6.147.19
POPULATION GEOGRAPHY
Suzanne Davies Withers
Department of Geography, University of Washington, USA
Keywords:
accessibility, clusters, demographic efficiency, distance decay, geography, geographic distribution,
geographic information systems, geographically weighted regression, geostatistics, life course,
LISA statistics, local statistics, location, location quotients, mapping, migration, migration
expectancy, mobility, population potential, scale, spatial autocorrelation, spatial demography,
spatial statistics, stability
Contents
1. Introduction
2. Population Geography and Contemporary Spatial Demography
2.1 Geographic concepts and spatial thinking
2.2 The making of population geography
2.3 The development of spatial demography
3. Methods of Population Geography
3.1The spatial basics – geographic distribution of population
3.1.1 Descriptive spatial statistics
3.1.2 Index of concentration – The Hoover Index
3.1.3 The Lorenz curve
3.1.4 Population potential – the accessibility index
3.1.5 Location quotients
3.1.6 Segregation measures
3.2 Measures of Migration - population on the move
3.2.1 Migration rates
3.2.2 Demographic efficiency of migration
3.2.3 Migration characteristics and expectancies
3.2.4 Spatial shift-share analysis
3.2.5 Gravity models and economic gravity models
3.3 Methods of spatial analysis
3.3.1 Measures of spatial autocorrelation: Global to Local Moran’s I
3.3.2 Global to Local: Getis-Ord G to Getis Gi
3.3.3 Agent-Based Computational Demography
3.3.4 Geographically weighted regression
4. Themes of Population Geography
4.1 Whither fertility and mortality
4.2 Migration events
4.3 Life course perspective
4.4 Intergenerational studies
4.5 Gendered family migration
4.6 Assimilation processes
4.7 Neighborhood effects and residential mobility
5. Conclusion and Future Challenges
Bibliography
Summary
1
This chapter provides an overview of the field of population geography. Population geography is a
subfield of the discipline of geography and a subfield of the discipline of demography. Population
geography addresses the spatial distribution, characteristics, and spatial variation of the population.
The importance of a spatial perspective for demographic research has received considerable
attention over the past few decades. Population geography addresses demographic issues and
population processes in an explicitly spatial manner, with a focus on the connection between people
and places. Spatial demography refers to the formal methods used to make these links. Geographic
concepts and spatial thinking are described, with particular attention given to the concept of scale.
The chapter reviews the intellectual heritage of population geography and explains the
contemporary correspondence between population geography and spatial demography. Conceptions
of space and place in population geography are quite sophisticated, and recent advancements in
spatial analysis have been enabled by geographic information systems, software development, and
the availability of spatially explicit data (geocoded) sources. A host of methods of population
geography are detailed, giving particular attention to the recent development of local, as opposed to
global, measures of spatial analysis. Methods are reviewed for the study of geographic distribution,
population movement, and spatial analysis.
Traditionally, population geographers addressed the three components of population change –
fertility, mortality, and migration, only. However, contemporary population geography is more
thematic and theoretically sophisticated. Various recent contributions to population geography are
reviewed and serve as exemplars of the themes of study. Life course studies, intergenerational
proximity and mobility, and gendered migration are discussed. By embracing theoretical and future
challenges, population geography is poised to make important contributions to our understanding of
people and places.
1. Introduction
September, 2004 marked the 150th anniversary of John Snow and the infamous water pump handle.
On August 31, 1854, there was a recurrent epidemic of cholera in London, England. Suspecting that
cholera was a waterborne contagious disease, Snow mapped the distribution of cholera deaths
surrounding the Broad Street pump which he suspected as the source of the disease. Snow (1849)
reviewed death records of area residents who died from cholera and interviewed household
members. He determined that most of the people having died from cholera lived near and had drunk
water from the pump. Snow presented his findings to community leaders and the pump handle was
removed on September 8, 1854, preventing additional cholera deaths. Often Snow is heralded as a
legendary figure in epidemiology because he provided one of the earliest examples of using
epidemiologic methods to study factors affecting the health and illness of populations. His study
advanced preventative action in the interest of public health. Snow is also heralded as a legendary
figure in population geography for he provides one of the earliest examples of the power of the map
to reveal spatial relationships and demographic processes.
Other early scientists used similar methods of mapping spatial distributions to examine
demographic phenomena. Mayhew (1861) mapped the incidence of crime in London along with
urbanization, poverty and disease. In the late 1880s, Charles Booth (1902) was one of the initial
scientists to map social and spatial distance based on his personal observations of neighborhood
differences in London. At the same time, Florence Kelley conducted a similar study by mapping
poverty in the slums of Chicago. From her field research Kelley learned that frequently several
families were sharing the same housing unit. Kelley was amongst the first to use cartograms to
represent family density in housing (CSISS, 2008). As these studies indicate, maps have enormous
analytical potential. Traditionally, maps have served a valuable role in understanding demographic
2
characteristics and events. Maps represent the geographic epistemology – thinking spatially. These
scientists were pioneers of population geography and spatial demography.
Population geography addresses the spatial distribution, characteristics, and variation of the
population. Often, the terms population geography and spatial demography are used
interchangeably. More specifically, however, population geography is a sub-disciplinary field of
geographic research that addresses demographic issues and population processes in an explicitly
spatial manner, with a focus on the connection between people and places. Spatial demography,
while similar to population geography, is the term usually used to refer to the formal empirical
methods used to make these connections between people and places. The importance of a spatial
perspective for demographic research has received considerable attention over the past few decades.
Geographers were amongst the first, but not the only population scientists, to embrace the
importance of spatial thinking. Today, population geography is a subfield of the discipline of
Geography. It is also a subfield of Demography.
This chapter provides an overview of population geography. Section two describes important
geographic concepts and spatial thinking, particularly the concept of scale. It then reviews the
intellectual heritage of population geography. Population geography is comprised of both a field of
study and methods of inquiry. The chapter then addresses the contemporary correspondence
between population geography and spatial demography. The significant role of Geographic
Information Systems is discussed, and special attention is given to demographic processes and their
spatial context. Section three describes the methods of population geography. Population
geographers embrace sophisticated conception of space and place. Particular attention is given to
the recent development of local, as opposed to global, measures of spatial association. Substantive
areas of research in population geography are reviewed in Section four. Traditionally, population
geographers have addressed the three components of population change – fertility, mortality, and
migration. Contemporary population geography has become more thematic and theoretically
sophisticated. This section reviews exemplars of recent themes in population geography. Section
five concludes with an assessment of theoretical challenges and future directions where population
geography is poised to make important contributions. Some issues also related to the
geographic/spatial aspects of population are addressed in other chapters in this volume. Small area
demography is addressed in Chapter 11 on applied demography, multi-regional population
projection is addressed in chapter 16 on multistate demography, and population projection by
countries and geographical regions of the world is addressed in chapter 22.
2. Population Geography and Contemporary Spatial Demography
There are three primary themes within a geographic perspective: spatial variation, humanenvironment interaction, and place-based populations. Geographers are interested in spatial
variation, aiming to understand how and why things differ from place to place across the landscape.
Geographers are interested in understanding how spatial patterns evolve over time. Central to the
study of Geography is the interaction between humans and the environment, both the impact of the
physical environment on society and the ways in which people adapt and change their natural
environment over time. The third theme considers the well-being of place-based populations and the
characteristics of places and regions. Population geography brings all three of these themes to the
comprehensive study of people and places.
2.1 Geographic concepts and spatial thinking
Location is the most basic geographic concept– where is a place? Yet, even this is a difficult
question to answer in the absence of other aspects of spatial thinking. Two of the most fundamental
geographic concepts are site and situation. Site refers to the characteristic features of a place,
3
whereas situation refers to the way in which a place is connected to other places either by natural
processes or by human processes, such as transportation or communication. A geographer’s
fundamental assessment of the site and situation of a place is a much more sophisticated conception
than one which depicts a place as merely a bounded area. For population geographers place is a
complex term that invokes numerous dimensions of spatial thinking, including such concepts as
analog, association, aura, accessibility, comparison, condition, connection, diffusion, distance
decay, distribution, exceptions, gradient, hierarchy, influence, interaction, mobility, pattern,
proximity, region, relationships and of course, scale – from the global to the local (Gershmel, 2008).
Tobler (1970) states that the first law of geography is that everything is related to everything else,
but near things are more related than distant things.
Typically population geographers ask questions that connect demographic phenomenon over time
and space. People and places are intricately interwoven, both spatially and temporally. It is common
for population geographers to ask how places are similar or different. How does one place compare
with another or with the nation or the world? How does an attribute affect the surrounding area?
Are there regional variations in demographic phenomena? What is the nature of transitions between
places? Are gradients gradual or abrupt? Are there spatial patterns such as imbalances, clustering,
and distinctive patterns or is the phenomenon random in occurrence? Is there evidence of spatial
associations or correlations? Are there places or locations that serve as exceptions to the rule or
aberrations from typical places? And how do things spread through time across space? What
avenues or barriers hinder or enhance diffusion? These are a few of the questions posed by
population geographers thinking spatially (Gershmel, 2008). Understanding spatial variation in
demographic phenomena is at the heart of population geography.
