The main sources for the historical data on rents are a handbook compiled by Kalweit
(1937) and a map constructed by Runge (1950). Using the Runge map and the Kalweit handbook,
we locate streets relative to the boundaries of blocks and evaluate rent for each block as the average
rent of all the streets that fall within its boundaries.26 In each year, we normalize rents by their mean
so that the resulting distribution of normalized rents has a mean of one. The main sources for the
historical data on population and employment are the 1933 population and employment censuses,
which report detailed breakdowns for Greater Berlin. To evaluate the impact of division and reunification on the transport network, we modify our ArcGIS shapefile for the 2006 transport network
using historical maps and documents.
Another advantage of focusing on Berlin is that its geographical boundaries are relatively well
defined over our time period. The 1920 law that defined Greater Berlin encompassed an extensive
area of 883 square kilometers, which diminished net commuting into Greater Berlin prior to the
Second World War.27 During the period of division, the 1920 law provided the basis for the outer
boundaries of West Berlin, which were completely sealed to local economic interactions after the
construction of the Berlin Wall. During the period since re-unification, the boundaries of Greater
Berlin have remained relatively well defined, partly because the overall size of the city population is
less than in the period prior to the Second World War.
While for the periods 1986 and 2006, our data comprise the entire areas of West Berlin and Greater
Berlin respectively, our 1936 data on land values and hence the balanced panel are restricted to the
main area of residential and commercial development. This area includes the Wilhelminian Ring
(“Wilhelminischer Ring”) of dense urban development that corresponds approximately with the “SBahn Ring,” as well as its surroundings as shown in Map 1. Comparing the areas of Berlin included
and excluded from the balanced panel either using our 1930s data on other characteristics (i.e. population and employment) or using our 1986 and 2006 data on land values and other characteristics
reveals that the balanced panel captures the vast bulk of economic activity in Greater Berlin and there
are no major economic centers in the areas outside the balanced panel.
25
Undeveloped blocks include forests and parks. While these undeveloped blocks themselves contain no economic
activity, they affect general equilibrium in the quantitative analysis of the model, because they influence the distance
traveled between developed blocks.
26
For further discussion concerning the data definitions and sources, see the data appendix. While Kalweit (1937) reports
rent values for 1936, as a robustness check, we compare them with the rent values for 1929 reported in Kalweit (1929).
27
For Greater Berlin in 1939, employment by workplace was 2,171,891, while employment by residence was 2,302,640.
15
5
Empirical Analysis
One of the model’s key predictions is that areas of West Berlin close to concentrations of economic
activity in East Berlin should experience a decline in rents relative to other areas of West Berlin following the construction of the Berlin Wall. To examine this prediction, we first characterize the evolution of the rent gradient in Berlin over time, before next estimating a “difference-in-differences”
econometric specification that controls for other potential determinants of rents.
5.1
Evolution of the Rent Gradient over Time
A distinctive feature of the model is that it allows a rich internal organization of economic activity
within the city without imposing mono-centricity. To examine the rent gradient empirically without
imposing a prior structure on the data, Map 1 displays the distribution of rents in 1936 across five
discrete classes, which are chosen to group together similar rent values and maximize the differences
in rent values between classes.28 As apparent from the map, Berlin’s rent gradient in 1936 was in
fact approximately mono-centric, with the highest rent values concentrated in the district Mitte just
East of the Berlin Wall. Based on this rent gradient, we determine the central point of the pre-war
Central Business District (CBD) as the intersection of Friedrich Str. and Leipziger Str., which lies
close to the U-Bahn station “Stadtmitte” and approximately one kilometer East of where the Berlin
Wall intersected Leipziger Str. at Potsdamer Platz.
Around this central point, there are concentric rings of progressively lower rents, which define
an ellipse around the pre-war CBD in Map 1. From the path of the Berlin Wall shown in Map 1, it is
clear that areas of West Berlin were differentially affected by division. For example, Kreuzburg was
particularly adversely affected, losing access to Mitte immediately North along a wide stretch of the
Berlin Wall. Similarly, Wedding lost access to Mitte immediately South along another stretch of the
Berlin Wall. In contrast, the white area immediately West of Mitte is the Tiergarten park. As a result
of this area of open space, Charlottenburg and Wilmersdorf were further away from the pre-war
CBD, and hence were less adversely affected by the loss of access to it.29 Just to the South-West of
the Tiergarten in Wilmersdorf, the Kudamm (“Kurfürstendamm”) had developed into a fashionable
shopping area prior to the Second World War. As this area was both part of the ellipse of concentrated
development surrounding the pre-war CBD and relatively centrally located within West Berlin, this
area was a natural possible location for the emergence of West Berlin’s CBD during division.
