Issues in Depicting Population Change
with Dot Maps
Eugene Turner and James P. Allen
ABSTRACT: Creating dot maps to show change s in racial and H ispanic popula tion distr ibutions
between two census pe riods can be an effective way to examine o ne of the most important dimensio ns
o f ch an ge with in an y metropolitan area. Using d ots of one color to show population increase and dots
of a second color to show population decrease vividly reveals where ch an ges have occurred within a
la rge r total population. We prepared such maps for t he book Changing Faces, Changing Places: MafJping
Southern Califomians, t he text of which a nalyzes a nd inte r prets the populatio n shifts evident on the
ma ps. ll1e maps show the expansion and contractio n o f racial and H ispanic populat ions in specific
ne ighborhood s so that community leaders and residents a like can easily relate genera l trends to their
localities. In this article we describe th e preparation of these dot maps and ex plain major problem s
encountered in lin king the 1990 and 2000 census population counts at th e tract level. We exp lain
our solutio ns, which we believe made possible mo re accu rate mapping of neighborhood change.
KEYWORDS: Do t mappin g , U.S. census, populatio n chan ge
Introduction
A
the end of each d ecade whe n the U.S.
Census of Popula tion data are released ,
ne of th e first questions people ask
is, "What has changed?" So me people look at
change in terms of counties or metropolitan
units, but ma ny elected officials, scho lars, and
busi ness people a re interested in how neighborhood s have cha nged . By measu ring the counts
of people gained or lost in census tracts between
decen nial cen suses, we have been able to effectively present this through dot mapping. Just
afte r the r-elease of the 1990 census, we created
dot maps that showed, by censu s tract, gains
in popula tion for Los Angeles and Orange
Counties with red dots a nd loss in popula tion
with blue dots (Turner and Allen 199 1). T hat
wor k was updated after Census 2000 into a mor·e
extensive book, Changing Faces, Changing Places
in 2002 (Allen a nd Turner 2002). In this article,
we d escribe the creation of such maps and the
difficulties encountered in trying to resolve the
counts of race and Hispanic population s ti-om
two successive cen suses.
Eugene Turner, Department of Geography, California State
University. Northridge, Northridge, California 91 330-8249. Tel:
(818) 407-0778. Fax: (818) 677-2723. E-mail: < eturner@csun.
edu > . J ames P. Allen. Department of Geography, California
State University, Northridge, North ridge, California 91330-8249.
Tel: (818) 886-8730. Fax: (818) 677-2723. E-mail: < jallen@
csun.edu >.
Representing Change
Using Maps
T he typical representation of cha nge using maps
is to pr·epar·e a series of maps at d iscre te time
per iods-usually on census da tes. Change must
be in ferred from the a ppare nt visual di ffere nces
in the individual pau erns. For example, a se t of
three dot maps portt·ays the distribution of slaves
at three census pe riods in the Historical Atlas ofthe
United States ( a tional Geographic Society 1988, p.
40). Also, in the same atlas, a se t of three isoline
maps re present a p opulation d e nsity surface at
three time period s (p . 42). Such ma ps are helpful in provid ing a broad overview of population
cha nges, but detai ls are more difficu lt to see.
An expression of the magnitude of change is often
presented th rough a chorople th map tha t shows
the percent of increase or d ecrease in population.
For example, in the Census Atlas ofthe United States
counties that gained a pe rcentage of p opula tion
are tilled with different shades of green while counties that lost a percentage of population are filled
with shades of pur·p le (Suchan et al. 2007 , pp .
15, 6 1, 24 1). To show the cha nge in the number·
of persons a graduated symbol map is typically
e mployed. In the Atlas de Fmnciens, var ious maps
containing graduated squares indicate a nume r ical population gai n in an area with a pink square
and a loss with a gr·een square (IAU RIF; INSEE
2000, p p . 4 5, 47, 65).
O ne obvious solution to portraying change
using ma ps is tO create a n a n ima tion. T ha nks to
Cartography and Geographic lnfonnation Science, Vol. 37, No. J, 20 10, pp. 189-197
-..
..
'
LOS
ANGlllS
1\~DINO
Sa
cou
I)
Gab
r i e
T\
TY
M
•.
I
• •
.,..'Chl~<1 .
"Hil~
_'
Perri•. ";:
~t.i ·
;,T\-tn . ,.,
. •~Torq
Major Road
County Boundary
o..____,o'---
20 Km.
11'
,..,
:/
fO-...~
: l
. .,. .\.
