Content uploaded by Jorge Alberto Montejano
Author content
All content in this area was uploaded by Jorge Alberto Montejano on Feb 15, 2022
Content may be subject to copyright.
February 2022
The findings and conclusions of this Working Paper reflect the views of the author(s) and have not been
subject to a detailed review by the staff of the Lincoln Institute of Land Policy. Contact the Lincoln Institute
with questions or requests for permission to reprint this paper. help@lincolninst.edu
© 2021 Lincoln Institute of Land Policy
Regulating Urban Sustainability: Land
Regulations, Urban Spatial Structure,
Transportation Infrastructure, and Greenhouse
Gas Emissions
Working Paper WP22PM1
Paavo Monkkonen
UCLA
Erick Guerra
University of Pennsylvania
Jorge Montejano Escamilla
CentroGeo
Camilo Caudillo Cos
CentroGeo
Abstract
Cities feature prominently in debates over how to deal with the existential threat of climate
change, alternating between the role of potential savior (e.g. Copenhagen or Amsterdam) and
prime culprit (e.g. Phoenix or Brisbane). The foundations of the debate on urban sustainability,
however, are too often based on research from data-rich regions of the world, such as the United
States and Europe. That bias may soon change. Globally available, harmonized datasets are
enabling new, robust, comparative analysis of cities. This paper is an effort to begin answering a
basic question—how do urban land policies shape sprawl, urban transportation, and greenhouse
gas emissions? To assess the relationships between these elements of urban sustainability, we
combine high-quality, globally available GIS data on urban footprints, population density,
transportation patterns, and carbon emissions with surveys of regulatory processes from the
World Bank’s Doing Business project for over 400 cities in nearly 40 countries. The sample
contains predominantly middle-income countries, but it is the beginning of research that will
eventually cover the entire world.
One set of findings confirms the importance of urban form for carbon dioxide (CO2) emissions.
Dense, compact cities with built-up downtowns and shorter roadway segments have lower CO2
emissions per capita. Our findings on the relationship between land regulations and emissions
tell a more complicated story. In the global sample of cities, we find that density is associated
with more time-consuming regulations, and places with ‘more’ regulations have lower
emissions. The fact that these are correlations is fundamental and underscores that regulations
are created when density is present. The finding is especially relevant when contrasted against
evidence in rich countries that emphasizes how regulations preventing dense urbanization
promote sprawl. We also examine the relationship between regulations and informality, finding
that cities with ‘more’ regulation do indeed have more slums.
The results highlight the need for a nuanced understanding of the relationships among
regulations, urban form, and sustainability. Regulations are necessary for urban density and
sustainability to be productive and functional, though in some contexts they are
counterproductive and promote sprawl. We propose a global model to frame this idea and argue
that finding the right regulatory balance is important, as is understanding where a given city is on
this spectrum. As a global research community, we need more and better data to unpack this
nuance. This paper’s limitations illustrate that point and establish a framework for assessing
relationships among regulations, urban form, and emissions, a surprisingly underemphasized
chain in urban research.
About the Authors
Paavo Monkkonen, Ph.D., (paavo.monkkonen@ucla.edu) is Associate Professor of Urban
Planning at the UCLA Luskin School of Public Affairs. He studies the way housing policies
shape urban development and segregation in cities around the world. More specifically, he
researches housing finance policy, land use regulations, socioeconomic segregation, land titling
programs, household formation, and property taxation. His work has been published in academic
journals such as the Journal of the American Planning Association, the International Journal of
Urban and Regional Research, the Journal of Urban Economics, Regional Science and Urban
Economics, Urban Studies, and the Journal of Peasant Studies. Dr. Monkkonen completed a
Master of Public Policy at the School of Public Affairs at the University of California, Los
Angeles, and a Ph.D. in City and Regional Planning at the University of California, Berkeley. He
was previously Assistant Professor of Urban Planning at the University of Hong Kong.
Erick Guerra, Ph.D., (erickg@upenn.edu) is Associate Professor in City and Regional Planning
at the University of Pennsylvania, where he teaches courses in transportation planning and
quantitative planning methods. His research focuses on the relationship between land use,
transportation systems, and travel behavior with an emphasis on rapidly motorizing cities, public
health outcomes, and transportation technologies. He has published recent articles on land use
and transportation in Mexico and Indonesia, public transport policy, land use and traffic safety,
and contemporary planning for self-driving vehicles. His 2017 book Beyond Mobility, with
Robert Cervero and Stefan Al, explores global challenges and opportunities to creating safer,
healthier, and more productive cities. Dr. Guerra holds a Ph.D. in City and Regional Planning
from the University of California Berkeley, a Master’s in Urban Planning from Harvard
University, and a Bachelor of Arts in Fine Arts and French from the University of Pennsylvania.
He served as a Peace Corps Volunteer in Gabon from 2002 to 2004.
Jorge Montejano, Ph.D., (jmontejano@centrogeo.edu.mx) is Professor (fulltime) at Centro de
Investigación en Ciencias de Información Geoespacial (CentroGeo-CONACYT) in Mexico City,
where he teaches courses in urban studies. He has published articles on land-use travel behavior,
urban structure, urban growth, social housing, and the impact of technology on the built
environment. He is a member of the National Research System in Mexico (S.N.I.); the Mexican
Academy of Urbanists (AMU), and the Network of Studies in Urban Form (REFU). His work
has been published in journals such as Urban Studies, Urbs, and Economía, Sociedad y
Territorio, among others. He studied Architecture at the Universidad Iberoamericana in Mexico
City and completed a Ph.D. in Urbanism at the Universitat Politécnica de Cataunya, in
Barcelona, Spain.
Camilo Caudillo has a Masters degree in Geomatics from the Centro de Investigación en
Geografía y Geomática “Ing. Jorge L. Tamayo”, A.C. (CentroGeo) and a Masters in Population
Studies from the Latin-American Social Sciences Faculty (FLACSO-Mexico). Currently, he is
Associate Professor and a doctoral student in Geomatics at CentroGeo, where he teaches spatial
analysis and spatial statistics. His research interests are urban geography, residential segregation,
spatial econometrics, detection of spatial and spatiotemporal patterns, and public safety. He has
co-authored articles and book chapters related to geospatial technologies in prestigious
publishers and journals like Springer and Urban Studies.
Table of Contents
1. Introduction: Understanding the Greatest Challenge of the 21st Century…………………1
2. Theory and Relevant Literature: Land Use Regulation, Urban Form, and Sustainable
Cities………………………………………………………………………………………2
3. Global Data on Regulations, Form, and Sustainability………………………………...…6
4. Measuring Regulations, Form, and Sustainability……………………………………….11
5. Analysis and Results……………………………………………………………………..18
6. Conclusion……………………………………………………………………………….23
1
1. Introduction: Understanding the Greatest Challenge of the 21st Century
Climate change is an existential threat to humanity. Given that a majority of humans live in
cities, and human activity is the main driver of climate change, the global understanding of how
cities can be made less damaging to the planet is surprisingly limited. One crucial question is
how regulating urbanization might be able to facilitate sustainability by making it denser rather
than working against sustainability goals by pushing urban growth toward a low-density
fragmented pattern of sprawl.
In this paper, we analyze the three sets of relationships—regulation and form, form and
environmental impacts, and regulations and environmental impacts—in a sample of more than
400 cities in over 40 countries. Our sample is constrained and determined by data on regulations,
which are rarely available in a consistent format across cities. We rely on the World Bank’s
subnational sample of Doing Business surveys, which have limited coverage (and only focus on
regulatory processes), yet also has the great advantage of being a harmonized set of indicators
consistently collected across numerous cities. Data on environmental impacts also constrain the
analysis. We use the best available, from the Global Gridded Model of Carbon Footprints
(GGMCF).