Population geographers have a heightened appreciation for the significance of scale. The scale of
analysis determines geographic representation. It is seldom a question of which scale is right, rather
which scale is appropriate. The important issue is to what extent the scale of analysis conditions the
findings of a study. Some studies are best undertaken with a global perspective, such as those
related to global climate, pandemics, and human rights. Others are distinctly the concerns of
families and households, such a completed fertility. Spatial aggregation increases as the scale of
analysis moves along the geographic hierarchy from people, to households, neighborhoods, cities,
regions, nations, international associations and the global context. As well, there are numerous units
of analysis that do not fit neatly into a geographic hierarchy as I have outlined. Stakeholders,
congregations of people, unions of nations, are just a few examples of groups that transcend
traditional scales.
The uncertainty associated with spatial aggregation is known formally as the modifiable area unit
problem (MAUP). First described by Openshaw (1984), the areal units used in many geographical
studies are arbitrary, modifiable, and subject to the caprice of the researcher conducting the
aggregation. Researchers take a variety of approaches to contend with the modifiable area unit
problem. Some studies conduct the same analysis at different levels of spatial aggregation to assess
the relative sensitivity of their findings. Others intentionally jump scales to produce alternative
knowledge. In population geography, as with other spatial sciences, the modifiable area unit
problem is intertwined with the practical challenge of integrating data across geographic scales.
2.2 The Making of Population Geography
Population geography owes its formal beginnings as a sub-discipline of Geography to the
Presidential address given by Glen Trewartha at the annual meeting of the Association of American
Geographers in 1953. At the time, Geography was divided into two areas, generally: Physical
Geography and Cultural Geography. In his address, Trewartha presented an organizational triade for
Geography comprised of population, the natural earth and the cultural earth. In more contemporary
4
terminology this triade includes people, the physical environment, and the cultural environment. He
placed the natural and cultural environments at the base and population at the top (Figure 19.1).
Figure 19.1. Trewartha’s depiction of Population as the apex of the field of Geography.
Trewartha was a physical geographer and a climatologist. Trewartha was making the call for an
anthropomorphic view of the geographical sciences. Population was what made the physical and
cultural studies important. He was encouraging the scientific study of populations, particularly as
they relate to the physical environment. In Trewartha’s words, “Population is the point of reference
from which all other elements are observed and from which they all, singly and collectively, derive
significance and meaning. It is population which furnishes the focus.” (Trewartha, 1953: 83).
In geographical texts it is commonplace to give Trewartha recognition as the one who initiated
population geography (although it has not come to be the pinnacle of the discipline of Geography as
Trewartha proposed). Trewartha made an impact on the discipline of geography, but he was not
alone in working to establish population as a scientific field of study. The classic and the most
influential book on demography “The Study of Population: An Inventory and Appraisal” by Hauser
and Duncan (1959) includes a chapter titled Geography and Demography. Clearly, population
geography is a typical interdisciplinary field of these two scientific disciplines.
Weeks (2004) is one of the few scholars to explain the larger context of Trewartha’s address.
Trewartha was at the University of Wisconsin which was, and still is, a vibrant environment for
population studies. The call for the scientific study of population was being heralded throughout the
academy. Moreover, the call was being heard! The Population Council, now a leading voice in
policy-oriented research, was established by the Rockefellers at this time. As well, the International
Union for the Scientific Study of Population (IUSSP) was working with the United Nations to plan
for the World Population Conference that was subsequently held in Rome in 1954.
Throughout the 1960s and 1970s population geography blossomed and, in a sense, came of age in
the 1980s by which time it had become dominated by spatial analysis and quantitative methods. It
was a sizeable sub-field of Geography whose domain was the scientific study of population.
Notable in this development was the work of Wilbur Zelinsky. In his 1966 text A Prologue to
Population Geography, Zelinsky distinguished the population geographer from the demographer in
the following manner. He depicted the demographer as being “concerned with the intrinsic nature,
the universal attributes, of populations, with the systematic principles governing their composition,
socioeconomic correlates, behavior, and changes: the spatial is incidental to this purpose.” In
contrast, he described population geography as the “science that deals with the ways in which the
geographic character of places is formed by, and in turn reacts upon, a set of population phenomena
that vary within it through both space and time as they follow their own behavioral laws, interacting
one with another and with numerous non-demographic phenomena.” (Zelinsky, 1966:3).
The scientific study of population influenced other areas of Geography as well during the 1970s.
The sub-field of medical geography (akin to spatial epidemiology) was established during the late
1970s. Melinda Mead’s 1977 article in The Geographical Review titled “Medical Geography as
Human Ecology: The Dimension of Population Movement,” is frequently cited as the cornerstone
of establishing medical geography as a sub-field in Geography. Interestingly, in this paper she
includes a diagram that is essentially the same as Trewartha’s triade, yet it focuses on health as the
outcome of the dynamic interactions between the three dimensions of population, environment and
culture. As well, the scientific study of population became a core concern of the sub-field of
economic geography and the field of regional science. It was Hägerstrand who integrated people
into regional science. Hägerstrand asserted that time and space were not independent of each other.
His contribution was as eloquent as it was influential. His research focused on innovation diffusion
5
as he studied the spread of new technologies. People and their ideas move together. Prior to his
writings, regional science and population geography did acknowledge the importance of both space
and time but tended towards keeping them separate analytically. Hägerstrand’s work was
overwhelmingly influential in Geography and Regional Science because his work situated people in
time and space. Situating people in time and place is at the core of contemporary population
geography.
2.3 The Development of Spatial Demography
Sociologists Voss et al. (2004) explain that until the mid-20th century virtually all demography, in
the United States at least, was spatial demography. They make this claim because they define
spatial demography as the formal demographic study of areal aggregates, i.e., of demographic
attributes aggregated to some level within a geographic hierarchy. They argue that in the years since
about 1950, the scientific study of population came to be dominated by attention to the individual as
the agent of demographic action. It was at this time that spatial demography (narrowly defined as
the study of areal aggregates) gave way to micro-demography. Moreover, the field of spatial
demography remained disengaged from the important trends that emerged in the 1980s in the
disciplines of geography, regional science and spatial econometrics. Luc Anselin published a
comprehensive book on spatial econometrics in 1988. Evidently, while Trewartha was encouraging
geographers to embrace population as their raison d’être, and geography and regional science were
developing conceptual approaches for studying the dynamic interaction of people and places,
sociologists were backing away from place-based studies. By the start of the 21st Century,
demography was described by some as possessing a rich methodology, but largely lacking a spatial
perspective. A spatial perspective is more complex than simple areal aggregation.
In a recent paper in Demography, Entwisle (2007) calls for ‘putting people into place’. Her paper
encourages demographers and population scientists to take the lead in explaining behavior and
outcomes in relation to a potentially changing local context. Entwisle is referring to the virtual
explosion of empirical studies on neighborhoods and health that have occurred over the past few
decades but as yet have made little headway towards understanding the significance of place and
location. Similar explosions of ‘studying the local’ can be found in other areas of research such as
segregation, criminology, educational attainment, intergenerational poverty, and so forth. She
asserts that all the necessary tools and techniques have been in place for 20 years or so, and yet
empirical evidence points to an absence of theoretical models where space plays a causal role in
demographic outcomes. Conspicuously absent from her article is geographic research. The only
explicitly mentioned geographic work is an example of agent-based modeling. More
interdisciplinary collaboration will enable social scientists to further appreciate the rich potential of
the spatial perspective. Echoing the distinction between demography and population geography
raised decades earlier by Zelinsky, Entwisle’s plea is directed towards the scientists within the
demographic community that continue to consider place as little more than a container. Population
scientists such as these are encouraged to embrace the advances made in spatial demography.
Theoretical models in which space plays a causal role in demographic outcomes are needed for
spatial demography to assume more of an analytical role in demographic research.
Indeed, spatial demography is flourishing, as some demographers and population scientists have
embraced developments in other disciplines, the most notable being spatial statistics, regional
science, and geography. Initially, the field of spatial analysis was dominated by regional science
and spatial econometrics. Contemporary spatial demography involves sophisticated spatial analysis.
In recent decades the range of spatial approaches in the social sciences has expanded considerably
to include analytic cartography, Geographic Information Systems (GIS), network analysis, timespace modeling and visualization, spatial optimization techniques, spatial interaction modeling,
agent-based modeling, simulation methods, point-pattern analysis, and spatial econometrics. Still, a
6
great deal of contemporary spatial demography is not being conducted by demographers but rather
by spatial scientists.