.
..
: .. .(
Te mecula •
Figure 1. Portion of map "Black Population Change: 1990 - 2000" from Changing Faces, Changing Places: Mappm,
Southern Californians. A red dot indicates an increase of 100 Blacks and a blue dot indicates a loss of 100 Blacks. Data
are based on census tracts outlined in white.
powe1·fu l ~oftware such as Adobe Fla ~ h . numer ous examples of a n imated map~ Gin be found on
the Internet (Univer ity of Syd ney Archaeological
Computing Laboratory 2004; Eg-vuckwc 20 I0;
Eth ington 200 I; Lodge 20 I0). llowever, lO caplUre detailed spatial chan ge~ over lime 1·equ ire.,
f1·equent data collection that is rarely available.
Funhermore, it is usuall)' ve1-y diflicuh for 1he
' ie\\er to gai n an) thi ng other than the mo'>t general
imp1·ession from ma p ani mations. For '>Cholars
190
and othe r researchers who wish w look closel)
a t patterns of cha nge, an imation does not seem
to be helpful. T hus, presenti ng the ch ange in
co un ~ of race and ll ispa nic populatiom \dth in
ccmus tracts between the 1990 and 2000 censu'
by meam of a series o f s1a1ic dot-maps seemed
especially appro priate.
Dot maps have common ly been u'led 10 G IJ)[Uit'
the spatial d etail of h uman population. The'><'
ha,·e take n a variety of forms I hat range from .1
Cartograph)' and Geographir Informal/on Srimu
basic map where each dot represents a numbe•·
of pe rsons to a mix of graduated symbols a nd
dots (Fawcett 1935; William-Olsson 1963, p . 246).
Cartographers have e mployed the latter method
in an attempt to deal with the large difference s in
population density which exist between urban and
rura l areas whe re a si ngle dot value would either
result in comp lete coalescence of dots in densely
seLtled areas or virtually e mpty .-ural areas.
Dot maps d o present a number of advantages for
portraying distributio ns of discrete phenomena
such as people, animals, or farms. They provide
considerable detail about the nature of the varying density in the underlying distribution and,
because they are d iscrete symbols, they can be
overlaid on other d ata such as shaded terrain,
chorople th pa tterns, o r isolines. They also can
be combined with other dot di stributio ns to portray the associations of the data Uenks 1953). For
example, a map of soil taxonomic orders across
the U.S. illustrates the mixing of ten soil orders,
where each dot represents eight square kilometers
of a soil type (Barrett 2008).
A challenge in d ot mapp ing is to select an appropriate dot size and a dot value since these will affect
the impact of the patterns o n the viewer (Dent
et al. 2009). Mackay ( 1949) and, more recently,
Kimerling (2009) have proposed methods to assist
in deter mining these two factors. Bak et al. (2009)
soug ht to improve the visua lization of dots by
gene rating a cartogram that could improve the
presentation of dots by e nlargi ng the map in very
densely populated areas. We found experimentation
through trial a nd e rror valuable for determining
a dot size a nd value. A dot value of I 00 persons
provided distribution detail across the counties
and enabled quick estimations of p opulations by
simply counting dots. The d ot size of about 0.6
mm was clearly visible over the map and presented
density variation in most a reas of great cha nge.
(Note the clustering of blue dots in Figure l.)
T he other requirement for dot mapping is to place
the dots in a way that captures the actual pattern of
the population being represented. Unfortunately,
the underlying population is rarely spread uniformly
within the a real unit, a nd so the cartographer
will attempt to d etermine how population density
varies within the reporting units. Trad itionally,
sup porting ancillary information such as aerial
pho tographs or topograph ic maps were used to
make the determination, but th is was very time
consuming.
Mapping and G IS software have automated the
placement of dots, but the software randomly places
the dots within the statistical units and creates
11Jl. 3 7, No.3
fairly uniform dot densities within each areal unit.
The resulting pattern often looks unrealistic in less
de nse a reas, and especially in hilly tenain when
relief is shown in the background. 'to improve
automatic dot placement, researchers have suggested strategies such as using smaller areal units
or using GIS to redefine the boundaries of the
geograph ic areas with the aid of satellite imagery,
land use data, or street networks (Mennis 2003;
Holt eta!. 2004; Andresen and Brantingham 2006).