The base data required to calculate measures of urban form have seen the most innovation. The
now easily accessed Global Human Settlement Layer (GHSL) created by the European
Commission (EC) provides detailed population density data for the entire globe on a small
geographic scale, enabling the creation of multiple measures of urban form. We present several
replicable metrics here, with a detailed explanation of their development in the paper’s appendix.
The first core finding confirms the importance of urban form for sustainability. Cities with
higher population densities, shorter roadway segments, built-up downtowns, which are more
geometrically compact (i.e. more circular) have lower CO2 emissions from mobile sources
(measured using the number of gas stations per capita). Doubling a city’s population density is
associated with a 29% to 47% reduction in CO2 per capita across our model specifications.
Doubling the amount of roadway per capita corresponds with a roughly 62% increase in
emissions per capita.
Our findings on the relationship between land regulations and emissions are more complicated.
In part, this result is explained by an important caveat to our data on regulations—at this stage
we only have data on regulatory processes. Recent work on regulation has made an important
distinction between restrictive regulatory prohibitions—rules that restrict the size and density of
buildings and neighborhoods like minimum lot sizes, maximum densities, or height limits—and
restrictive regulatory processes – rules that introduce long delays or uncertainty through
complicated or politically determined approval processes or ones that require developers to pay
high costs for building permits (Monkkonen et al., 2020).
These two groups of regulations impact urban expansion or dense housing production in
different ways. Density prohibitions would most obviously be related to lower densities and
sprawl. The impact of complex and costly regulatory processes on urban form is less clear.
Growing, prosperous cities may tend to increase regulatory bureaucracy, because people prefer a
2
more carefully controlled built environment, and delays in permitting occur simply because there
is more development activity. Thus, a simple correlation across cities may find a positive
relationship between density and process, and it will be more difficult to identify the
counterfactual impact of more restrictive processes. We expect that the impacts of regulatory
density prohibitions will be more easily measured once data are more widely available.
We find that higher densities are associated with more time-consuming regulations. Doubling
population density is associated with a roughly 21% increase in property registration time.
Similar to previous work (Monkkonen and Ronconi, 2013), we find that higher GDP locations
also have more time-consuming regulations—a $1000 increase in GDP per capita is associated
with a roughly 3% increase in time to register property. Again, this finding is expected, because,
at this point, we can only measure regulatory processes, rather than density prohibitions.
We also find that places with ‘more’ regulations have lower emissions. This finding is especially
important, because, combined with the connection to urban form, it challenges a sometimes
oversimplified understanding of the relationships between regulations and urban form: that
regulation only works in one direction to limit density and increase prices; and that more strictly
regulated cities will be more sprawling, produce less housing, and be more expensive. However,
not all land use regulations arise for the same reasons, nor do they have the same effects. Thus,
we posit that our finding about the relationship to emissions reflects the increase in regulatory
bureaucracy associated with more functional urban governments. Of course, it does not suggest
that land use regulations cannot be overly strict or create problems for urban development,
because correlational analysis does not test the counterfactual of identical places with differing
levels of regulatory bureaucracy.
2. Theoretical Framework and Relevant Literature
In order to conceptualize the relationship between land regulation and urbanization, we propose
an inverted U model, as shown in Figure 1. It posits the existence of a hypothetically optimum
degree of regulation—not too much or too little—that facilitates functional and sustainable urban
density. Much of the research to date has focused on cities on the lower right side of this model
(low density, high regulation), especially those in California (Quigley and Raphael 2005,
Jackson 2016, Jackson 2018, Kahn, 2011). The present study focuses on cities on the left side,
near the center of this model. This approach adds nuance to the broad, and at times
oversimplified, understanding of the regulations/form relationship. Cities need regulations to
densify in an efficient manner (more like Hong Kong, less like Dhaka), and richer locations will
tend to have more regulations. Finding the right balance and level is important. As a research
community, we need more and better data to get at this global nuance.
3
Figure 1: Model of Regulatory Burden and Functioning Urban Compactness
Source: Authors
Finally, we examine a commonly hypothesized secondary impact of land regulations: their
relationship to informality and slums. As with other research on this topic, we find that more
time-consuming rules are associated with higher proportions of slum areas in a city. A city in the
top quartile of ‘onerous’ land regulations has roughly 30 percentage points more residents living
in slums than an average city (in terms of regulatory bureaucracy). This finding reinforces the
folly of blanket assertions about regulations being problematic or beneficial.
Ascertaining the ‘right’ regulatory balance is important and context dependent, and analysts
should seek to understand the details of any city before making assumptions. As a global
research community, we need more and better data to get at this nuance.
On account of this study’s limitations, this paper may be seen as a provocative first step in a
research agenda to improve our understanding of how land regulations can shape less
environmentally destructive cities. The results outlined above should motivate further global
work on this topic. Moreover, they inform urban policy discussions worldwide, especially in
countries where urban development is under-regulated. Urban governments interested in
sustainability should use this study as a call to action, and a counter to simple arguments for
deregulation.
This study draws on three distinct bodies of academic research: studies of land use regulation
and how they impact various outcomes; research on urban form—especially the challenges of
measuring it; and the environmental impacts of cities, emphasizing the role of urban form and
transportation emissions.
4
2.1 Land Use Regulations
A common theoretical framework in international development work on urban sustainability
(UN Habitat; Buckley and Kalarickal 2006) is related to the body of economic research on land
use regulations, which has its origins in the United States (Glaeser et al., 2005; Quigley and
Raphael, 2005; Saiz, 2010) and Europe (Cheshire and Sheppard, 2002) and has been tied to
carbon emissions in that context (Magnum, 2014). Application of this work in developing-
country contexts, where the difference between written rules and their application is often quite
large (Monkkonen, 2013; Monkkonen and Ronconi, 2013), continues to grow.
One basic theoretical insight is that urban population growth can occur either horizontally
through the urbanization of rural land, or through higher-density development on already
urbanized land. The demand for housing is shaped by tradeoffs between access to central city
jobs and amenities and cheaper space at the periphery. Developers respond to demand in
building housing and substitute between land and capital (buildings) by building denser in more
central locations. Governments can support or constrain the process of urbanization in multiple
ways, and policies make cities more or less sprawling than they would be otherwise given
economic and demographic conditions.
Land use regulations make housing development and urbanization more difficult, but these
regulations tend to focus on existing urban areas, because existing residents exert political
pressure not to change their neighborhoods and not to add more people to existing service
networks. Thus, regulatory prohibitions against higher density developments or increased
regulatory scrutiny of the processes of development reduce what could be built below the
underlying demand to live in central urban locations. As Figure 2 illustrates, one tendency is for
density or height restrictions to limit development where demand is most intense, which reduces
a city’s overall population growth and pushes where urbanization can occur further from the city
center (Bertaud and Brueckner, 2005). It is this latter push, towards urban sprawl, that makes
cities less environmentally sustainable.
Figure 2: How Density or Height Restrictions Encourage Sprawl
Source: Authors, based on Bertaud and Brueckner (2005)
5
Therefore, this paper builds on the growing global perspective within the vibrant area of research
on land use regulations and urbanization (e.g. Monkkonen, 2013; Collier and Venables, 2017;
Kleemann et al., 2017; Zhou et al., 2017; Han and Go, 2019), and a policy focus on the
increasingly sophisticated globally comparative work on urbanization (Angel et al., 2012). In
terms of policy, the study’s assessment of the relationship between specific types of land,
property, and infrastructure policies, and more sustainable urban forms, is useful for city leaders
interested in sustainability. We began by examining three types of local land use policies:
construction permitting, property registration, and obtaining electricity (as a proxy for urban
infrastructure). However, due to data limitations we focus on property registration.