Fearon conducted a study in 2005 of published demographic papers that use spatial analysis based
on the Spatial Population Index. He found that a quarter of all articles published from the period
1990 to 2003 were in the discipline of geography. A third was published in economics due largely
to the prolific use of spatial econometrics. Other disciplines that were well represented were urban
studies, sociology, public health, regional science, urban and regional planning, and environmental
studies and policy. Amongst the 2000-plus demographic articles that included spatial analytic
methods the following approaches were recorded: neighborhood effects, spatial diffusion, spatial
data analysis, locational decision making, spatial econometrics, spatial regression, spatial dynamics,
spatial cluster, spatial effect, mapping, population potential (accessibility), spatial economy, spatial
statistics, spatial decision, spatial autocorrelation, gravity model, cartography agglomeration
economies, spatial heterogeneity, spatial equilibrium, spatial correlation, spatial interaction, spatial
cluster analysis, urban hierarchies, industrial location, spatial autocorrelation and locational
analysis. And this study doesn’t include papers since 2003. Spatial demography is alive and well
and very sophisticated.
The diffusion of methods of spatial demography has been fostered by the integration of spatial
analysis and geographic information systems (GIS) and emerging sources of spatially referenced
data (geo-coding). The Center for Spatially Integrated Social Science (CSISS) at the University of
California Santa Barbara, which was established by MF Goodchild, RP Appelbaum and Luc
Anselin, has facilitated this integration greatly. The ‘spatial awakening’ in the social sciences can be
attributed to a combination of factors. Spatial demography became a funding priority by both the
National Science Foundation and the National Institute for Child Health and Development. A
number of innovative Population Centers funded by the National Institute of Health supported
spatial demography programs (Brown, Wisconsin, University of North Carolina, and Penn State).
As well, the integration of theory, greater computational ability, and the availability of specialized
software led to further developments in spatial statistics and spatial econometrics. CSISS now
partners with the Population Center at the Pennsylvania State University specializing in GIS, spatial
statistics, cartographic visualization and demographic science.
Research in population geography has a long tradition of respect for scientific rigor, conceptual
clarity, objective empirical measurement and solemn concern with the nature and quality of data. In
the next section, the traditional and contemporary methods of population geography will be
reviewed along three themes: the geographic distribution of populations, migration, and spatial
analysis.
3. Methods of Population Geography
Population geography is concerned with the spatial distribution and composition of the population.
Like other population scientists, geographers commonly use measures such as median age, sex
ratio, the dependency ratio, labor-force participation measures, population pyramids, percentage
change, rates of change, crude death rates, age-specific and cause-specific death rates, standardized
death rates, crude birth rates, period and cohort measures of fertility and life table models.
Population geographers also model rates of population growth (geometric, exponential, and logistic)
to understand demographic change. These are standard techniques in the demographer’s toolbox
and are well reviewed elsewhere in this volume.
This section presents analytical methods used by population geographers and spatial demographers.
Many of the methods have been derived from techniques developed in regional science, spatial
econometrics and geostatistics. The methods are grouped into three general areas: 1) analyzing the
7
geographic distribution of the population, 2) analyzing population movement and migration, and 3)
recent developments in spatial analysis towards local spatial statistics. Much of the material
presented in this section is distilled from the 1994 book by David Plane and Peter Rogerson titled
The Geographical Analysis of Population with Applications to Planning and Business. This book is
a rich and detailed collection of geographical techniques for population scientists and a must read
for any student of population geography and spatial demography.
3.1 The Spatial Basics – Geographic Distribution of Population
Population density is a fundamental measure in demographic research which is of particular interest
to population geographers. Simple to calculate, population density is measured by the ratio of the
size of the population to the size of the geographic area that contains the population. It is an
explicitly spatial concept. According to the United Nations, in 2005 the Netherlands had a
population density of 395 per km2, whereas Canada’s population density was 3 per km2. The
variation in population density is striking. This example illustrates another important concern for
the population geographer- spatial variation. As a nation, Canada has a large land mass and a
relatively small population size. However, the population is distributed unevenly across the urbanrural continuum, concentrated in a few very large cities.
Figure 19.2 World Population Dot Density Map by Country
3.1.1
Descriptive spatial statistics
For descriptive measures of a spatial distribution (such as a point pattern) the most basic approach is
to calculate the mean center and the standard distance. These are spatial analogs of the classic
descriptive measures of mean and standard deviation. Unlike an aspatial variable, a spatial point
pattern has locational attributes which can be measured in 2-dimensional space, using either latitude
and longitudinal, or locations on a grid referenced with an x-axis and a y-axis. The mean center of a
point pattern has two coordinates: the mean of the x-values and the mean of the y-values such that
 X i Y   Yi
Xc 
c
n .
n ,
(1)
The standard distance is measured as
SD 
 X
i
 Xc
   Y  Y 
2
i
n
2
c
.
(2)
To describe a variable or attribute that is distributed across the point pattern the same approach is
used, but the mean center and the standard distance are weighted by the values of the attribute at
each location, producing a weighted mean center and a weighted standard distance. The Census of
the United States calculates the weighted mean center of the population of the United States after
each census. The mean center of the population has been steadily moving west and south with each
decennial census.
3.1.2
Index of Concentration – The Hoover Index
In 1941, E.M. Hoover used a method to measure the level of population concentration of a given
population distribution. Originally Hoover was studying interstate population movement. The index
is practical and popular and has come to be named after Hoover.
r
H  50 pi  ai
i 1
8
(3)
where pi is the share of the entire study regions population P found in a subarea i at a given moment
in time, and ai is the share of the entire study regions land area A in subarea i. Part of the appeal of
this measure is the ease of interpretation of the index. The index is theoretically bound by 0
(meaning the population is equally distributed across all areas relative to their size) and 100
(meaning the population is concentrated entirely in a very small area of the larger study region).
The two theoretical bounds are extremely unlikely. The value of the index can be interpreted as the
proportion of the population that would need to relocate to make the population of similar density
across the study region.
3.1.3
The Lorenz curve
The Lorenz curve was originally developed by M.O. Lorenz in 1905 to study the distribution of
wealth. It has become a very common technique to graphically portray patterns of inequality. As it
happens, the initial call to use the Lorenz curve came from an article published by JK Wright in
1937. Wright was the director of the American Geographic Society from 1938 to 1949. Wright’s
article encouraged the use of the Lorenz curve to study geographic areas.
The Lorenz curve is a simple but effective graphing technique that compares the cumulative
percentage of two variables displayed on an x and a y axis, respectively. If there is complete
correspondence between the two variables (equivalence) then the graph displays a diagonal line
from the bottom left corner to the top right corner of the graph. Any deviation from this diagonal
line represents the amount of difference between the two cumulative percentage values, or more
simply, inequality in the relative distribution of two variables. Wright made the case for the Lorenz
curve to be used to compare geographic areas for differences in the association between population
distribution and income, and other important measures. Nowadays the Lorenz curve is used in
studies of geographic inequality, such as environmental racism and accessibility.
3.1.4
Population potential - the accessibility index
Accessibility is an important concept in geographic research. Measuring and comparing the
accessibility of a population to a particular site can be vital for successful site location. Accessibility
can be measured in a number of ways, such as the distance traveled, the cost of travel, or the time to
travel to a particular site from all areas in a geographic region. Warntz (1964) developed a measure
of accessibility called the population potential of a site. Population potential is measured in the
following manner:
r 
p 
 j   i 
 
i 1  d ij 
(4)
where  j is a measure of accessibility (population potential) at any location (site) j; pi is the
population of each subarea and dij is the distance from the centroid of each area i to the site j. Since
this ratio of population to distance is summed across all the surrounding areas the larger the
population potential of a given location the more accessible a site is for the surrounding population.
These measures are often used to determine the optimum location for services targeted towards
specific populations. For example, in choosing where to locate a public health facility the
population potential could be calculated using the complete population counts or it could be
calculated relative to the spatial distribution of a specific subgroup of the population in need. By
comparing the population potential of various sites, one can determine the optimal location to
maximize accessibility to a site for a target population. This index is used in marketing studies
considerably.
9
3.1.5
Location quotients
The traditional use of location quotients in geographic research has been the study of industrial
specialization across the landscape. Location quotients (LQs) are now a common method within
population geography to study the spatial distribution and clustering across the landscape by
population groups and subgroups. Location quotients are computed in the following way:
  Pij  
   
 Pj 
LQ      *100
P
 i
  Pt  


(5)
where Pij is the population in subarea j in subgroup i, Pj is the total population in subarea j, Pi is the
population in subgroup i, and Pt is the total population of a region. In the numerator we have the
ratio of a particular subgroup of the population in a specific area to the total population of that
specific area. In the denominator we have a ratio that represents the proportion of the population
across the whole region in that particular subgroup. Therefore, values above 100 represent spatial
concentrations of a population subgroup and values below 100 represent fewer members of a
population subgroup than expected given their presence in the larger region. A value equal to 100
indicates a subarea has the same proportion of a population subgroup as is present in the larger
region. Practically, subareas are measured at the level of zip code, census tract, or neighborhood.
Location quotients can be readily mapped to assist with visualizing spatial concentrations.