To provide another perspective, Panel A of Figure 1 displays Berlin’s 1936 rent gradient across
28
As discussed in the data section above, rents are normalized to have a mean of one in each year, and hence we focus
on relative changes in rents within the city over time.
29
West Berlin’s East-West and North-South axes are approximately 30 kilometers long and the Westernmost edge of the
Tiergarten is approximately 3 kilometers from the Berlin Wall.
16
blocks in three dimensions using a latitude and longitude grid.30 Again the pre-war CBD in Mitte is
evident as the highest rent peak. Also evident are the concentric rings of progressively lower rents
around the pre-war CBD, including the area around the Kudamm, as well as the area of open space
in between the Kudamm and Mitte. As a point of comparison, Panel B of Figure 1 displays the 1936
rent gradient only for blocks in West Berlin. As apparent from this panel, the two areas of West Berlin
with the highest pre-war rents were parts of the concentric ring around the pre-war CBD: the area
around the Kudamm and the area just West of Potsdamer Platz and the Berlin Wall.
To illustrate the impact of division, Panel C of Figure 1 displays the 1986 rent gradient across
blocks within West Berlin. A number of striking features are apparent. First, while the mean rent in
each year is equal to one by construction, the standard deviation of rents within West Berlin is much
smaller during division than prior to the Second World War. As division substantially reduced the
total population and overall scale of economic activity in the city, this reduced dispersion of rents
is consistent with reduced production externalities in the model. Second, the loss of access to the
pre-war CBD entirely eliminated one of pre-war rent peaks in West Berlin: the area just West of
Potsdamer Platz. Third, West Berlin’s CBD coincided with the other pre-war rent peak around the
Kudamm, which benefited from its relatively central location in West Berlin. Both of these features
are consistent with the role played by the surrounding concentration of economic activity in determining the rent gradient in the model. Fourth, in the aftermath of division, a much smaller rent
peak began to form in the area around Schlossstrasse, in between the universities of West Berlin and
Tempelhof airport and close to the affluent suburb of Dahlem.
To illustrate the impact of reunification, Panel D of Figure 1 displays Berlin’s rent gradient in
2006. Again several striking features are apparent. First, while the mean rent is again equal to one by
construction, there is a dramatic increase in the standard deviation of rents following reunification
towards levels observed during the pre-war period. Second, there is a re-emergence of the second
rent peak in West Berlin in the area just West of Potsdamer Platz. Third, following reunification, the
pre-war CBD in Mitte has again become a major center of economic activity. Fourth, while Mitte and
the area just West of Potsdamer Platz had developed into major rent peaks by 2006, they remained
dominated by the rent peak in the area around the Kudamm in contrast to the pattern observed prior
to the Second World War. Fifth, in the aftermath of re-unification, the much smaller rent peak in the
area around Schlossstrasse remains apparent, although its relative magnitude compared to the rent
peak around the Kudamm is diminished.
Taken together, these findings suggest that division and reunification did indeed impact differ30
To construct the figure, blocks are first arrayed on a discrete grid of around 9,000 points of 0.0025 intervals of latitude
and longitude. A surface is next constructed through the points in the discrete grid using linear (triangular) interpolation.
The surface passes through the observations for each block and exhibits the same pattern for a wide range of intervals for
the discrete grid.
17
entially on areas of West Berlin depending on their location relative to concentrations of economic
activity in East Berlin. There is also some evidence that following reunification pre-war patterns of
economic activity within Berlin began to reemerge. In the next section, we examine the robustness of
these findings to controlling for other potential determinants of rents in addition to location relative
to pre-war concentrations of economic activity in East Berlin.