2.2 Urban Form
Myriad forces shape the ultimate form a city takes, and scholars have wrestled with the relative
importance of different factors for over a century. Recent efforts to assess trends in global urban
expansion (Angel et al., 2005) highlight six groups of factors: the natural environment,
demographics, the economy, the transport system, consumer preferences, and governance. These
groups do not act independently on urban form, but even stylized facts assessing these
relationships using global data are yet unknown.
2.3 Environmental Impacts of Cities
High-profile land policies intended to influence urban form and curb sprawl—such as urban
containment perimeters and greenbelts—have been implemented by relatively few cities, though
they garner the greatest attention. Whereas the strict greenbelts of Seoul and Brasilia are well-
known, most cities shape urban form through a mix of transportation investments, land use rules,
and bureaucratic decisions that make it easier or harder to build certain types of buildings in
different locations. These policies are harder to study, but in a global sense, more important.
Recent studies have highlighted the importance of public finance in sprawl in urban China (Liu
et al., 2018), the relevance of industrial mix and agricultural productivity for urban expansion in
Europe (Oueslati et al., 2015), the urban consequences of a fragmented governance structure
(Ahrend et al., 2017), and the interaction between housing policy and private vehicle travel in
Mexico (Guerra, 2015).
Existing literature on urban form and sustainability is consistent in its findings about the role of
density and the shape of cities in the carbon emissions of their residents. Much of this research is
from the United States (as pointed out in a literature review by Kahn and Walsh, in 2015) where
researchers find that doubling urban density correlates with a nearly 50% reduction in emissions
from travel (Lee and Lee, 2014) and a 40% increase in overall CO2 efficiency (Gudipudi et al.,
2016). Additionally, scholars from China have found that higher densities and less complex
shapes of cities, i.e. more geometrically compact ones, have lower CO2 emissions (Fang et al.,
2015).
6
3. Global Data on Regulations, Form, and Sustainability
3.1 The Sample
Data on land use regulation in the Doing Business database determined the sample of cities for
this paper, from the more than 400 cities for which data on regulations were available in
subnational samples. The table in Appendix I presents a list of countries and the number of cities
from each included in this study. The sample is balanced across world regions, with nearly 90
cities from Africa and the Americas, 119 in Europe, and 156 in Asia. It is a diverse set of
contexts, though primarily middle- and low-income, with a median GDP per capita of $6,500 in
2018 for this sample. For most countries we have data on two to 12 cities; though for China,
Russia, Mexico, Colombia, and Nigeria, we have data on over 30 cities. Figure 3 is a map of the
cities in our study.
Figure 3: Sample of 433 Cities in this Analysis
Source: Authors
One initial challenge was determining the boundaries of these urban areas, as administrative
jurisdictions often do not map onto urbanized areas. To do this, we used the idea of Functional
Urban Areas (FUAs), defined by the Organization for Economic Co-operation and Development
(OECD) as “sets of contiguous local (administrative) units composed of a ‘city’ and its
surrounding, less densely populated local units that are part of the city’s commuting zone”
(Schiavina et al., 2019, p. 2).
The FUA concept is especially useful because it harmonizes a definition of metropolitan area
globally. It characterizes the functional dependence and spatial contiguity of interconnected and
adjacent urban centers. Most countries use distinct methods for denoting metropolitan areas, so
the FUA approach, based on previous OECD work in defining a common European Functional
7
Urban Area (Dijkstra et al., 2019), is important. The detailed methodology is reported in
(Schiavina et al., 2019) and (Moreno-Monroy et al., 2020).
Figure 4a: Missing FUA (in red) in Central Italy
Source: Authors, with data from Florczyk et al., 2019 and Schiavina et al., 2019.
However, nearly 20% of the sample of cities from the Doing Business database were missing
FUAs. Figure 4a shows central Italy, where L’Aquila FUA (in red) is a significant urbanized
area missing from the original worldwide FUA database (9,031 FUAs in green). L’Aquila has
nearly 70,000 inhabitants and is considered a FUA, according to (Dijkstra et al., 2019). Thus, we
constructed FUAs by either adding them from the OECD database or building a proxy of a
missing FUA by joining municipalities that contain contiguous urban areas.
We completed the FUA calculation approach for missing FUAs in the same way as the OECD
FUA database outlines,
1
joining administrative boundaries of the main city with adjacent
municipalities when they were covered by the main city’s urban footprint. Administrative
boundaries were obtained from GADM maps.
2
Figure 4b provides an example of FUA
boundaries in the Mediterranean.
1
Available at: www.oecd.org/cfe/regional-policy/functionalurbanareasbycountry.htm
2
Available at: https://gadm.org/index.html
8
Figure 4b: FUAs in the Mediterranean Region (green) City Centroids (red)
Source: Authors, with data from Florczyk et al., 2019 and Schiavina et al., 2019.
3.2 Data on Land Use Regulation
Assessing the impacts of land use regulation is complex in part because of the different types of
rules the term encompasses, which range from regulatory prohibitions, like height and density
restrictions, to regulatory procedures that make the process for approving a new building more or
less complex and costly. Additionally, for many regulations, there is a strong correlation between
higher levels of regulations, urban growth, and prosperity.
The importance of land use regulations has been recognized for decades, and early efforts to
measure them to understand their impacts better were a core component of the ambitious World
Bank / UN Habitat 1989 Housing Indicators program for data collection. That program sought to
monitor housing markets, evaluate their performance, and set out an agenda for a regulatory
audit to draw attention to how rules distort the market. Their data collection had several goals,
including regulatory reform. The book Housing Policy Matters (Angel, 2000) reflects the
culmination of this data collection effort. It analyzes mostly national data to assess the impact of
the various dimensions of indicators on housing outcomes.
This study relies on the World Bank’s Doing Business survey project, which began in the 1990s
as an effort to systematically record the regulatory “costs of doing business” in every country in
the world, with an initial focus on starting a business and enforcing contracts. It soon expanded
to focus on other areas, including obtaining a construction permit, registering a property transfer,
and getting electricity. For each of the topics, a detailed description of each step required to carry
9
it out is recorded, along with the monetary cost, time, and relevant actors. In places where
obtaining a construction permit, for example, takes longer, requires more bureaucratic steps, and
costs more, the regulatory environment is considered to be more burdensome. These quantitative
indicators are generated through multiple surveys of expert practitioners and government
officials.
The Doing Business project takes several steps to make indicators globally comparable. In
conducting surveys, they use the same hypothetical project for construction permits and access to
electricity, and the same type of property for property transfers. Additionally, they present
information on a similar type of company requesting these bureaucratic procedures so as to
control for any influence the requestor may have on the process. Finally, they report costs in
relative terms—normalized by an estimate of the building’s value in that city—to make
comparison possible.
Doing Business survey data are available for every ‘economy’ in the world, a total of 190. The
urban indicators are obtained for the largest city in the country. This sample presents a challenge
for analysis, separating national level factors from the role of local regulations and government.
Fortunately, Doing Business surveys are also available for 510 local economies in 75 countries.
These subnational samples enable us to model the local effects of urban policies while
controlling for national context.