3.1.6 Segregation Measures
Population geographers are often interested in the spatial distribution of the population, particularly
the extent to which ethic groups are segregated in metropolitan areas. There are a variety of
measures of segregation. Two that are frequently used in geographic analysis are the Index of
Dissimilarity and the Exposure Index. Frequently, these methods have been used to develop school
busing plans aimed at achieving racial integration within schools. Both of these measures are easy
to calculate, and are readily understood in non-technical terms.
The Index of Dissimilarity (D) measures the extent to which the ethnic composition of each subarea
(usually some representation of a neighborhood) reflects the aggregate ethnic composition of a city
or metropolitan area. The index takes on values from 0 to 100, and the larger the dissimilarity index,
the greater the amount of segregation.
For example, if we let Bi and Wi represent the Black and White populations respectively in a
subarea i and let B and W represent the Black and White populations in the entire metropolitan area,
then the formula for the index is:
B
W
D  100( 12 )i  i    i 
 B  W 
(7)
The Dissimilarity index can be interpreted as the percentage of either population group that would
have to move in order for each subarea to reflect the ethnic composition of the metropolitan area.
The interpretation is similar to the Hoover index, but the scale of analysis is different.
10
Another commonly used measure of segregation is the Exposure index (E). The basic concept of the
Exposure index is to define the probability that a member of a specific subgroup comes into contact
with a member of another specific subgroup in any random “exposure” to a member of the general
population of his or her own subarea of residence. This index must be calculated separately for each
group. If Ti represents the total population in subarea I, the exposure index can be calculated for the
intragroup or intergroup exposure.
The intragroup exposure index is used to determine the probability that a member of a specific
subgroup comes into contact with someone within their own group,

 
Ebb  100 i  Bi  *  Bi 
 B   Ti  
(8)
The intergroup exposure index is used to determine the probability that a member of a specific
subgroup comes into contact with someone from a different group,

 
Ebw  100 i  Bi  *  Wi 
 B   Ti  
(9)
These indices provide comparable and intelligible measures of segregation for metropolitan areas.
Note that the indices can be used for other types of population diversity as well. More recent
developments with measures of segregation enable more than two groups to be considered
simultaneously.
3.2.
Measures of migration – population on the move
Migration is an explicitly spatial demographic process, so it is not surprising that the lion’s share of
research in population geography is focused on migration.
3.2.1
Migration rates
Measures of migration begin with rates for in-migration, out-migration, net-migration and grossmigration, where O represents the size of the population moving out of an area, I the size of the
population moving into an area, and P is the size of the population in the area initially. The
measures are multiplied by a constant, usually 1000, to create a rate. Although they are measured in
the same manner, outmigration and immigration refer to local scales of population movement. At
the international scale the terminology becomes emigration and immigration.
O
*1000
Outmigration rate = P
(10)
I
*1000
Inmigration rate = P
(11)
I O
*1000
Net migration rate = P
(12)
I O
*1000
Gross migration rate = P
(13)
Gross and net migration rates can pose difficulties in making comparisons of population growth due
to differences in the population size of places. Small and large denominators can be difficult to use.
11
Consequently, some researchers prefer to use a measure of demographic effectiveness to assess the
impact of population exchange that results from migration.
3.2.2 Demographic efficiency of migration
How effective are migration streams across a system? This question has been tackled by Thomas
(1941) by calculating a measure of demographic effectiveness, and further addressed by Shyrock
(1964) who called it demographic efficiency. Letting N represent net-migration, and T represent
total migration, the demographic effectiveness of a given region j is measured in the following
manner:
N 
E j  100  j 
 Tj 


(14)
Some researchers prefer this approach to measuring the impact of migration flows on population
change because it is based on population flows in and out of the same region, and therefore it is not
influenced by the population size of a place. Plane (1984) extended this approach and developed a
summary measure for system effectiveness to assess the short-term changes in U.S. population at
the interstate level.
3.2.3 Migration characteristics and expectancies
Migrant characteristics vary across time and space. Nonetheless, age reliably dominates all other
migration characteristics. The age profile of migration is sufficiently reliable that Rogers (1966)
developed and applied model migration schedules (Rogers, 1966). In a similar vein, population
geographers have developed tables of migration expectancies by using age-specific rates of mobility
and mortality. This approach is similar to the use of a life table to determine life expectancies. Life
tables are used to determine life expectancy beyond a certain age group. Given that someone lives
to be a certain age, and assuming the age-specific profile of mortality holds true, the typical life
expectancy can be read of the life table.
Similarly, with a table of migration expectancies one can ask – how likely is someone to move once
they have reached a certain age? Or, what is the expected number of moves after a certain age? If Ex
is the migration expectancy at a given age, Ma is the mobility rate for people in the age group
beginning at age a, La refers to the size of the stationary population for the age group, and l x is the
population at exact age x, then the formula to calculate the migration expectancy for a given age is:


  M a La 

Ex   a x
lx
(15)
Both La and lx are derived from a standard life table. Measures of migration expectancies are
particularly useful for anticipating future age patterns of migration.
3.2.4 Spatial shift-share analysis
Regional economists and economic geographers have used shift-share analysis to understand
changes over time in the sectoral structure of employment. Plane (1987) used shift-share analysis to
understand the changing age structure of migration. Shift-share analysis is a technique used to
decompose migration patterns. Just as regional economists study sectoral shifts in the employment
structure, population geographers study age shifts in the population migration structure. Like many
of the migration methods previously discussed, the aim is to gain an understanding of the changes
in the volumes of population flows between places.
12
Spatial shift-share analysis is derived from a matrix composed of interregional migration over time.
Each dimension of the matrix represents a different moment in time. The shift-share analysis is
designed to decompose the migration flow patterns across the matrix on the basis of three separate
population quantities: a population base component (Bij), a mobility component (Uij), and a
geographic distribution component (Gij), such that the change in migration flows from one time
period to another ( M ij ) becomes:
M ij  Bij  U ij  Gij
(16)
The base population component represents the different population at risk across various places due
to differences in the original population size. Pi represents the amount of population change in an
area. The base component measure uses the migration rate from the first time period and applies it
to the population at risk in the second time period:
M 
Bij   ij  Pi
 Pi 
(17)
The mobility component captures the amount of population change (positive or negative) that is
attributable beyond the base population because of the rate of change from migration to all areas
(Mi*),
 M   P 
i*
U ij  
   i  M ij

M
 i *   Pi 
(18)
The third component is the geographic distribution component which gives the proportion of the
change of migration from region i to j which is attributable to the destination specific flow.
 M ij   M  
i*
Gij  

M
 M ij   M i*   ij






(19)
Gij can be determined readily after calculating the first two components, such that
Gij  M ij  Bij  U ij
(20)
In this form the shift-share technique assumes spatial independence between the geographic units.
Since this is frequently an unrealistic assumption, recent applications of this technique utilize a
spatial weight matrix to represent the geographic dependence between places.
3.2.5 Gravity models and economic gravity models
The gravity model has been used for decades to crudely predict gross migration between two places.
It is based on Newton’s universal law of gravitation.
PP
M ij  i 2 j *1000
d ij
(21)
Simply stated, the size of the migration flow anticipated from one area to another is directly related
to the collective size of the populations (mass) and inversely related to the distance between two
places. Implicit in this model is the concept of distance decay. The expected interaction between
two places declines as the distance between them increases. The friction of distance serves as a
deterrent for cultural and spatial interactions. Gravity models have been used for decades with
13
considerable refinement to this basic representation. More sophisticated versions of the gravity
model include different specification for the friction of distance, the use of different distance
deterrent metrics for different origins and destinations. Other common refinements focus on
differences in economic opportunities between two places and not just population size.
3.3 Methods of spatial analysis
The recent integration of methods of spatial analysis with software programs and geographic
information systems has vastly increased the number of studies using spatial analytical methods,
even though many of these methods have been around for many decades. At the other end of the
spectrum, the use of computers for computationally intensive procedures has enabled the
development and integration of more specifically local spatial analytical methods. Spatial analysis
has experienced a turn from the global to the local – a turn that has significant promise for the
development of geographic and spatial theory. In this section we will review global measures of
spatial autocorrelation and then their local counterparts.
3.3.1. Measures of spatial autocorrelation: Global to Local Moran’s I
There are a variety of analytical techniques to test for spatial dependence. These tests specifically
measuring the extent of spatial autocorrelation found in a spatial distribution. One such measure has
been devised by Moran (1950) and can be applied to area patterns and to point patterns. For areal
data the equation for Moran’s coefficient is:
n c X i  X X j  X
I
2
J X  X





(22)
where I is Moran’s spatial autocorrelation coefficient, n is the number of areas in the study region, J
is the number of joins, X is a value for an area (ordinal or interval), Xi and Xj are two contiguous
areas (on either side of a join) and c is a pair of contiguous areas.