5.2
Difference-in-Differences Estimates
Two sets of determinants are important for rents in the model. First, there are exogenous differences
across blocks in natural advantage and consumption amenities. Second, there are endogenous differences across blocks in accessibility to production externalities and commuting possibilities, which
depend on the transport network and the concentration of employment and residents within each
block. To capture these two sets of determinants, we consider the following empirical specification:
ln Qit = η i + f (ait ) + ln Xi µt + ν t + uit ,
(8)
where i denotes blocks and t corresponds to time; η i is an unobserved block fixed effect; f (ait ) is an
arbitrary function of block accessibility, ait , to concentrations of economic activity within the city; Xi
is a matrix of observable block characteristics and µt are time-varying coefficients to be estimated; ν t
is a time fixed effect; and uit is a stochastic error.
To examine the impact of division and reunification, we take long differences in (8) to obtain the
following cross-section regression specification:
4 ln Qi = ψ + g(ait , ait−T ) + ln Xi ζ + χi ,
(9)
where 4 is the difference operator and T is the time interval for the differencing; ψ = ν t − ν t−T is the
regression constant; g(ait , ait−T ) is a function of accessibility in the two time periods; ζ = µt − µt−T ;
and χi = uit − uit−T .
A key empirical challenge in estimating (9) is that changes in rents and accessibility are typically
jointly and endogenously determined. To address this challenge, we exploit the exogenous source
of variation in accessibility provided by Berlin’s division and reunification. We model the change in
accessibility of each block in West Berlin as a non-parametric function of distance from the pre-war
CBD East of the Berlin Wall, which yields our baseline econometric specification:
4 ln Qi = ψ +
J
X
dij ξ j + ln Xi ζ + χi ,
(10)
j=1
where dij is a (0, 1) dummy variable which equals one if block i lies within distance grid cell j and
zero otherwise; and ξ j is a coefficient to be estimated for each distance grid cell j. We begin by
18
considering distance grid cells of one kilometer intervals from one to six kilometers from the prewar CBD, where the excluded category is blocks more than kilometers from the pre-war CBD.31 We
discuss the robustness of the results to alternative specifications below. We estimate this specification
both for division (taking differences between 1936 and 1986) and for reunification (taking differences
between 1986 and 2006). To allow the errors for neighboring blocks to be correlated, we cluster the
standard errors by the 69 statistical areas (“Gebiete”) in our sample.32
This baseline econometric specification has a “differences-in-differences” interpretation, where
the first difference is over time and the second difference is between areas of West Berlin at differing
distances from the pre-war CBD East of the Berlin Wall. The key coefficients of interest ξ j capture the
treatment effects of division and reunification on blocks in West Berlin proximate to the pre-war CBD.
Our specification allows for a flexible functional form for the relationship between rent growth and
distance from the pre-war CBD. We allow for time-invariant unobserved determinants of the level of
block rents, which are differenced out when we take long differences. We also allow observable block
characteristics to have time-varying effects on rents. We consider a wide range of observable block
characteristics as controls, including block area, the average number of floors per building in the
block, distance to the nearest U-Bahn station, distance to the nearest S-Bahn station, distance to the
nearest park, distance to the nearest canal, lake or river, distance to the nearest school, the percentage
of the block’s area destroyed during the Second World War, and the district of Berlin in which the
block is located. While we wish to demonstrate the robustness of our results to the inclusion of these
controls, some of them could themselves be affected by division, and hence we report results both
with and without these controls.
Table 1 reports the results of estimating this baseline specification for division (Columns (1)-(4))
and reunification (Columns (5)-(8)). In Column (1), which includes only the distance grid cells, the
estimated treatment effect of division is negative, highly statistically and declines monotonically
with distance from the pre-war CBD. From the estimated coefficients, blocks in West Berlin within
one kilometer of the pre-war CBD experienced a 300 percent reduction in relative rents between
1936 and 1986 compared to other blocks in West Berlin more than six kilometers from the pre-war
CBD. In contrast, blocks within 2-3 kilometers of the pre-war CBD experience a reduction in relative
rents of around half this magnitude. Together the six distance grid cells alone explain around half
of the variation in rent growth following division (R2 = 0.49), which suggests a powerful effect of
proximity to the surrounding concentration of economic activity. In Column (2), we include district
31
Overall, 1,743 of the 5,162 blocks in our sample are within six kilometers of the pre-war CBD. Eleven blocks lie within
the first distance grid cell, eight-nine blocks lie within the second distance grid cell, and more than two hundred lie within
each of the remaining grid cells. The maximum distance to the pre-war CBD in our sample is over 15 kilometers.