Recent work on measuring land use regulations by Lewis and Marantz (2020) and O’Neill and
colleagues (2019) challenges the utility of the survey approach, because of inaccuracies in what
those surveyed know about this complex topic. Although the Doing Business approach is more
reliable than many of the approaches taken by United States scholars (Doing Business surveys
several actors about a smaller number of processes), this recent work highlights the challenge of
measuring regulatory prohibitions. Doing Business surveys focus exclusively on processes.
Density and height restrictions are expected to have more significant impacts on urbanization,
but they are not covered.
3.3 Urban Form Data
We calculate a set of Urban Form Metrics (UFM) using the newly harmonized Global Human
Settlement Layer (GHSL) as the primary input. The metrics are not new, but we have brought
them together from different sources. Moreover, they capture slightly different aspects of the
shape of cities, which we consider important to understanding how these different dimensions of
urban form relate to carbon emissions. It is unlikely that a specific functional and sustainable
density or urban form can be identified that would apply globally, but we hope to assess that
possibility.
The GHSL, funded by the European Commission (EC) and developed by the Joint Research
Centre (JRC), is a geospatial database that incorporates a variety of satellite imagery, ranging
from the 1975 60m resolution Landsat satellite to the 2014 38m resolution Landsat satellite. The
GHSL has three main products. The first, the Global Human Settlement Built Grid (GHS-
BUILT), focuses on the built and non-built environment and is available for 1975, 1990, 2000
10
and 2014. This layer’s resolution is 30 x 30 m, but the 250m resolution level has a value between
0 and 100 representing the share of urbanized land.
The second product is the population layer for the same time periods (GPWv4 or Global
Population of the World Project, or GHS-POP). Each cell has an imputed population, calculated
using local census data for the corresponding administrative area. Population data are collected
for administrative areas at different levels of resolution in each country (Freire et al., 2016). The
final layers are 250-meter population distribution grids for 1975, 1990, 2000, and 2015, where
the “distribution of resident-based population in each period […]is expressed as the estimated
number of people per cell” (Freire et al., 2016, p. 5).
Despite the novelty of the GHSL, its usability and exhaustive scientific documentation, it does
present some challenges. First, the GHS-BUILT layer is not as accurate as we would like it to be
as a representation of the built environment. The classification method relies on a symbolic
machine learning method, which is a supervised classification technique used for large multiple-
scene satellite data processing scenarios. The estimated degree of built-up within each cell of the
raster is obtained by implementing a fuzzy model. This model estimates a percentage of the cell
covered by a building. The 2014 GHSL has been assessed against the European Land Use and
Land Cover Survey (LUCAS) database, to a 0.96 accuracy level and a Kappa level of 0.32,
which is considered ‘fair’. It has also been tested against a more detailed cartographic database,
reaching nearly 90% total accuracy.
A recent version of the GHSL (used in this research) has been improved by refining spatial
coverage, spatial resolution (from 38 to 30 m), and by fixing the learning algorithm. This means
more built-up areas are detected, and there are fewer false positives (Florczyk et al., 2019, p. 8).
Nonetheless, there is still some overestimation and underestimation of built-up areas. Figure 5
presents an example. Small built-up areas and built-up areas in a desert or sandy context are very
hard to detect properly and thus accumulate errors. The same happens with sprawling cities in a
green context.
Third, the GHS-POP layer is not perfect in imputing population per cell. As noted by (Florczyk
et al., 2019, p. 13), the population was disaggregated from census or administrative units to grids
cells, based on the distribution and density of built-up areas mapped in the GHSL. Since the
latter may under or overestimate built-up areas, inconsistencies in population estimates also
exist.
There are no established city centers in the GHSL, and the location of the city center can shape
the UFM value. Centroids are crucial for computing density gradient metrics (both population
density gradients and built-up density gradients). In order to identify centroid coordinates, we
used city and Bing Maps reverse geocoding to obtain point shapefiles
(https://www.gpsvisualizer.com/geocoder/). We determined that Bing Maps was more accurate
than Google Maps or Map Quest, based on a few test cases. Nonetheless, some small errors were
found and adjusted based on our best estimation.
11
Figure 5: Example of Under and Overestimation Errors
Source: Own with data from Florczyk et al., 2019 and Schiavina et al., 2019.
3.4 Data on Emissions and Slums
To measure environmental conditions we used data from the Global Gridded Model of Carbon
Footprints (GGMCF)
3
. This model provides a globally consistent, spatially resolved (250m)
estimate of carbon footprints (also called Scope 3 emissions) in per capita and absolute terms for
189 countries. It also incorporates existing subnational models for the US, China, Japan, EU, and
UK. The model is described in greater detail in the open-access publication Carbon Footprints of
13,000 cities.
The data on slums are from the World Bank’s database. Unfortunately, it is national level data,
which limits precision. Nonetheless, it provides some measure of informality.
4. Measuring Regulations, Form, and Sustainability
4.1 Measuring Land Use Regulations
The Doing Business data includes 18 measures of the ease of getting construction permits,
registering properties, and starting a business. The property registration data are the most
complete and only required dropping one of our 433 initial urban observations. Each of the other
measures is missing data for at least another hundred urban areas.
3
Available here: http://citycarbonfootprints.info
12
On average, in our sample, it takes six steps to register property (ranging from 1 to 15) and 32
days (ranging from 1 to 250), and costs 4.3 percent of a project’s total budget (ranging from 0 to
25%). We then used two measures in our analysis: the amount of time that it takes to register a
property (Table 3) and a combination of all three measures using principal components (Table
4), explained below. Time is the most clearly uniform measure, since wealthier countries
generally charge more for most services (even as a percentage of project cost), and the number of
steps may include relatively easy or relatively complex steps. The principal component combines
the correlates between the different indicators used by the Doing Business reports.
Table 1: Principal Components of the Difficulty of Registering Properties
Property registration
PC1
PC2
PC3
Number of steps
0.61
-0.06
-0.79
Amount of time
0.57
-0.67
0.48
Total cost
0.56
0.74
0.37
Proportion of variance
0.61
0.22
0.17
Source: Authors
Table 1 summarizes how the three different aspects of property registration processes correlate
with one another through a principal-components analysis (a technique used to combine
variables – like an index). Each principal component (PC1, PC2, PC3) expresses different areas
of shared variance of the data in the three variables. The first (PC1) explains over 60% of the
total data variance and draws equally from the three measures. Since it is also strongly positively
associated with the three measures of property registration, we consider it the best combined
measure of the difficulty of registering a property.
4.2 Measuring Urban Form
Urban form encompasses several dimensions. Urban sprawl, presumably the most
environmentally destructive urban form, is characterized by low-density horizontal urban
expansion, but also fragmentation and polycentricity. In this study, we build on a prior set of
urban spatial structure metrics—population density, the density gradient, urban compactness,
concentration, and fragmentation—described in Montejano et al. (2017) and adapt them to the
raster data format of the global GIS data. We also incorporate recent work in this area (e.g.
Biderman, et al., 2018) and as part of the research have created geospatial tools, including a
raster USSM plugin for Qgis, suitable for long-term release. This free and open source tool will
aid other researchers in calculating similar measures of urban spatial structure.
The measures of urban form in this project can be divided into three main categories: one that
represents the geometric compactness of the urbanized area, one that represents the compactness
of human population settlement, and a third that measures sprawl using street networks. For the
first we calculate Marshall’s Geometric Compactness (Marshall et al., 2019) and compare it to
Amindarbari and Sevtsuk’s Discontiguity metric (Amindarbari & Sevtsuk, 2015) in order to
validate them. For the second, we follow the density gradient approach as operationalized by
Bertaud and Malpezzi (1999). The third approach is inspired by work by Barrington-Leigh and
13
Millard-Ball (2019), though we ended up using simpler measures than the index they developed.