Moran’s I is calculated and then compared with the expected value derived from either a random
distribution or a normal distribution to determine if the observed spatial pattern came about by
chance. Evidence of spatial autocorrelation indicates the pattern did not occur randomly. Of course,
this is merely an indicator of spatial dependence, it does not constitute an explanation for the
observed spatial distribution.
The test for spatial autocorrelation in point patterns is a revised version of Moran’s coefficient.
With point data, instead of considering the relationship between pairs of contiguous area values, it
is necessary to measure the relationship between all pairs of point values, taking into account the
distances between them. If there are n points, there will be n(n-1)/2 possible pairs of points. The
technique can be quite easily computerized for it is very tedious to calculate manually. The form of
Moran’s I for point pattern is
n  pWij X i  X X j  X
I
2
 pWij   X  X 
(23)
where I is Moran’s spatial autocorrelation coefficient, n is the number of points, Wij is the weight
given to the relationship between two points i and j and p is a pair of points. The weight is usually
the reciprocal of the distance between the two points. The distance between points i and j are
defined as dij, thus Wij = 1/dij. Each weight is meant to be a measure of the influence exerted by one
point on another. The use of the reciprocal of distance as a weight implies that the influence
decreases with distance.



14
Both of these measures are global measures in that the procedure provides a single coefficient to
describe the complete spatial pattern. Sometimes the researcher wants to know if there is a cluster
of events around a single or small number of places. For example, does disease cluster around a
toxic waste site? Does crime cluster around exotic dancing establishments? We need a method to
detect spatial clustering, and local spatial statistics are the suitable method.
Anselin (1995) has developed a local spatial autocorrelation statistic called Local Moran’s I. The
local measure is called a LISA statistic – local indicator of spatial association. Anselin defines LISA
as having the following properties: 1) The LISA for each observation gives an indication of the
extent of significant spatial clustering of similar values around that observation, 2) the sum of
LISAs for all observations is proportional to a global indicator of spatial association. The formula to
calculate the local measure of Moran’s I is,
X X
Ii  i 2
 wij  X j  X 
Si
j 1,i  j
(24)
where
N
Si 
2

j 1, j i
wij
N 1
 X2
.
(25)
In the case of applying the weights for distance bands
1
wij (d )  m
d .
(26)
The interpretation of the local Moran’s I is very similar to that of global Moran’s I. Values of Ii that
exceed the expected values indicate positive spatial autocorrelation, in which similar values, either
high values or low values are spatially clustered around point i. Values of Ii below expected values
indicate negative spatial autocorrelation, in which neighboring values are dissimilar to the value at
point i.
There are two types of spatial weighting methods that may be used. The first spatial weighting
method uses a spatial weights matrix recorded and referenced in a file. This enables researchers to
introduce their own conceptualization of the spatial structure. The second type uses a weight that
raises the distance by some power function. This statistic is calculated for distance bands only
(Figure 19.3). Note that each Ii value for a given site i represents association between the ith site
and only the j values in a given band.
Figure 19.3 Distance Bands for Calculating LISA using Local Moran’s I
3.3.2. Global to Local: Getis-Ord G to Getis Gi
Getis-Ord General G (1992) is a global measure of spatial association. It is a multiplicative measure
of overall spatial association of values which fall within a critical distance of each other. An
observed G(d) value higher than the expected G(d) value indicates a clustering of high values, and
an observed G(d) lower than the expected G(d) indicates a clustering of low values. It can only be
computed for positive variables. For a chosen critical distance d, G(d) is
15
N
G d  
N
 w (d ) x x
ij
i 1 j 1
N N
i
 x x
i 1 j 1
i
j
, j 1
j
(27)
where Xi is the value if the ith point and wij(d) is the weight for point i and j and for distance d. This
is then compared to the expected value, which is derived theoretically. This global measure
provides a number that determines if there is a general pattern of high values clustering or low
values clustering.
Getis also developed a local version of this spatial statistic to detect clustering – the Gi statistic.
There are two variations of this statistic, depending on whether the unit (observation) i around
which the clustering is measured is included in the calculations. Gi does not include the observation
around which the measure is being calculated. Gi* does include the observation around which the
measure is being calculated.
The purpose of this measure is to test whether spatial clustering exists around a certain location (i)
in the following manner,
 jWij X j , j  i
Gi 
 jXj
(28)
where Gi is the measure of local clustering of attribute X around i, Xj is the value of X at j, Wij
represents the strength of the spatial relationship between units i and j which can be measured as
either a binary contiguity variable or a continuous distance-decay measure. If high values of X are
clustered around i, Gi will be high. If low values of X are clustered around i, Gi will be low. With no
clustering of values around i then Gi will be an intermediate value.
3.3.3 Agent-Based Computational Demography
Microsimulation approaches such as agent-based models are at the forefront of many disciplines.
Agent-based models have been used to study mobility, residential segregation and tipping points.
These approaches capture the multifaceted aspects of neighborhoods. Simulations provide a means
to experiment with conditions and constraints and observe the effect on micro-level behavior and
system level outcomes. The method is largely inductive. Agents are individuals that follow rules
governing the system. One of the appeals of these models is that both macro and micro processes
are endogenous. In geography ‘spatially explicit’ agent-based models have been developed to
describe landuse change in a variety of settings. If these methods could be designed to include
dynamic social networks then these methods would make an enormous contribution to spatial
demography and population geography. Currently, this is the leading edge of research in spatial
demography. The general appeal of agent-based models is the ability to explore complex behaviors
and outcomes stemming from simple rules.
3.3.4 Geographically weighted regression
When using ordinary least squares regression (OLS) with spatial data researchers frequently violate
the assumptions of the model. Regression assumes independence among the observations. To
determine if model assumptions have been violated, researchers examine regression residuals for
patterns deviating from homogeneity. Spatial autocorrelation in the residuals violates the
assumptions of OLS regression modeling.
16
Traditional multiple regression analysis is a global approach, in that the OLS criteria generate a
single equation with one intercept and one parameter estimates for each of the variables in the
model. OLS regresses towards the mean. It is a global measure that indicates the generalized nature
of the relationship between a dependent variable and independent variables. Values of the
regression coefficients are global in the sense that they apply to the study region as a whole.
Geographers and regional scientists have traditionally mapped the residuals of a regression model to
detect places where the global model performed well or poorly, depending on the size and direction
of the residuals. Once mapped, these residuals can also be scrutinized using methods to detect
spatial autocorrelation and clustering as outlined above.
Fortheringham, Brunsdon and Charleton (2002) have developed a method of geographically
weighted regression analysis that provides local measures rather than a global measure.
Realistically, it is most probable that coefficients vary across space. Geographically weighted
regression is a regression approach that accounts for spatially varying parameters. The GWR
technique is based on local views of regression, as observed from each data point. In a GWR
relatively large weights are ascribed to points near the location and relatively small weights are
given to observations far from the location. There is a continuous surface of parameter values. In
the calibration of the GWR model it is assumed that observed data near to point i have more of an
influence in the estimation of the regression coefficients than do data located farther from i. The
dependent variable at location i is modeled such that
p
yi  bi 0   bij X ij   i
j 1
(29)
X
where P are independent variables, ij is the observation on variable j at location i. Note the b
coefficients have i subscripts making them location specific. In GWR an observation is weighted in
accordance with its proximity to point i so that the weighting of an observation is no longer constant
in the calibration but varies with i. The most common form for weights is a negative exponential
function of squared distance, so points farther away will be assigned lower weights. In essence, the
GWR equation measures the relationships inherent in the model around each point i. The weights
depict spatial relations that are in keeping with Tobler’s first law of geography. GWR produces
localized parameter estimates. It also produces localized versions of all standard regression
diagnostics including goodness-of-fit measures such as r2.
GWR is available as a stand-alone software package and it is very easy to use. Users need to
consider an appropriate kernel and bandwith to use in their analysis. This makes the user determine
a priori the metric that best reflects their theoretical conception of the influence of data points
within certain distances (bandwidth). Studies have used geographically weighted regression to
examine different causes of spatial nonstationarity. The software can accommodate a
geographically weighted regression model for logistic regression models as well. Researchers in
spatial demography and population geography are beginning to take full advantage of the potential
of GWR for making conceptual advances in our understanding of the significance of place and
space.
Multilevel modeling is another regression-type approach that has been developed to accommodate
the endogenous and nested nature of the various scales of measurement. For example, individuals
are nested within families, which are nested within neighborhoods, which are nested within labor
markets or school districts, and so forth. Standard regression methods can not handle the spatial
dependence that is introduced when several independent variables are included in a model that are
related or scale dependent. In these situations population scientists apply multilevel models which
explicitly include in the model terms to account for the various levels of dependence. These models
17
are very potent in their ability to integrate social and spatial measures. They have also served to
reinforce the significance of scale in demographic research.
To summarize, there are a host of methods used in population geography. The methods have been
grouped within three general areas of inquiry: geographic distributions, population movement, and
spatial analysis. Many were developed decades ago within regional science and some have been
developed more recently, enabled by the power of computational processes. These methods and
techniques enable population geographers to situate people in time and place. Clearly, the methods
of population geography have come a long way from areal aggregation. In the next section we turn
to the substantive themes of geographic demographic inquiry.