A detailed database covering much of the world,
the LandScan Global Population Database, can
im prove the position of dots in more rural a reas
(Oak Ridge National Laboratory 2003). LandScan
gives an estimate of how many people are expected
to be within a roughly I square kilometer gr id cell
at a ny time of d ay for any day of the year. The
LandScan database is regularly updated, but only
the most recent version is made available a nd it
is not recomme nded for calculating populatio n
change over time d ue to com pilation differences.
'<\There we fou nd automated d ot placeme nt to be
unsatisfactory, we used topograph ic maps, other
large-scale maps, and occasionally ae•·ial photographs to position our d ots.
Representing Population
Change with Dots
We have found d ot maps to be especially useful
for rep resenting metropol itan population
change between census years (Turner and Allen
199 1; Allen a nd Turner 2002). In several maps
of race and Hispanic popu lation cha nge in metropolitan Los Ange les census tracts, we used a
red d ot to indicate an increase of 100 persons
and a blue dot to indicate a loss of I 00 persons
(see Figu•·e I). T he maps vividly capture the
areas of change within the metropoli tan a rea
and are not influenced by the total count of persons. For maps printed in black and white, black
d ots represent population gain and population
loss is ind icated with hollow dots or white dots
against a gray background.
An analysis presented in Changing Faces, Changing
Places of the colored dot map of Black population c hange between 1990 a nd 2000 illustrates
how impo rtant spatia l patterns are revealed (see
Figure I). Perhaps the most str iking pattern on
the map is the decrease in Black po pulations in
the old once-segregated concentrations, shown by
the blue dots. The largest a nd best known of these
former ghettos is South Central L.A. or South
L.A., but the map also shows Black population
/9/
numbers declining in former subu rban ghetto
of Pasade na, Lo ng Beach, Pacoima in the San
Fe rnando Valley, and Monrovia, as well as in cities
beyond Los Angele County.
In all these areas, the decline of Black population numbers ha been due to both a departure
of a growing Black middle-class population and
their replacement by Hispanic, especially Mexican,
immigrants (Allen and Turner 1997). T he departure of middle-class and more affiuent Blacks has
meant that poverty, crime, and poor schools have
come to characterize the former g he ttos. Forty
percent of the remaining Blacks in outh Central
still own their own homes (Myers 2002), but the
lower housing prices in this a nd other poor areas
have attracted Mexicans, with the result that Blacks
comprise only about half the population in areas
shown in blue. Overall, the out-movement of Blacks
from fo rmer gheuos into more White neighbo rhoods has reduced Black-White egregation in
Los An geles County.
T he red dots show tracts in which Blacks have
increased during the 1990s. Blacks moved to
nearby areas west and south of South Central, but
high housing pr ices restricted movement to the
north, and dense Hispanic settlement discouraged
movement directly east. The arrivals of Black were
widespread in more distant areas with moderate
housing prices, particularly in older suburbs, such
as in Bellflowe•; Lakewood, and the an Fernando
Valley. For the San Fernando Valley, a mostly White
area fifty years ago, the new settlement of Blacks
represents a major change. Especially sig nificant
is the fact that Blacks in the Valley are now wide ly
dispersed in contrast to the former g hettos. Also,
Blacks have settled on the urban fri nge in a dispersed pattern: outhern Orange County, the 1-15
freeway corridor in Riverside County, and new
developments in San Bernardino County.
\Nc began the process of creating the metropolitan Los Angeles maps by first e timating the
2000 single-race counts of persons for census
tracts in the 5-county Los Angeles metropolitan
area, a problem that wi ll be addressed later in
this pa per. This was fo llowed by reallocating the
1990 population counts into 2000 census tracts
by aggregating 1990 block data into the 2000
tract boundaries. Change data we•·e generated by
subtracting the 1990 population counts fo r each
tract from those of 2000. Positive and negative
va lues were •·epresented as positive numbers in
separate columns of a spreadsheet and then a dot
map of each category (red dots indicating a gain
and blue d ots indicating a loss) was generated in
ArcMap . The two patterns were then overla id and
192
the composite map exported to Adobe Illustrator.
Because ArcMap randomly places the dots within
the statistical areas, the positions of some individual
dots within larger tracts were adjusted visually from
open to settled an~as by consulting background
map layers of streets and parks and by referencing
the printed Thomas Guide county atlases. While
the above ste ps fa irly describe the production of
the dot maps con iderable data preparatio n was
necessary to make the census tabu lations lium the
two decades comparable. ext we discuss issues
relating to the use and comparability of census
data.