We describe the three approaches below and have included more details in the appendix.
Geometric Compactness
Marshall et al. (2019) recently developed a measure of urban compactness that relies solely on
geometric elements. They start with the basic notion that a circle is the most compact theoretical
shape (Marshall et al., 2019) and build on Angel and others’ research on the compactness
properties of circles (Angel et al., 2012) to create a new measure.
Figure 6: Compactness calculations for Cali and Barcelona
Cali Colombia (4.3) Barcelona, Spain (0.2)
Source: Authors, with data from Florczyk et al., 2019 and Schiavina et al., 2019
Figure 6 presents two examples of how this metric is computed for very different cities. Cali,
Colombia, has a compactness of 4.3, because it has very few ‘urban fragments’ disconnected
from the central core, which fills a circle much more than Barcelona, Spain, which has a
compactness of 0.2.
In the Marshall et al. (2019) compactness metric, a perfectly circular built-up area has a
compactness value of 100. As one can see in the example above, Cali is more geometrically
compact than Barcelona, because the metropolitan area of Barcelona, defined by its FUA, is
more scattered compared to the Cali metropolitan area. Cali has only 33 contiguous urban
patches, whereas Barcelona has 485.
14
Compactness weighted by area
The overall compactness metric for each FUA does not consider the weight of each urban
patch’s size. This situation implies that one can have individual urban patches, no matter their
size, with compactness values close to 1, because either (a) in the case of smaller patches of just
one pixel, the square pixel almost fills a circular area (see Figure 7); or (b) in the case of bigger
urban built-up areas, the more it resembles a square (or a circle), the higher its compactness
value.
Figure 7: The Importance of Patch Size
Notes: On the left (a), a small, square, urban patch with an individual compactness close to 1; on the right (b), a
bigger urban patch with a smaller compactness value.
Source: Authors, with data from Florczyk et al., 2019 and Schiavina et al., 2019
By weighting patch size, we make their importance proportional. The compactness values of
each patch changes, as does the FUA’s overall compactness. For example, Cali’s compactness of
4.3 becomes a weighted compactness of 12.0 (an increase of 180%), whereas Barcelona’s
compactness of 0.2 becomes a weighted compactness of 0.5 (an increase of only 105%). Patches
in Cali on average were larger.
Discontiguity
Amindarbari and Sevtsuk (2015) developed a suite of urban metrics within their Metropolitan
Form Analysis (MFA) toolbox. Among them, discontiguity measures the fragmentation level of
a certain set of patches. It is essentially an inverse to geometric compactness. Figure 8 provides
an illustration for two cities—Mandalyong in the Philippines, and Zrenjanin in Serbia. We
computed this metric and correlated it with the Marshall et al. measure. Amindarbari and Sevtsuk
highlight that their discontiguity metric considers the rank order and relative size of
discontiguous urban clusters in order to ‘weight’ them by area, as we do above. The MFA
15
discontiguity tool is useful in that it computes the ratio between every polygon’s area and the
area of smaller polygons for all polygon features.
Figure 8: Mandaluyong is much less discontiguous than Zrenjanin
Source: Authors, with data from Florczyk et al., 2019 and Schiavina et al., 2019
Figure 9 presents a scatterplot comparing our calculations of urban compactness and
discontiguity. The plot clearly shows a strong inverse relationship between the two metrics.
Figure 9: Comparison between Compactness and Discontiguity
Source: Authors
Street-Network Measures of Urban Form
First, the street-network disconnectedness index (SNDi) is a scalar measure that is expressed as
the degree of connectivity among a specific street network. Barrington-Leigh and Millard-Ball
(2019) developed this metric as a proxy measure of sprawl. As Barrington-Leigh and Millard-
Ball describe it, sprawl can be conceptualized as “a low connectivity in the street network […]
sprawl is characterized by a low nodal degree of intersections” (2015: 15).
R² = 0.9206
-2
-1
0
1
2
3
4
5
6
7
-9 -8 -7 -6 -5 -4 -3 -2 -1 0
LN+1 OF AMINDARBARI AND SEVTSUK
DISCONTIGUITY
LN OF MARSHALL ET AL. COMPACTNESS
Compactness vs Discontiguity (n=433 worwide cties)
Mandaluyong, Philippines. Discontiguity: 0.0
Zrenjanin, Serbia. Discontiguity: 1.136
16
Figure 10: Three Prototypical Street-network Layouts
Notes: (A) The grid paradigm has the highest connectivity and an SNDi of -1.01. (B) The medieval paradigm is
relatively highly connected (SNDi of 0.66) but far from a grid. (C) The culs-de-sac paradigm is the most sprawling,
with SNDi of 7.9
Source: Barrington-Leigh & Millard-Ball, 2019, p. 5.
In their research they find a strong and positive correlation between disconnected urban street
networks and more vehicle travel, energy use, and CO2 emissions. Figure 10 shows the three
prototypical street-network layouts. Barrington-Leigh and Millard-Ball generously provided this
information as another vector of the multifactorial sprawl phenomena for our sample of 433
cities.
Obtaining road statistics and other metrics using OSMnx
In addition to the street-network index, we use three other road statistics and metrics as a proxy
for the degree of urbanization. These metrics—the number of intersections and street length per
capita—proxy how sprawling an urban region is. The more intersections in a street network, the
more compact the city (Barrington-Leigh and Millard-Ball, 2019). We calculate the number of
intersections and the number of intersections per unit area (intersection density) to have a
straightforward measure of road infrastructure development. The disconnectedness of the road
network and a lower node density point to a more sprawled territory.
We use the OSMnx (Boeing, 2017) Python package, which enables spatial geometries to be
downloaded directly from Open StreetMap (OSM), then calculate three street-network measures:
1. Number of intersections in the graph (nodes with >1 street emanating from them)
2. Sum of all edge lengths in the graph, in meters
3. Sum of all edges in the undirected representation of the graph
The original FUA files are projected in World Mollweide, and OSM data are in geographic
coordinates, thus the polygons had to be reprojected to WGS84 (EPSG 4326). OMSnx enables
street networks contained inside polygons to be downloaded.
17
The polygon geometry is fixed, if necessary, to avoid self-intersections and topological errors.
By default, during simplification, OSMnx removes nodes outside of polygon, as well as isolated
nodes. It only retains the graph’s largest weakly connected component. In this case, retaining
other components is desirable, even if they are not connected.
Figure 11: Data on Street Networks for Prague and Wuhan
Location: Prague
Intersections: 38,892
Total edge length (m):
17,669,303
Total street length (m):
9,828,924
Location: Wuhan, China
Intersections: 19,787
Total edge length (m):
14,333,207
Total street length (m):
9,807,955
Source: Authors, with OSM
Figure 11 presents two examples of data processing and results, for the cities of Prague and
Wuhan. The two cities have a very similar total street length, but Prague has roughly twice as
many intersections as Wuhan, meaning it is less sprawling.
4.3 Measuring Urban Environmental Impacts
We rely on estimated greenhouse gas emissions from the Global Gridded Model of Carbon
Footprints (GGMCF). We also collected data on the number of gas stations per capita as a proxy
18
for carbon dioxide emissions from the transport sector. As shown in Figure 12, the two are
strongly correlated and move proportionally together. A one percent increase in GHG is strongly
associated with a one percent increase in the number of gas stations per capita.