32
Bertrand et al. (2004) examine several approaches to serial correlation and show that clustering the standard errors
performs well in settings with at least 50 clusters as in our application.
19
fixed effects to control for differences in policies across the districts and across the occupation sectors
that were defined based on district boundaries. While districts lie at differing distances from the
pre-war CBD, this specification exploits only the variation in distance within districts. Nonetheless,
we continue to find a strong and statistically significant division treatment.
In Column (3), we further augment the specification with our full set of controls for observable
block characteristics. In this specification too, we find a strong and statistically significant division
treatment, although the estimated coefficients fall in magnitude. In Column (4), we examine other
measures of proximity by including distance grid cells based on proximity to the inner boundary
between East and West Berlin and the outer boundary that separated West Berlin from its surrounding hinterland in East Germany. The distance grid cells for the pre-war CBD and inner boundary
are separately identified, because the former is measured relative to a fixed point in the center of
East Berlin, while the latter is measured relative to the closest point on the inner boundary between
East and West Berlin. While we continue to find a division treatment of around the same magnitude
for the pre-war CBD, we find a very different pattern of results for the inner and outer boundaries.
In particular, there is no statistically significant division treatment for the outer boundary, and the
division treatment for the inner boundary is if anything positive and of a much smaller magnitude.
This pattern of results is reassuring because it suggests that the reorientation of West Berlin’s rent
gradient following division did indeed reflect a loss of access to existing concentrations of economic
activity rather than other disamenities associated with proximity to the Berlin Wall.33
In Columns (5)-(8), we find the reverse pattern of results for reunification, confirming the beginning of a re-emergence of pre-war patterns of economic activity in Berlin noted above. The reunification treatment is positive, highly statistically and declines monotonically with distance from
the pre-war CBD. In Column (5), which includes only the distance grid cells, we find that blocks in
West Berlin within one kilometer of the pre-war CBD experienced a 170 percent increase in relative
rents between 1986 and 2006 compared to other blocks in West Berlin more than six kilometers from
the pre-war CBD. When this specification is augmented district fixed effects (Column (6)) or further
augmented with our full set of controls for block characteristics (Column (7)), we continue to find
the same pattern of results. In Column (8), we show that the reunification treatment is driven by
proximity to the pre-war CBD rather than distance to the inner and outer boundaries of West Berlin.
Again this provides further support for an explanation based on the surrounding concentration of
economic activity rather than proximity the Berlin Wall per se.
33
While a negative treatment effect of proximity to the inner and outer boundaries of West Berlin could also in principle
be explained in terms of a loss of access to surrounding economic activity, the positive estimated treatment effects suggest
that this was not important in practice. This is consistent with most commuting in pre-war Greater Berlin occurring along
radial transport arteries to the city center and the relatively small levels of pre-war commuting from beyond the boundaries
of Greater Berlin.
20
We have also undertaken a number of other robustness checks, which in the interests of brevity
are not reported in the paper. For example, we find a similar pattern of results when we include
a quadratic in observable block characteristics to allow for a more flexible functional form for the
controls. We also find a similar pattern of results if we re-estimate the specifications in Table 1 excluding blocks in which government buildings are located to address concerns about government
construction. These findings are in line with the observation above that one of the key features of the
evolution of the rent gradient since reunification is the re-emergence of the rent peak in West Berlin
immediately West of Postdamer Platz. This emerging rent peak is driven by private-sector development in and around the Sony Center and not by government construction, which is concentrated
further North in the area around the parliament building (“Reichstag”).
While the distance grid cells specification enables us to flexibly explore the impact of proximity
to the pre-war CBD and control for other potential determinants of rents, we find a similar pattern
of results using other related non-parametric techniques. In Figures 5 and 6, we display normalized
rent growth for each West Berlin block in our sample against distance from the pre-war CBD and the
fitted values from a locally-weighted linear least squares regression for the periods of division and
reunification respectively. As evident from the figures, there is a strong and non-linear relationship
between rent growth and distance from the pre-war CBD, which extends to a distance of around six
kilometers as used in the grid cells specification above.