4. Themes of population geography
Since the initiation of the journal, Progress in Human Geography has provided an annual progress
report on the state of the art of population geography. These reports are a good resource from which
to chronicle the themes in population geography over time. They also serve periodically as a forum
to draw attention to what is absent in the field. Traditionally, population geography focused on the
spatial aspects of the three demographic themes of fertility, mortality and migration only. This is no
longer the case. The scope of population geography has shifted with the times. Currently,
population geography practically is obsessed with migration and anything related to migration. In
part this is because migration (more so than fertility and mortality) is inherently spatial. In many
places, migration is the driver of population change rather than natural increase. As well,
historically speaking we live in an age of migration and enormous levels of population circulation.
Yet, we also live in an age of adoption, aging, assimilation, below replacement fertility, care giving,
famine, fear, flexibility, global warming, global inequality, information, overconsumption,
vulnerability, and wellbeing, to name just a few contemporary concerns. The modern world is
vastly different than it was a generation ago, and the themes of research in population geography
have changed with it. The demographic perspective remains central – but the power of the
demographic perspective comes from situating people in place, across time and space.
4.1 Whither fertility and mortality
During periods of high fertility, such as the baby boom, population geographers have given
attention to fertility and the spatial variation of fertility. However, the study of fertility is not as
visible in population geography now as it has been in the past. Fertility studies are conspicuously
absent. This is likely due to very low levels of fertility in many nations. Arguably, this is an area in
need of more research. There is a distinct geography to fertility differentials, and the geography of
the consequences of below-replacement fertility is as yet poorly known, but of considerable import
for population policy. Recently, Boyle conducted studies of fertility differentials in Scotland and
found there to be striking geographic variations. What has caused these local variations in fertility
levels remains largely unknown. Understanding global to local variations in fertility levels is an
area set for study.
Similarly, with increases in life expectancy, population geographers have also turned away from
mortality studies. There has been a decline in interest amongst population geographers in the study
of mortality. What little interest there has been is largely historical or based in the developing
world. In contrast, there is a vast health inequalities literature with which population geographers
are becoming more fully engaged. Recently, Boyle provided evidence of a striking geography to
mortality and morbidity. Geographic inequalities in mortality are at an historic high in Britain,
which begs interesting questions about the health selectivity of migrants and the impact of
inequalities on the health and wellbeing of a population. As with fertility, understanding global to
local variations in mortality and morbidity is an area ripe for research.
18
4.2 Migration events
In contrast, population geographers overwhelmingly have been engaged with the study of migration
and mobility. The vast majority of studies in population geography are related to the movement of
population across all scales, and the consequences of these movements for local areas and larger
societal concerns. The scales of migration studies range from international to transnational to
regional to the local. Migration refers to population movement across a political boundary, whereas
residential mobility refers to population movement within the same political boundary. Mobility
studies serve as a good example of the modifiable area unit problem. Residential mobility is
conceptualized as mobility within the same labor market area, with adjustments in the housing
market. These are often motivated by job changes or shifts in housing consumption, and tend to be
of short distance. There is a close link between the life course of individuals and residential
mobility. Migration has generally been conceptualized as long-distance moves between labor
markets, motivated by employment reasons or familial ties and social networks. Another important
distinction when conceptualizing mobility is the difference between voluntary and forced mobility.
Forced moves occur in the context of evictions, natural disasters or political turmoil involving
human rights violations.
Contemporary studies in population geography either study migration directly or indirectly by
analyzing the material consequences stemming from migration. Some of the main themes that have
been prevalent in research are life course studies and longitudinal analysis, intergenerational
studies, gendered family migration, assimilation processes, and neighborhood studies and
residential mobility.
4.3 Life course perspective
Like other disciplines that embrace population studies, the life course perspective serves as a
unifying theme for a great deal of research in population geography. Life course studies examine
the interconnection between different aspects, called careers or trajectories, of people’s lives. A
career is the sequence of events that occur over time in any one aspect of someone’s life.
Individuals have household careers, family career, educational careers, employment careers, and so
forth. Life course studies make the connection between the timing and sequencing of events in one
aspect of people’s lives with their state or events in other aspects. For example, studies have shown
repeatedly the strong association between residential mobility and the timing of marriage, or the
timing of child bearing, etc. The life course perspective has been a prominent theme in studies of
residential mobility and migration. This perspective is usually associated with longitudinal data
analysis due to the emphasis on process rather than association. Early life course research tended to
examine the impact of an event in one career on the likelihood of an event in another career.
Research on life course geographies carries themes of relational ties while contributing new
knowledge on topics that include mobility, work, housing, childhood, changing family
arrangements, social networks, age, generation, disability, health and well-being inequalities and
vulnerability.
An excellent example of life course research is the study by Plane, Henrie and Perry (2005) titled
‘Migration up and down the urban hierarchy and across the life course’. The authors tabulate origindestination specific migration flow data from the Census and Internal Revenue Service tax-return
administrative records for the period 1995-2000. The scale of the study is both national and along
the urban hierarchy. As they examine the recent structure of migration flows up and down the
hierarchy they find support for the concept of step migration as originally presented by Ravenstein
(1885) but instead the flow is down the hierarchy. They find effects of spatially focused
immigration, an age profile to migration, including migration of young adults associated with
19
college and military service, as well as moves characteristic of stages of family formation,
childrearing and retirement. Their research methods include calculating demographic effectiveness,
components of population change, rates of natural increase and differential birth and death rates,
age decomposition of current net-migration streams, and a focus on life course events.
4.4 Intergenerational studies
Now, with the availability of longitudinal data sources and methods, more sophisticated life course
research is able to examine the intergenerational aspects of the life course. Demographers have
always had an interest in families. Life course analysis has progressed to the stage where studies
include more than one generation. There are many intriguing research areas that benefit from
intergenerational analysis. Within geographic research intergenerational studies tend to examine
family proximity and caregiving. Do senior parents live near their adult children? Do grandparents
relocate to be near grandchildren? International variations in intergenerational relationships and
proximity vary enormously. In some nations the distance metric for such studies is measured in
meters whereas in other contexts it is measured in thousands of kilometers. Of note, census 2000
measured the number of grandparents providing regular childcare for their grandchildren.
Intergenerational care relationships have always been of importance, but in the contemporary
context proximity (or more aptly the lack of proximity) creates an additional challenge.
An example of this type of research is provided by Rogerson and Kim (2005) in a paper titled
‘Population distribution and redistribution of the baby-boom cohort in the United States: Recent
trends and implications.’ This study examines the spatial redistribution of the baby-boom cohort.
Since 70 million people were born into this cohort spanning 1946 to 1964, the cohort exhibits
considerable influence on regional population change. The study highlights the temporal and
geographic implications for intergenerational care giving. They focus on the spatial and temporal
dimensions of intergenerational relationships. Their research methods include geovisualization
techniques, and mapping age ratios across counties of the United States.
4.5 Gendered family migration
Within population geography a strong case has been made for the study and understanding of
gendered migration. To the extent that men’s and women’s lives and situations differ, migration is
gendered. These studies draw attention to the social and political context of migrants and migration.
Migrants are situated within the greater context of patriarchy, familial and gendered norms and
expectation, power structures, discrimination and opportunity structures. The impetus for gendered
studies of migration has come from research in developed and developing countries. In the United
States and Britain, for example, family migration has been studied at great length, particularly with
an interest in the impact of migration on the labor-force participation, and earnings potential, of
women and wives. As women’s roles in the labor market have changed (or not) so has the
propensity for families to migrate, given the increased ties to labor markets. Gendered migration
studies advance this field further by acknowledging that men and women have differing
responsibilities in the home, especially regarding caregiving, which articulate into spatial and
temporal variations in labor force attachments amongst married pairs. Population geographers have
made significant advances in this research area. As well, there are geographic differences in the
propensity of families to move, some of which can be explained by spatial variations in the labor
and housing markets, and social networks.
4.6 Assimilation processes
Population assimilation has traditionally been a central focus within the discipline of demography,
and is reviewed elsewhere in this volume. Population geographers contribute to this body of
20
knowledge with an emphasis on the spatial scale of assimilation. A recent example of a the
importance of scale for assimilation studies is the work of Ellis and Wright (2005) titled
‘Assimilation and differences between the settlement patterns of individual immigrants and
immigrant households’. Their research methods include classification, analysis of secondary data
(the Current Population Survey), calculating entropy measures to explore the question of dispersion,
calculating and mapping location quotients and intergenerational conceptions of households. This
study is innovative in the approach to counting immigrants and hence the framing of immigrant
assimilation. Rather than count individual immigrants by neighborhood, metropolitan area, state or
region, these scholars classify immigrants and their descendants into household types. The study
compares distributions of first- and second-generation immigrants with different households that
include first- and second-generation immigrants. Their study shows that the picture of immigrant
geographies is different for households than the geography for immigrant individuals. They draw
attention to scale dependence. They argue that the unit of analysis shapes assimilation research
results. They further argue that the analytic choice is not independent from the politics of
immigration. They intentionally jump scale to compare the sensitivity of their results to scale
dependence.