Figure 12: Greenhouse gas emissions and gas stations per capita
Source: Authors, with Global Gridded Model of Carbon Footprints
5. Analysis and Results
Our analysis proceeded as follows. First, we evaluated associations between urban form and
CO2 emissions per capita. Then, we assessed the relationship between land use regulations and
urban form and asked whether “more” land use regulations correlate with lower carbon
emissions. Finally, we correlated land use rules and the share of slum areas per capita.
We tested each of the relationships using three models. Model one includes no geographic
controls. Model two introduces a control variable for a metropolitan area’s world region (Central
Asia, West Africa, etc.), in order to reduce the impact of regional differences. Model three goes
further to include a control variable for each country. In the interest of parsimony and to help
avoid overfitting, we excluded measures of urban form that are not statistically significantly
associated at the 90% confidence interval in any of our models.
5.1 Urban form and CO2 emissions per capita
We first tested the direct relationship between measures of urban form and CO2 emissions. If the
ease of property registration contributes to CO2 emissions, we suspect that it does so because of
19
its impact on built form. Table 2 presents the model results of CO2 per capita as a function of
urban form. We have included additional statistically significant predictors. We found that CO2
emissions are inversely correlated with population density, short roadway segments, built-up
downtowns, and geometric compactness. A doubling in population density is associated with a
29% to 47% reduction in CO2 per capita across our specifications. A doubling in the amount of
roadway per capita corresponds to around a 62% increase in emissions per capita. GDP per
capita remains statistically insignificant.
Table 2: OLS and hierarchical linear models predicting CO2 emissions per capita
Dependent variable:
CO2 per capita (nat. log metric tons)
(1)
(2)
(3)
Population density
-0.294***
-0.476***
-0.472***
(natural log of people per hectare)
(0.091)
(0.125)
(0.119)
Roadway per capita
0.623***
0.615***
0.624***
(natural log of meters)
(0.173)
(0.166)
(0.163)
Share of CBD built out
-0.004*
-0.005*
-0.005**
(0 to 100)
(0.002)
(0.002)
(0.002)
Geometric compactness
-6.503***
-6.217***
-6.196***
(0.890)
(0.844)
(0.830)
GDP per capita
-0.044
-0.12
-0.11
1000's USD
(0.088)
(0.153)
(0.134)
Constant
0.799
2.512
2.393
(1.367)
(1.529)
(1.472)
Observations
334
334
334
R2
0.212
Adjusted R2
0.2
Log Likelihood
-549.357
-547.241
Akaike Inf. Crit.
1,114.71
1,110.48
Bayesian Inf. Crit.
1,145.20
1,140.97
Notes: *p<0.1; **p<0.05; ***p<0.01; Standard errors in parentheses; Model 1 OLS, Model 2 random regional
interceptions, Model 3 random national intercepts
20
5.2 Regulations and Urban Form
Next, we assessed the relationship between regulatory bureaucracy and urban form. Table 3
presents the results of three models predicting the amount of time it takes to register a property in
each country. Population density tends to be positively associated with higher property
registration times, though only weakly so. Because the independent and dependent variables are
log-transformed, the parameter estimate has an interpretation as an elasticity. In the model with
regional intercepts, the relationship is statistically significant and suggests that a doubling of
population density is associated with around a 21% increase in property registration time. We
found a much more consistent and significant relationship with national income. Each $1000
increase in GDP per capita is associated with a 0.02 to 0.04 decrease in the natural log of
property registration time. Taking the exponent of parameter estimates, this corresponds to a
reduction of approximately 2% to 4%.
Table 3: OLS and hierarchical linear models predicting total property registration time
Dependent variable:
Property registration time (nat. log)
(1)
(2)
(3)
Population density
(natural log of people per
hectare)
0.08
0.207***
0.057
(0.049)
(0.069)
(0.046)
Share of CBD built out
(0 to 100)
-0.002
-0.001
-0.001
(0.001)
(0.001)
(0.001)
GDP per capita
1000's USD
-0.038***
-0.021*
-0.042**
(0.006)
(0.011)
(0.017)
Constant
3.205***
2.055***
3.310***
(0.417)
(0.580)
(0.414)
Observations
334
334
334
R2
0.225
Adjusted R2
0.218
Log Likelihood
-351.72
-207.64
Akaike Inf. Crit.
715.43
427.28
Bayesian Inf. Crit.
738.30
450.15
Notes: *p<0.1; **p<0.05; ***p<0.01; Standard errors in parentheses; Model 1 OLS, Model 2 random regional
interceptions, Model 3 random national intercepts
The models of ease of property registration produced similar results (shown in Table 4). Since
the principal component is scaled, the easiest interpretation is in terms of standard deviations in
change. For example, a percentage point increase in the share of the CBD that is built is
associated with a 0.002 to 0.004 decrease in the standard deviation of the difficulty of registering
a property.
21
Table 4: OLS and hierarchical linear models predicting difficulty of property registration
Dependent variable:
Difficulty of property registration PC
(1)
(2)
(3)
Population density
0.236***
0.264***
0.071
(natural log of people per hectare)
(0.082)
(0.093)
(0.066)
Share of CBD built out
-0.004*
-0.003
-0.002*
(0 to 100)
(0.002)
(0.002)
(0.001)
GDP per capita
-0.065***
-0.074***
-0.067***
USD 1000's
(0.009)
(0.020)
(0.023)
Constant
-0.748
-1.156
0.107
(0.706)
(0.799)
(0.584)
Observations
334
334
334
R2
0.266
Adjusted R2
0.26
Log Likelihood
-429.361
-323.717
Akaike Inf. Crit.
870.722
659.434
Bayesian Inf. Crit.
893.589
682.301
Notes: *p<0.1; **p<0.05; ***p<0.01; Standard errors in parentheses; Model 1 OLS, Model 2 random regional
interceptions, Model 3 random national intercepts
5.2 Regulations and Carbon Emissions
Finally, we directly assessed the relationships between measures of the difficulty of property
registration and the natural log of CO2 emissions per capita. Table 5 present the estimates. We
found a somewhat consistent relationship across specifications, with an increase in the difficulty
of property registration associated with a decrease in CO2 emissions per capita. Somewhat
surprisingly, we found no statistically significant relationship between CO2 emissions per capita
and national GDP.
Table 5: OLS and hierarchical linear models predicting CO2 emissions per capita
Dependent variable:
CO2 per capita (nat. log metric tons)
(1)
(2)
(3)
Difficulty of property registration PC
-0.140**
-0.148*
-0.098
(0.064)
(0.083)
(0.096)
GDP per capita
-0.010
-0.021
-0.011
USD 1000's
(0.011)
(0.019)
(0.017)
Constant
1.728***
1.925***
1.881***
(0.123)
(0.212)
(0.196)
Observations
334
334
334
22
R2
0.015
Adjusted R2
0.009
Log Likelihood
-586.783
-587.479
Akaike Inf. Crit.
1,183.57
1,184.96
Bayesian Inf. Crit.
1,202.62
1,204.01
Notes: *p<0.1; **p<0.05; ***p<0.01; Standard errors in parentheses; Model 1 OLS, Model 2 random regional
interceptions, Model 3 random national intercepts
5.2 Regulations and Slums
Finally, we examined the relationship between the share of people living in slums and the ease of
property registration. Since slums are measured at the national level, relationships deserve
additional caution when interpreting. Nevertheless, national and urban slum rates roughly
correspond at the cross-sectional level of our analysis. Table 6 presents the results. Each
percentage point increase in the share of the national population living in slums is associated
with a 0.02 to 0.04 increase in the standard deviation of the difficulty of registering a property.