6
Quantitative Analysis of the Model
7
Conclusions
21
A
Data Appendix
Pre-Second World War data: Data on land rent by street in 1936 in Reichsmark per square meter are
from Kalweit (1937) Die Baustellenwerte in Berlin, Berlin: Carl Heymanns Verlag. Data on population by street in 1933 are from Mitteilungen des Statistischen Amts der Stadt Berlin, No. 18, Volks-,
Berufs- und Betriebszaehlung in Berlin 1933. Data for each street were matched to statistical blocks.
Land rent for each statistical block is the average of land rents for all streets within that block. Land
rents are normalized to have a mean across statistical blocks of one. Population for each statistical
block is the sum of the population of all streets within that block. Data on employment by workplace
and residence by district (“Bezirk”) are from Mitteilungen des Statistischen Amts der Stadt Berlin,
No. 18, Volks-, Berufs- und Betriebszaehlung in Berlin 1933.
Division data: Data on land rent by statistical block in 1986 in Deutsche Mark per square meter is
from Der Senator für Bau- und Wohnungswesen (1987) Bodenrichtwerte 31.12. 1986. Land rents
are normalized to have a mean across statistical blocks of one. Data on population, employment by
workplace and employment by residence by statistical block are from from Senatsverwaltung für
Stadtentwicklung Berlin.
Re-unification data: Data on land rent by statistical block in 2006 in Euro per square meter is from
Senatsverwaltung für Stadtentwicklung Berlin – Abt. III, Geoinformation, Vermessung, Wertermittlung (ed.) (2006): Bodenrichtwerte 01.01.2006. Data on population, employment by workplace and
employment by residence by statistical block are from from Senatsverwaltung für Stadtentwicklung
Berlin.
Geographical Charactertisitics: Geographical Information Systems ArcGIS Shapefiles from Senatsverwaltung für Stadtentwicklung Berlin, Digitale Grundkarte 1:5000, DIGK5, 12/2005. Projection: Soldner
Berlinshape files. Geographical Characteristics include: location and land area of statistical blocks,
statistical areas (“Gebiete”) and districts (“Bezirke”); suburban (“S-Bahn”) and undergound (“UBahn”) rail network); location of parks and open spaces; location of schools. Historical suburban
and underground rail networks were constructed by amending the modern ArcGIS shapefiles using
historical maps of the transport network.
22
References
[1] Ahlfeldt, Gabriel M. and Nicolai Wendland (2008) “Die Zentralität war schon da! Die Bedeutung
des öffentlichen Schienennahverkehrs für die Entstehung der Berliner City West,” DISP 174.
[2] Alonso, W. (1964) Location and Land Use, Cambridge MA: Harvard.
[3] American Embassy (1965) Handbook of Economic Statistics: Federal Republic of Germany and the
Western Sectors of Berlin, Economic Affairs Section, Bonn.
[4] Arzaghi, Mohammad and J. Vernon Henderson (2008) “Networking Off Madison Avenue,” Review of Economic Studies, 75, 1011-1038.
[5] Bertrand, Marianne, Esther Duflo and Sendhil Mullainathan (2004) “How Much Should We
Trust Differences-in-Differences Estimators?” Quarterly Journal of Economics, 119(1), 249-275.
[6] Book, Tommy (1995) “The Urban Field of Berlin: Expansion–Isolation–Reconstruction,” Geografiska Annaler. Series B, Human Geography, 77(3), 177-196.
[7] Brakman, Steven, Harry Garretsen and Marc Schramm (2004) “The Strategic Bombing of German Cities during WWII and its Impact on City Growth,” Journal of Economic Geography, 4(2),
201–18.
[8] Ciccone, A. and R. E. (1996) “Productivity and the Density of Economic Activity,” American
Economic Review, 86, 54-70.
[9] Combes, Pierre-Philippe, Gilles Duranton, Laurent Gobillon and Sébastian Roux (2009) “Estimating Agglomeration Economies with History, Geology, and Worker Effects,” Chapter 1 in
(ed.) Edward L. Glaeser, The Economics of Agglomeration, Chicago: University of Chicago Press.
[10] Crafts, N. F. R. and T. Leunig (2009) “Agglomeration Economies from Better Transport: Nineteenth Century England and Wales,” London School of Economics, mimeograph.