4.7 Neighborhood studies and residential mobility
Population geographers have studied residential mobility for many decades. Rossi’s 1955 book
titled Why Families Move remains a classic in this regard. For decades the field of residential
mobility was dominated by spatial choice theory and microeconomic behavioral models of rational
behavior. Recently there has been a ‘spatial turn’ within the residential mobility literature, with the
focus turning towards the impact of neighborhoods on peoples daily experiences and opportunity
structure. Residential mobility literature, and place-based studies are interrogating the role the
quality of residential neighborhoods has on peoples life chances. Unfortunately, as explained
previously, many of these studies use unsophisticated concepts of neighborhoods by simply
comparing variation across census tracts. There has been some very creative work that aims to
understand whether racial assimilation achieved by residential mobility is beneficial. The driving
question is whether neighborhoods play a salient role in demographic processes and outcomes. For
example, is being situated in an area of extreme poverty, or an area of hypersegregation, more
explanatory in behavior outcomes than individual measures of the same. We have a long tradition of
measuring explanatory variables at the level of the family or individual, now we paying more
attention to situating these people within different environments. Is there a neighborhood effect?
A novel study in the area is Clark’s (2005) paper titled ‘Intervening in the residential mobility
process: Neighborhood outcomes for low-income populations’. In this study, the US government
sponsored Move-to-Opportunity program provides a real world experimental context to assess the
impact of housing vouchers on families moving out of poverty neighborhoods. His research method
involves empirical analysis of a survey, comparisons across time and space, and geographic
visualization of the neighborhoods in Baltimore. His study considers moves out of poverty and into
integrated communities and how the neighborhood context affects families that live in them.
Collectively these papers address a variety of contemporary themes found in population geography
research. They use a variety of data sources, a host of research approaches and many of the tools of
spatial demography. None of the papers falls neatly into the traditional demographic triad of
population geography. Nonetheless each makes a significant contribution to population geography
and spatial demography. Collectively these themes reveal much of what is currently valued in the
discipline. In the final section of this chapter we turn our attention to the challenges and future
directions of population geography.
5. Challenges and future directions
21
Throughout the first decade of the 21st Century, population geographers have been putting people in
place, temporally and spatially. Population geographers tend to focus on distinct events, such as
births, deaths and movement. Paradoxically, the unintended consequence of focusing on events is
that research seldom addresses the absence of events. For example, residential stability is given far
less attention than residential mobility. In 2005 Susan Hanson addressed this issue in her paper
titled ‘Perspectives on the geographic stability and mobility of people in cities’. As she studies
entrepreneurship and place, she makes an important claim that research on mobility/stability to date
has tended to emphasize mobility and to treat stability as the absence of an event (mobility) rather
than as an occurrence worthy of analysis. People become rooted to place. She draws our attention to
the way in which stability has been suppressed as an analytical category. Population geographers
would be well served to consider whether other analytical categories have been suppressed by our
practices and paradigms.
Not only have some population groups been underresearched by population geographers, certain
issues have not had the benefit of sufficient geographic or demographic inquiry: human security
issues, population-health-environment nexus; population growth, economic development and
poverty issues. Arguably, paths to human-environment sustainability, studies of migration,
circulation, transnationalism, refugee and internally displaced person movements, and of immigrant
communities must be major themes of future empirical research for population geography to count.
Still, some populations are slow to gain recognition in the field – specifically internally displaced
persons and children.
Future progress in the field of population geography will derive from more research at the
intersections of population processes and societal issues and concerns. For example, Hugo (2006)
urges population geographers to step up to the challenge presented by Asia, which he figures must
loom large in any assessment of the dominant global issues – poverty, inequality, sustainable
development, climate change, conflict, spread of disease, famine, and environmental degradation.
The explosion of personal mobility is a key issue in the Asian context. To date population
geographers have already contributed to understanding the complex link between migration and
development, especially in the areas of skilled migration, remittances, diasporas, and return
migration. International migration is related to development and knowledge about the migrationdevelopment nexus has been expanded through analyses of gender. Population geographers need to
give more attention to the interdependencies between places, as well as the interdependencies
amongst scales. Hugo has suggested that in this highly mobile society we need better data on
expatriates. The census of any origin nation should include a question which asks all adult females
on the location of their children, and include a question on residence of all siblings. Information on
these geospatial relationships of children and siblings would make a real contribution to research in
the migration-development nexus.
Acquiring more data does not come without additional concerns, however. VanWey and colleagues
(2005) have addressed this concern in their paper titled ‘Confidentiality and spatially explicit data:
concerns and challenges’. Granted, social-spatial linkage is the key to the advancement of science in
a variety of fields. Recent theoretical, methodological, and technological advances in the spatial
sciences create an opportunity for social sciences to address questions about the reciprocal
relationship between context and individual behavior. There exists an uneasy tension between
respondent confidentially and spatially explicit data. This tension is also present in maps and
displays of spatially referenced data. Collectively, the discipline and supporting agencies are
determining an acceptable disclosure risk level. Issues of confidentiality and spatially explicit data
become more pressing as additional information is gathered and analyzed.
22
An excellent discussion of the future challenges for population geography can be found in Bruce
Newbold’s 2007 book titled Six Billion Plus: World Population in the Twenty-First Century. He
outlines five demographic forces that have the potential to shape the world in the 21 st Century: i)
Population growth and population decline, ii) The HIV/AIDS Epidemic, iii) International
Migration, iv) Refugees and Internally Displaced Persons, and v) Interactions and Conflicts.
Clearly, many of the problems faced by society and in scientific study require consideration of the
complex patterns in interrelated social, behavioral, economic, and environmental phenomena. The
critical issues of the 21st Century, such as economic health, environmental degradation, ethnic
conflict, health care, global climate change, and education, embody fundamental geographic
dimensions. In particular, an area of current and future importance is the geographic dimensions of
vulnerability. These issues also represent opportunities for theoretical and policy relevant research
for geographers. Population scientists are grappling with these issues which derive from
fundamental questions concerning sustainable population change, human security and welfare, and
environmental preservation and conservation. We are moving in the direction of being poised to
contribute as society tackles these critical issues. It remains to be seen whether population
geographers will engage sufficiently with these questions.
Population geography has come of age in the 21st Century. It is no longer a field comprised of
spatial applications of fertility, mortality and migration only. Contemporary population geography
is theoretically sophisticated. Spatial demography has also come of age. It is equally sophisticated
and complex, integrating spatial analysis, geographic information science and geo-referenced data.
An important challenge for population geography is to continue to embrace spatial analysis. Many
of the perceived short-comings of earlier quantitative methods are negligible or meaningless in the
new regime of spatial analysis. As well, spatial demographers need to keep it real. Space is not a
container, and the presence of geocoded information does not make a study inherently spatial.
Population geographers and spatial demographers, alike, need to keep the geographical
epistemology at the forefront.
Bibliography
Anselin, L. (1988). Spatial econometrics: methods and models. Dordrecht; Boston, Kluwer Academic Publishers.
[Classic text on spatial economics]
Anselin L. (1995). Local Indicators of Spatial Association - Lisa. Geographical Analysis 27(2), 93-115. [An
introduction to the method of local indications for spatial association].
Anselin L. (2003). Spatial Externalities. International Regional Science Review. 26(2), 147-52. [Describes the
complexities of incorporating spatial externalities in modeling]
Bailey A. (2005). Making Population Geography. New York: Oxford University Press. [Reviews the development of
population geography in relation to larger disciplinary shifts]
Booth C. (1902). Life and Labour of the People in London (17 Volumes). London: Macmillan and Co. [Provides an
example of early demographic study using maps and spatial distributions]
Boyle P. (2002). Population geography: transnational women on the move. Progress in Human Geography 26, 531-543.
[Provides an assessment of the state of the field of population geography with emphasis on transnational women]
Boyle P. (2003). Population geography: does geography matter in fertility research? Progress in Human Geography 27,
615-626. [Reviews fertility research conducted within population geography]
Boyle P. (2004). Population geography: migration and inequalities in mortality and morbidity. Progress in Human
Geography 28, 767-776. [Reviews mortality research conducted within population geography]
Clark W.A.V. (2005). Intervening in the residential mobility process: Neighborhood outcomes for low-income
populations. The Proceedings of the National Academy of Sciences 102, 15307-15312. [Assesses the merits of the
federal residential mobility program for low-income households]
23
Cliff A.D. and Ord J.K. (1973). Spatial Autocorrelation. New York. Methuen. [A classic text on spatial autocorrelation
issues and processes]
CSISS.
(2008).
Classics
in
Demography.