To put this number in perspective, a place with an average difficulty of property registrations has
around 20 to 40 fewer percentage points of residents living in slums than a place at the top
quartile of difficulty.
Table 6: OLS and hierarchical linear models predicting property registration
Dependent variable:
Difficulty of property registration PC
(1)
(2)
(3)
Slums per capita national
0.035***
0.042***
0.024
(0.006)
(0.008)
(0.019)
GDP per capita
-0.056
0.014
-0.055
USD 1000's
(0.040)
(0.075)
(0.118)
Population density
-0.223**
0.07
0.066
(natural log of people per hectare)
(0.110)
(0.096)
(0.089)
0.001
-0.002
-0.002
(0.002)
(0.002)
(0.001)
Constant
1.56
-1.406
-0.458
(1.025)
(0.969)
(1.279)
Observations
230
230
230
R2
0.319
Adjusted R2
0.307
Log Likelihood
-271.63
-255.452
Akaike Inf. Crit.
557.26
524.903
Bayesian Inf. Crit.
581.327
548.97
Notes: *p<0.1; **p<0.05; ***p<0.01; Standard errors in parentheses; Model 1 OLS, Model 2 and Model 3 random
regional interceptions.
23
6. Conclusion
In this paper, we empirically examine a simple but important question: How do urban land
policies shape sprawl, urban transportation, and greenhouse gas emissions? A common narrative
in the United States is that strict regulation of urban development has pushed sprawl and created
environmentally harmful cities. But these relationships have not often been examined globally,
primarily due to data limitations. Recent advances in globally available GIS data and the
relatively underutilized survey data from Doing Business allow us to test these relationships in
over 400 cities in nearly 40 countries. The make-up of the sample is not random, but
predominantly middle-income countries. This composition makes it particularly relevant to
global debates, though the results should be taken within the context of this sample.
Our findings can be considered in two buckets. The first is a strong confirmation of the
importance of urban form for CO2 emissions. We find that dense, compact cities, with built-up
downtowns and shorter roadway segments, have consequentially lower CO2 emissions per
capita. This is one more piece of evidence supporting efforts to make urbanization more
sustainable.
The second group of findings address how our measure of land regulation relates to density and
emissions. We find that more time-consuming regulations are associated with more density,
lower emissions, but also more slums. These findings give nuance to one common narrative of
regulations and urban form—that excessive rules eventually lead to less sustainable growth
patterns. They highlight that regulations are necessary for urban density to be productive and
functional, without contradicting the idea that they can be counterproductive. Our global model
to frame this idea suggests that finding the right regulatory balance is important, as is
understanding where a given city is on the spectrum of land regulation.
As a global research community, we need more and better data to get at this nuance. Beyond the
findings of this analysis, we hope this paper lays the foundation for future work in this area by
outlining an approach to measuring the relationships among regulations, urban form, and
emissions, a surprisingly underemphasized chain in urban research. Especially important will be
regulatory measures in a globally random sample of cities, as well as better measures of
regulatory prohibitions that limit density. The latter issue is a true challenge, but one that will
yield an important distinction in the impacts of prohibitions as compared to processes.
24
References
Albouy, D. & Ehrlich, G. (2018). Housing Productivity and the Social Cost of Land-Use
Restrictions. Journal of Urban Economics, 107, 101-120.
Angel, S., Sheppard, S.C., & Civco, D. L. (2005). The Dynamics of Global Urban Expansion.
Transport and Urban Development Department, The World Bank: Washington D.C.
Ahrend, R., Farchy, E., Kaplanis, I., & Lembcke, A. C. (2017). What makes cities more
productive? Evidence from five OECD countries on the role of urban governance. Journal of
Regional Science, 57(3), 385-410.
Amindarbari, R., & Sevtsuk, A. (2015). Metropolitan Form Analysis Toolbox for ArcGIS v 10.2
and v 10.3.1. City Form Lab.
Angel, S., Parent, J., & Civco, D. L. (2012). The fragmentation of urban landscapes: global
evidence of a key attribute of the spatial structure of cities, 1990–2000. Environment and
Urbanization, 24(1), 249-283.
Barrington-Leigh, C., & Millard-Ball, A. (2015). A century of sprawl in the United States.
Proceedings of the National Academy of Sciences, 112(27), 8244–8249.
Barrington-Leigh, C., & Millard-Ball, A. (2019). A global assessment of street-network sprawl.
PloS One, 14(11).
Bertaud, A., & Malpezzi, S. (1999). The spatial distribution of population in 35 World Cities:
The role of markets, planning and topography [Working Paper]. The Center for urban land and
economic research, The University of Wisconsin.
Bertaud and Brueckner (2005) Analyzing Building Height Restrictions: Predicted Impacts,
Welfare Costs, and a Case Study of Bangalore, India. World Bank Working Paper.
Biderman, C., Hiromoto, M., & Ramos, F. (2018). The Brazilian Housing Program Minha Casa
Minha Vida: Effect on Urban Sprawl. Lincoln Institute of Land Policy Working Paper.
Boeing, G. (2017). OSMnx: New methods for acquiring, constructing, analyzing, and visualizing
complex street networks. Computers, Environment and Urban Systems, 65, 126–139.
Cheshire, P. and S. Sheppard, (2002). The welfare economics of land use planning. Journal of
Urban Economics, 52(2): 242-269.
Collier, P. and A. Venables (2017). Urbanization in developing economies: the assessment.
Oxford Review of Economic Policy 33(3) 355–372.
Corbane, C., Pesaresi, M., Kemper, T., Politis, P., Florczyk, A. J., Syrris, V., Melchiorri, M.,
Sabo, F., & Soille, P. (2019). Automated global delineation of human settlements from 40 years
of Landsat satellite data archives. Big Earth Data, 3(2), 140–169.
25
Dijkstra, L., Poelman, H., & Veneri, P. (2019). The EU-OECD definition of a functional urban
area.
Fang, Chuanglin, Shaojian Wang, and Guangdong Li. (2015). Changing urban forms and carbon
dioxide emissions in China: A case study of 30 provincial capital cities. Applied Energy, 158:
519-531.
Florczyk, A. J., Corbane, C., Ehrlich, D., Freire, S., Kemper, T., Maffenini, L., Melchiorri, M.,
Pesaresi, M., Politis, P., & Schiavina, M. (2019). GHSL Data Package 2019. Luxembourg. EUR,
29788.
Freire, S., Doxsey-Whitfield, E., MacManus, K., Mills, J., & Pesaresi, M. (2016). Development
of new open and free multi-temporal global population grids at 250 m resolution.
Gelman, A, & Hill, J. (2007). Data Analysis Using Regression and Multilevel/Hierarchical
Models. New York: Cambridge University Press.
Glaeser E.L., Gyourko J., Saks R.E. (2005). Why is Manhattan so expensive? Regulation and the
rise in housing prices Journal of Law and Economics, 48 (2), 331-369.
Gudipudi, Ramana, Till Fluschnik, Anselmo García, Cantú Ros, Carsten Walther, and Jürgen P.
Kropp. (2016). City density and CO2 efficiency, Energy Policy 91: 352 – 361.
Guerra, E. (2015). Has Mexico City’s Shift to Commercially Produced Housing Increased Car
Ownership and Car Use? Journal of Transport and Land Use 8 (2): 171–189.
Han, A.T. and M.H. Go. (2019). Explaining the national variation of land use: A cross-national
analysis of greenbelt policy in five countries. Land Use Policy 81 644-656.