[11] Davis, Morris A., and Franois Ortalo-Magn (2007) “Housing Expenditures, Wages, Rents.” CESifo Working Paper, 2156.
[12] Davis, Donald R. and David E. Weinstein (2002) “Bones, Bombs, and Break Points: The Geography of Economic Activity,” American Economic Review, 92(5), 1269–89.
[13] Dekle, R and J. Eaton (1999) “Agglomeration and Land Rents: Evidence from the Prefectures,”
Journal of Urban Economics, 46, 200-14.
23
[14] Desmet, Klaus and Esteban Rossi-Hansberg (2009) “On Spatial Dynamics,” Journal of Regional
Science, forthcoming.
[15] Duranton, Gilles and Diego Puga (2004) “Micro-foundations of Urban Agglomeration
Economies,” Chapter 48 in (eds.) J. V. Henderson and J.F. Thisse, Handbook of Regional and Urban
Economics, Volume 4, Amsterdam: Elsevier North Holland.
[16] Eaton, Jonathan and Kortum, Samuel (2002) “Technology, Geography, and Trade,” Econometrica,
70(5), 1741-1779.
[17] Ellger, Cristof (1992) “Berlin: Legacies of Division and Problems of Unification,” The Geographical
Journal, 158(1), 40-46.
[18] Erlander, Sven and Neil F. Stewart (1990) The Gravity Model in Transportation Analysis – Theory
and Extensions, Utrecht: VSP.
[19] Feder, Gottfried (1939) “Arbeitstätler - Wohnstätte,” Schriftenreihe der Reichsarbeitsgemeinschaft für Raumforschung an der Technischen Hochschule Berlin, No. 3, Berlin : Julius Springer.
[20] Franklin, William M. (1963) “Zonal Boundaries and Access to Berlin,” World Politics, 16(1), 1–31.
[21] Friedensburg, F. (1967) “The Geographical Elements in the Berlin Situation,” The Geographical
Journal, 133(2), 137-144.
[22] Fujita, M. and Ogawa, H. (1982) “Multiple Equilibria and Structural Transformation of Nonmonocentric Urban Configurations”, Regional Science and Urban Economics, 12, 161-96.
[23] Fujita, M., Krugman, P. and Venables, A. J. (1999) The Spatial Economy: Cities, Regions and International Trade, MIT Press.
[24] Gibbons, Stephen and Stephen Machin (2005) “Valuing Rail Access Using Transport Innovations,” Journal of Urban Economics, 57(1), 148-169.
[25] Glaeser, Edward L. and D. C. Mare (2001) “Cities and Skills,” Journal of Labor Economics, 19(2),
316-342.
[26] Glaeser, Edward L., H. D. Kallal, J. A. Scheinkman and A. Shleifer (1992) “Growth in Cities,”
Journal of Political Economy, 100, 1126-1152.
[27] Gyourko, J. and Tracy, J. (1991) “The Structure of Local Public Finance and the Quality of Life,”
Journal of Political Economy, 99, 774-805.
24
[28] Head, Keith and Thierry Mayer (2004) “The Empirics of Agglomeration and Trade,” in Handbook
of Regional and Urban Economics, (Vol. 4), J. Vernon Henderson and Jacques-François Thisse (eds.),
2609–69, Amsterdam: North Holland.
[29] Henderson, J. V., A. Kuncoro and M. Turner (1995) “Industrial Development in Cities,” Journal
of Political Economy, 103, 1067-1085.
[30] Kain, J and Quigley, J. (1970) “Measuring the Value of Housing Quality,” Journal of the American
Statistical Association, 65, 330.
[31] Kalweit, F. (1929) Die Baustellenwerte in Berlin 1928, Teil 1, Berlin: Emro.
[32] Kalweit, F. (1937) Die Baustellenwerte in Berlin: Strassen-Verzeichnis, Neubearbeitung Abgeschlossen
am 30. Juni 1936, Berlin: Carl Heymanns.
[33] Kettenacker, Lothar (1989) Krieg zur Friedensicherung: Die Deutschlandplanung der britischen
Regierung während des Zweiten Weltkrieges, Göttingen: Vandenhoeck & Ruprecht.