Center
for
Spatially
Integrated
Social
Sciences.
http://csiss.ncgia.ucsb.edu/GISPopSci/resources/classics/ [This website provides information on the CSISS program].
de Castro M.C. (2007). Spatial Demography: An Opportunity to Improve Policy Making at Diverse Decision Levels.
Population Research and Policy Review 26, 1573-7829. [Reviews papers using spatial analysis to study fertility,
mortality, and population policy]
Ellis J.M. and Wright R. (2005). Assimilation and differences between the settlement patterns of individual immigrants
and immigrant households. The Proceedings of the National Academy of Sciences 102, 15325-15330. [Innovative
measurement of the distribution of immigrants at the household level]
Entwisle B. (2007). Putting People into Place. Demography 44, 687-703. [A call for more theoretical development of
the causal role of space in local area studies]
Fearon D. (2005). Spatial Population Bibliography Report. Center for Spatially Integrated Social Sciences. University
of California, Santa Barbara. [Provides statistics on disciplines and methods of recently published spatial demography
papers]
Forest B. (2005). The changing demographic, legal, and technological context of political representation. The
Proceedings of the National Academy of Sciences 102, 15331-15336. [Outlines demographic challenges for political
representation]
Fotheringham A.S., Brunsdon C. and Charlton M.E. (2000). Quantitative Geography: Perspectives on Spatial Data
Analysis. London: Sage Publications. [An important text for introductory spatial quantitative methods]
Fotheringham A.S., Brunsdon, C. and Charlton, M.E. (2002). Geographically Weighted Regression: The Analysis of
Spatially Varying Relationships. Chichester: Wiley. [Introduces geographically weighted regression]
Gaile G.L. and Willmott C.J. (eds.). (1984). Spatial Statistics and Models. Dordrecht: D. Reidel Publishing Company.
[A collection of spatial methods representative of the 1980s]
Gersmehl P. (2008). Teaching Geography, 2nd Edition. New York: Guilford Publishing. [A primer on the geographic
perspective]
Getis A. and Ord K. J. (1996). Local spatial statistics: an overview. in Longley P. and Batty M.
eds. Spatial analysis: modelling in a GIS environment. New York: Cambridge, 261-277. [A good introductory article
for understanding local spatial statistics]
Goodchild M.F. and Jannell D.G. (2004). Spatially Integrated Social Science. New York: New York. Oxford University
Press. [An overiew of spatially integrated social science by founders of the center at Santa Barbara]
Hagerstrand T. (1970). What about people in regional science? Papers of the Regional Science Association 24, 7-21.
[Classic article on the analytical integration of time and space in regional science]
Hanson S. (2005). Perspectives on the geographic stability and mobility of people in cities. The Proceedings of the
National Academy of Sciences 102, 15301-15306. [Emphasizes stability in geographic population studies]
Hauser P.M. and Duncan O.D. (1959) The Study of Population: An Inventory and Appraisal. The University of Chicago
Press, Chicago. [Presenting a classic and systematic introduction of the primary fields of population studies]
Hoover E.M. and Giarratani, F. (1984). An Introduction to Regional Economics (3rd edition). New York: Alfred Knopf.
[Classic text introducing concentration index]
Hugo G. (2006). Population Geography. Progress in Human Geography 30: 513-523. [Review of state of the art of
population geography]
Hugo G. (2007). Population Geography. Progress in Human Geography 31: 77-88. [Review of state of the art of
population geography]
24
Matthews
S.
(2004).
Population
Research
Institute
at
The
Pennsylvania
http://www.csiss.org/GISPopSci/ [Reference source for classics in spatial demography]
State
University.
Mayhew H. (1861). London Labour and the London Poor, Volumes 1-4. [Early example of mapping and demographic
study]
Meade M. (1977). Medical Geography as Human Ecology: The Dimension of Population Movement. The Geographical
Review 67(4), 379-393.[Foundation article suggests medical-geography as a subfield]
National Academies Press. (2006). Learning to think spatially: GIS as a support system in the K-12 curriculum.
Washington, D.C., National Academies Press. [Ideas from Panel of experts on spatial education]
Newbold K.B. (2007). Six Billion Plus. World Population in the Twenty-First Century. Second Edition. New York:
Rowman & Littlefield Publishers. [Providing an overview of thematic population issues in 21 st Century]
Openshaw S. (1984). The Modifiable Areal Unit Problem. Norwich: Geo Books. [Classic text on MAUP]
Plane D.A. (1987). The geographic components of change in a migration system. Geographical Analysis 19: 283-299.
[Provide details on geographic component methods]
Plane D.A., Henrie C.J. and Perry M.J. (2005). Migration up and down the urban hierarchy and across the life course.
Proceedings of the National Academy of Sciences of the United States of America 102(43), 15313-15318. [Providing a
life-course study of mobility across space]
Plane D.A. and Rogerson P.A. (1994). The Geographical Analysis of Population with Applications to Planning and
Business. New York: John Wiley & Sons, Inc. [Important collection of methods in geographic population analysis]
Ravenstein, E.G. (1985). The laws of migration. Journal of the Royal Statistical Society 48,167-227. [Classical account
of six laws of migration]
Raudenbush S.W. and Bryk A.S. (2002). Hierarchical Linear Models: Applications and Data Analysis Methods. (2nd
Ed). Thousand Oaks: Sage Publications. [Advanced text on hierarchical linear models]
Rogerson PA and Kim D. (2005). Population distribution and redistribution of the baby-boom cohort in the United
States: Recent trends and implications. The Proceedings of the National Academy of Sciences 102, 15319-15324.
[Traces the population movements of the babyboom over time]
Shryock H.S. (1964). The effectiveness of migration. In Population Mobility within the United States. Chicago:
University of Chicago, Community and Family Study Center, 289-294. [Classic text on migration effectiveness]
Snow J. (1849). On the Mode of Communication of Cholera, London. [An early account of mapping and demographic
analysis]
Sweeney S. (2002). Enabling Spatial Demography: Concepts, Tools and Resources. University of Madison, Wisconsin.
[Explains the spatial turn in the social sciences]
Thornthwaite C.W. (1934). Internal Migration in the United States. Philadelphia, PA: University of Pennsylvania Press.
[Classic text is for measuring migration]
Tiefelsdorf M., Seliac M., Bhattacharji S., Friedman G.M. and Neugebauer H.J. (eds.) (2000). Modelling Spatial
Processes: The Identification and Analysis of Spatial Relationships in Regression Residuals by Means of Moran’s I.
Berlin: Springer. [Provides a sophisticated overview of spatial models]
Tobler, W. (1970). A computer movie simulating urban growth in the Detroit region. Economic Geography, 46(2), 234240. [Classic article contains first law of geography]
Trewartha, G.T. (1953). A Case for Population Geography. Annals of the Association of American Geographers 43 (2),
71-97. [Presidential address on population geography to the Association of American Geographers]
VanWey L.K., Rindfuss R.R., Gutmann M.P., Entwisle B. and Balk D.L. (2005). Confidentiality and spatially explicit
data: concerns and challenges. The Proceedings of the National Academy of Sciences 102, 15337-15342. [Reviews the
tensions between data access and privacy]
25
Voss P.R., Curtis White K.J. and Hammer R.B. (2004). The (Re-)Emergence of Spatial Demography. CDE Working
Paper No. 2004-04. Center for Demography and Ecology. University of Wisconsin- Madison. [Provides an overview of
spatial demography in Sociology]
Wachter K.W. (2005). Spatial demography. Proceedings of The National Academy of Sciences of The United States of
America 102, 15299-15300. [Editorial introduction to special issue]
Warntz W. (1964). A new map of the surface of population potentials for the U.S.1960. Geographical Review 54, 170184. [Classic article for measuring population potentials]
Weeks J.R. (2004).What did he know, and when did he know it? Putting Glenn Trewartha's call for population
geography into historical perspective. Forum: fifty years since Trewartha. Population, Space and Place 10 (4), 279283.[Provides the historical context of Trewartha’s presidential address]
Weeks, J. (2004). The role of spatial analysis in demographic reseach. Chapter 19 in Goodchild M.F. and Janelle D.G.
[Editors]. Spatially Integrated Social Science. New York, NY: Oxford University Press. [Describes contribution of
spatial analysis for demographic study]
Westert G.P. and Verhoeff R.N. (1997). Places and People: Multilevel Modelling in Geographical Research. Utrecht:
The Royal Dutch Geographical Society. [A mid-level text on multilevel modelling]
Wilbur G.W. (1963). Migration expectancy in the United States. Journal of the American Statistical Association 58,
444-453. [A classic article on migration expectancies]
Wright J.K. (1936). A Method of Mapping Densities of Population: With Cape Cod as an Example, Geographical
Review 26,103-110. [An early example of mapping in demographic research]
Wright R.A. and Knight P.L. (1988). Dynamic Shift-share Analysis, Growth and Change 19, 1-10. [Providing a
dynamic application of the shift-share method]
Zelinsky W. (1966). A Prologue to Population Geography. Englewood Cliffs, N.J.: Prentice-Hall. [A classic text in
population geography]
26