Jackson, K. (2016). Do land use regulations stifle residential development? Evidence from
California cities. Journal of Urban Economics, 91, 45-56.
Jackson, K. (2018). Regulation, Land Constraints, and California’s Boom and Bust. Regional
Science and Urban Economics, 68, 130–47.
Kahn, M.E., R. Vaughn, & J. Zasloff. (2010). “The Housing Market Effects of Discrete Land
Use Regulations: Evidence from the California Coastal Boundary Zone.” Journal of Housing
Economics, 19(4), 269–79.
Kahn, Matthew E. (2011). “Do liberal cities limit new housing development? Evidence from
California.” Journal of Urban Economics, 69, 223–228.
Kahn, Matthew and Randall Walsh. (2015). Cities and the Environment. Handbook of Regional
and Urban Economics, 5: 405-465.
26
Kleemann, J. J. Nanalnkoom, M. Theil, S. Shankar, S. Lautenbach, and C. Furst (2017). Peri-
urban land use pattern and its relation to land use planning in Ghana, West Africa. Landscape
and Urban Planning 165 280-294.
Lee, Sungwon and Bumsoo Lee. 2014. The influence of urban form on GHG emissions in the
U.S. household sector. Energy Policy 68: 534-549.
Lewis, P.G., & N.J. Marantz. (2020). “What Planners Know: Using Surveys about Local Land-
Use Regulation to Understand Housing Development.” Journal of the American Planning
Association (forthcoming).
Liu, Yong, Peilei Fan, Wenze Yue, and Yan Song. (2018). Impacts of land finance on urban
sprawl in China: The case of Chongqing, Land Use Policy, 72: 420-432.
Marshall, S., Gong, Y., & Green, N. (2019). Urban compactness: New geometric interpretations
and indicators. In The Mathematics of Urban Morphology (pp. 431–456). Springer.
Monkkonen, Paavo. (2013). Urban Land-Use Regulations and Housing Markets in Developing
Countries: Evidence from Indonesia on the Importance of Enforcement. 2013. Land Use Policy,
34: 255-264.
Monkkonen, P., and Ronconi, L. (2013). Land-Use Regulations, Compliance and Land Markets
in Argentina. Urban Studies, 50(10): 1951-1969.
Monkkonen, P., and Ronconi, L. (2016). Comparative Evidence on Urban Land Use Regulation
Bureaucracy in Developing Countries. In Birch, E., Chattaraj, S., and Wachter, S. (eds.) Slums:
How Informal Real Estate Markets Work. Philadelphia, PA: University of Pennsylvania Press.
Monkkonen, P., Montejano, J., Guerra, E., & Caudillo, C. (2019). Compact cities and
economic productivity in Mexico. Urban Studies, 0042098019869827.
Monkkonen, P. Lens, M. and Manville, M. (2020). Built out cities: How cities in California
restrict housing production with process and prohibition. UC Berkeley Terner Center Working
Paper.
Montejano, J., Monkkonen, P., Guerra, E., & Caudillo, C. (2017). The costs and benefits of
urban expansion: Evidence from Mexico, 1990-2010. Lincoln Institute of Land Policy Working
Paper.
Moreno-Monroy, A. I., Schiavina, M., & Veneri, P. (2020). Metropolitan areas in the world.
Delineation and population trends. Journal of Urban Economics, 103242.
O’Neill, M., G. Gualco-Nelson, & E. Biber. (2019). “Comparing perceptions and practice: why
better land use data is critical to ground truth legal reform.” UC Berkeley: Terner Center for
Housing Innovation Land Use Working Paper Series.
27
Oueslati, W., Alvanides, S., & Garrod, G. (2015). Determinants of urban sprawl in European
cities. Urban Studies, 52(9), 1594-1614.
Pesaresi, M., Ehrlich, D., Ferri, S., Florczyk, A., Freire, S., and Syrris, V. (2016). Operating
procedure for the production of the Global Human Settlement Layer from Landsat data of the
epochs 1975, 1990, 2000, and 2014. Publications Office of the European Union, 1-62.
Pesaresi, M., Ehrlich, D., Ferri, S., Florczyk, A., Freire, S., Halkia, M., Julea, A., Kemper, T.,
Soille, P., & Syrris, V. (2016). Operating procedure for the production of the Global Human
Settlement Layer from Landsat data of the epochs 1975, 1990, 2000, and 2014. Publications
Office of the European Union, 1–62.
Quigley, J. M., & S. Raphael. (2005). “Regulation and the High Cost of Housing in California.”
The American Economic Review, 95(2), 323–28.
Rose, A. N., McKee, J. J., Urban, M. L., & Bright, E. A. (2018). LandScan 2017. Available at:
https://landscan.ornl.gov/
Saiz, A. 2010. The Geographic Determinants of Housing Supply. Quarterly Journal of
Economics, 125:1253-1296.
Schiavina, M., Moreno-Monroy, A., Maffenini, L., & Vener, P. (2019). GHSL-OECD
Functional Urban Areas 2019. Eur. Publications Office of the European Union.
https://doi.org/doi:10.2760/67415
Zhou, Y., X. Huang, Y. Chen, T. Zhong, G. Xu, J. He, Y. Xu, and H. Meng. (2017). The effect
of land use planning (2006–2020) on construction land growth in China. Cities 68 37-47.
28
Appendix I
Country
Year
Number of Cities
Afghanistan
2017
5
Albania
2011
3
Bosnia and Herzegovina
2011
3
Bulgaria
2017
6
China
2008
30
Colombia
2017
32
Costa Rica
2015
1
Croatia
2018
5
Czech Republic
2018
7
Dominican Republic
2015
4
Egypt
2014
15
El Salvador
2015
4
Guatemala
2015
4
Honduras
2015
4
Hungary
2017
7
India
2009
17
Indonesia
2012
20
Italy
2013
13
Kazakhstan
2019
16
Kenya
2016
11
Kosovo
2011
2
Macedonia FYR
2011
3
Mexico
2016
32
Moldova
2011
2
Montenegro
2011
3
Morocco
2008
7
Mozambique
2019
10
Nicaragua
2015
4
Nigeria
2018
36
Pakistan
2010
13
Panama
2015
1
Philippines
2010
25
Poland
2015
18
Portugal
2018
8
Romania
2017
9
Russia
2012
30
Serbia
2011
6
Slovakia
2018
5
South Africa
2018
9
Spain
2015
19
29
Appendix II
This appendix describes how to calculate the various urban form metrics, including relevant
formulas.
Compactness
To calculate Marshall’s Compactness, minimum bounding circles are computed for all the built-
up areas within a specific FUA. It can be done over raster or over vector layers, which is
necessary to compute all the radiuses of all circles that surround built-up areas.
Where is the number of builit-up areas; is the summation of all the the built-up areas within a
specific FUA; is the summation of all the built-up areas’ perimeters within a specific FUA;
and is the summation of all the diameters within a specific FUA, taken from the minimum
bounding circles of the built-up areas. The number four ensures a circle has a maximum
compactness of 1 (Marshall et al., 2019, p. 440):
Compactness Weighted by Area
One limitation of the compactness metric is that it fails to consider the size of patches. In order to
consider patch size (i.e. a bigger contiguous urban area should account for more compactness
than a smaller one, by rewarding the closeness between built-up areas), we propose a variation of
this geometrical compactness metric by weighting it with the size of each patch.
Where:
=
Discontiguity
Where is the number of urbanized clusters, and the area of cluster (p. 7).