[34] Lucas, R. E. (2000) “Externalities and Cities”, Review of Economic Dynamics, 4, 245-74.
[35] Lucas, R. E. and Rossi-Hansberg, E. (2002) “On the Internal Structure of Cities”, Econometrica,
70(4), 1445-76.
[36] Marshall, Alfred (1920) Principles of Economics, London: Macmillan.
[37] McDonald, J. F. and C. I. Osuji (1995) “The Effect of Anticipated Transportation Improvement
on Residential Land Values,” Regional Science and Urban Economics, 25, 261-278.
[38] McMillen, D. (1996) “One Hundred and Fifty Years of Land Values in Chicago: A Nonparametric Approach,” Journal of Urban Economics, 100-24.
[39] Merritt, Richard L. (1973) “Infrastructural Changes in Berlin,” Annals of the Association of American Geographers, 63(1), 58-70.
[40] Mills, E. (1967) “An Aggregative Model of Resource Allocation in a Metropolitan Centre,” American Economic Review, 57, 197-210.
[41] Muth, R. (1969) Cities and Housing, Chicago: University of Chicago Press.
[42] Overman, Henry G., Stephen J. Redding, and Anthony J. Venables (2003) “The Economic Geography of Trade, Production and Income: A Survey of Empirics,” in Handbook of International
Trade, (Vol. 1), James Harrigan and E. Kwan Choi, 353–87, Oxford: Blackwell.
25
[43] Rauch, James E. (1993) “Productivity Gains from Geographic Concentration of Human Capital,”
Journal of Urban Economics, 34, 380-400.
[44] Redding, Stephen J. (2009) “The Empirics of New Economic Geography,” Journal of Regional
Science, 50th Anniversary Special Issue.
[45] Redding, Stephen J. and Daniel M. Sturm (2008) “The Costs of Remoteness: Evidence from
German Division and Reunification,” American Economic Review, 98(5), 1766-1797.
[46] Redding, Stephen J., Daniel M. Sturm and Nikolaus Wolf (2009) “History and Industrial Location: Evidence from German Airports,” London School of Economics, mimeograph.
[47] Roback, J (1982) “Wages, Rents and the Quality of Life,” Journal of Political Economy, 90, 1257-78.
[48] Robinson, G. W. S. (1953) “West Berlin: The Geography of an Enclave,” Geographical Review,
43(4), 540-557.
[49] Rosenthal, Stuart S. and William C. Strange (2004) “Evidence on the Nature and Sources of
Agglomeration Economics,” Chapter 49 in (eds.) J. V. Henderson and J.F. Thisse, Handbook of
Regional and Urban Economics, Volume 4, Amsterdam: Elsevier North Holland.
[50] Rosenthal, Stuart S. and William C. Strange (2008) “The Attenuation of Human Capital
Spillovers,” Journal of Urban Economics, 64(2), 373 89.
[51] Runge, E. (1950) Bodenwertkarte Gross-Berlin, Berlin: Hansa Druck.
[52] Sen, Ashish and Tony E. Smith (1995) Gravity Models of Spatial Interaction Behavior, Heidelberg:
Springer-Verlag.
[53] Sharp, Tony (1975) The Wartime Alliance and the Zonal Division of Germany, Oxford: Clarendon
Press.
[54] Sveikauskas, L. A. (1975) “The Productivity of Cities,” Quarterly Journal of Economics, 89, 393-413.
[55] Taylor, Frederick (2006) The Berlin Wall, London: Bloomsbury.
26
Figure 2: West Berlin Rents 1936 Base Sample
35
30
Normalized Rent
25
Kudamm
20
15
10
52.6
5
52.55
0
13.2
52.5
13.25
13.3
52.45
13.35
13.4
Longitude
13.45
52.4
Latitude
Figure 3: West Berlin Rents 1986 Base Sample
35
30
Normalized Rent
25
20
Kudamm
15
10
52.6
5
52.55
0
13.2
52.5
13.25
13.3
52.45
13.35
13.4
Longitude
13.45
52.4
Latitude
-4
Normalized Rent Growth
-2
0
2
4
Figure 5: West Berlin Rent Growth 1936-1986
0
5
10
Distance to pre-war CBD
Actual
Fitted
Note: Fitted values based on locally-weighted linear least squares regression
15
-2
Normalized Rent Growth
-1
0
1
2
3
Figure 6: West Berlin Rent Growth 1986-2006
0
5
10
Distance to pre-war CBD
Actual
Fitted
Note: Fitted values based on locally-weighted linear least squares regression
15
Table 1: Long Differenced Rents for West Berlin Blocks following Division and Reunification
(1)
(2)
(3)
(4)
(5)
(6)
 ln Rent
 ln Rent
 ln Rent
 ln Rent
 ln Rent
 ln Rent
1936‐86
1936‐86
1936‐86
1936‐86
1986‐2006 1986‐2006
PWCBD1
‐3.159*** ‐2.767*** ‐1.881*** ‐2.072*** 1.684***
1.973***
(0.353)
(0.404)
(0.320)
(0.337)
(0.422)
(0.361)
PWCBD2
‐2.705*** ‐2.275*** ‐1.272*** ‐1.472*** 1.355**
1.506***
(0.362)
(0.398)
(0.332)
(0.349)
(0.562)
(0.430)
PWCBD3
‐1.686*** ‐1.319*** ‐0.567*** ‐0.802*** 0.274*
0.456***
(0.129)
(0.203)
(0.171)
(0.219)
(0.139)
(0.167)
PWCBD4
‐1.314*** ‐1.107*** ‐0.426*** ‐0.591*** 0.233**
0.285***
(0.196)
(0.204)
(0.159)
(0.184)
(0.115)
(0.096)
PWCBD5
‐1.122*** ‐1.007*** ‐0.441*** ‐0.572*** 0.183*
0.233***
(0.165)
(0.165)
(0.135)
(0.144)
(0.106)
(0.080)
PWCBD6
‐0.734*** ‐0.658*** ‐0.256*** ‐0.345*** 0.109
0.119*
(0.153)
(0.134)
(0.095)
(0.091)
(0.091)
(0.063)
INNER1
0.456*
(0.233)
INNER2
0.514**
(0.226)
INNER3
0.345*
(0.183)
INNER4‐INNER6
yes
OUTER1
0.033
(0.166)
OUTER2
0.164
(0.159)
OUTER3
0.143
(0.138)
OUTER4‐OUTER6
yes
District Fixed Effects
yes
yes
yes
yes
Block controls
yes
yes
Observations
5162
5162
5162
5162
5162
5162
R‐squared
0.49
0.71
0.82
0.83
0.19
0.50
(7)
 ln Rent
1986‐2006
1.348***
(0.308)
0.981***
(0.333)
0.094
(0.127)
‐0.029
(0.076)
‐0.069
(0.074)
‐0.131**
(0.056)
yes
yes
5162
0.62
(8)
 ln Rent
1986‐2006
1.285***
(0.294)
0.975***
(0.310)
0.167
(0.130)
0.035
(0.087)
‐0.024
(0.081)
‐0.091*
(0.050)
‐0.370***
(0.104)
‐0.436***
(0.108)
‐0.390***
(0.097)
yes
‐0.299***
(0.111)
‐0.203**
(0.084)
‐0.154**
(0.076)
yes
yes
yes
5162
0.66
Notes: Regression sample includes blocks in West Berlin for which rent data are available for 1936, 1986 and 2006. PWCBD1‐PWCBD6 are distance grid cells in 1 km intervals from the pre‐war CBD. WALL1‐WALL6 are distance grid cells in 1 km intervals from the Berlin Wall. OUTER1‐OUTER6 are distance grid cells in 1 km intervals from the outer boundary of West Berlin. District fixed effects are for the 11 districts ("Bezirke") of West Berlin (based on pre‐war and division boundaries). Block controls are: log block area, log average number of floors per building in the block, log distance to the nearest U‐Bahn station, log distance to the nearest S‐Bahn station, log distance to the nearest school, log distance to the nearest canal, lake or river, log distance to the nearest park, and the proportion of the block's area destroyed during the Second World War. Distance to the nearest U‐Bahn and S‐Bahn station is measured in 1936 for the 1936‐
1986 regression and in 1986 for the 1986‐2006 regression. Robust standard errors in parentheses are adjusted for clustering on statistical areas ("Gebiete"), 69 Clusters. * significant at 10%; ** significant at 5%; *** significant at 1%.