Problems in Dot Mapping
of Population Change
T he U.S. Census is constantly being modified by
changes in the choices and definitio ns of variables, collection procedures, and in geographic
un its (U.S. Census Bu reau 1985; Martinet al.
2002; U.S. Census Bureau 2002; Martin et al.
2005; U.S. Census Bureau 2008b). For mapping
purposes these issues might be categorized as
those related to the collection and reporting of
variables such as race and Hispanic identity and
those related to geography.
Chan ge in Cen us Questions
Regarding Race
Each decade, changes in the definition of variables present challenges to ensuring that measures of change in populations are comparable
(U.S. Census Bureau 2008a). Fortunate ly, the
Census Bureau auempts to document the comparability of different variables to tho e in the
previous census. But even with this documentation not a ll changes can be confidently accounted
for between census decades because the changes
may affect where people were counted, how they
chose their race or I l ispanic identity, o r how
they interpreted the census questions.
TI1e Office of Management and Budget (1997;
2000) has established standards for presenting
race and Hispanic data. Between 1990 and 2000
,the OMB modified the te rms used to describe
SOllie basic race groups. T he term " Black" became
"Black or African Amer ican," "American Indian ,
Ale ut, or Eskimo" became "American Indian and
Alaska Native," and "Asian or Pacific Islander" was
split into "Asian" and " 1 alive Hawaiian and Other
Pacific Islander." The term for Hispan ics in 1990,
" Hispanic origin," became " Hispanic or Latino" in
CartografJhy and Geographic lnfonnatiou Scimce
2000. T he Ce nsus Bureau uses the epa rate terms
"Spa nish, I lispanic, and Latino" intercha ngeably.
In this paper, we use the earlier terms and categories. Generally the Census Bureau defines race
as "self- identification by people according tO the
race or races with which they most closely identify.
T hese categories are socio-political constructs and
sho uld not be interpreted as being scientific or
anthropological in nature. Furthermore , the race
categories include both racia l and national-origin
groups" (U.S. Census Bureau 2003, p. B-38). Fo r
H ispa nics the term "origin" is meant to ind icate
"the heritage, nationality group, lineage, or country
of birth of the person or the perso n's parents o r
ancestors before their ar rival in the United States"
(U .. Census Bureau 2003, p. B- 13).
T he questions and categories used in 1990 a nd
the modified questions and categories used in 2000
and 20 I 0 can be see n by examining the race and
Hispanic questions placed in the respective cen u es.
Because race and Hispanic information is gathered
from two separate questions, the tabulations of race
a nd Hispanic identity are no t mutually exclusive.
For example, a person indicating a Whi te race
may also be Hispanic and will be co unted in the
tabulations for each question. Ho1vever, the Census
Bureau does produce a few race tables (e.g., P8 in
Sum mary File 1 from 2000) where race category
counts are stratified by those persons who did a nd
did not indicate a Hispanic identity. T h us, it is
possible to separate the Hispanic perso ns from
any of the race categories.
The most significa nt change to the race counts
of Ce nsus 2000 was that for the fi rst time, people
were a llowed to mark more than o ne race o n the
ce nsus questionna ire. People could indicate up tO
six races on the fo rm and in the PL94- 17 1 data
file, 6 3 race category combina tions are reported
(U .S. Census Bureau 200 lb). In 1990 and prior
ce nsuses, peop le were forced to choose one race,
and so the data on race of Census 2000 a re no t
directly comparable tO earlier censuses.
In o rder to dete rmine the change in counts of
race groups between 1990 and 2000, one must
decide how to handle the count of the m ixedrace populatio n. Fo rtunately, only 2.4 percent of
the populatio n na tio nally and 4.7 percent of the
population in the Los Angeles 5-cou nty region
reported two o r more races and so the m;Uority of the populatio n still fa lls imo a single-race
only category. everthe less, whethe•· the 2000
single-race only counts are used to calculate the
percem change or the 2000 single and mixed•·ace counts are used ca n make a large difference
in the percem populatio n change reported for
101. 37, No. J
the two census periods. For example, in the worst
case situation, if the 1990 Los Angeles five-county
coum of American Indians, Aleuts, or Eskimos is
co mpared to the 2000 single-race o nly coum of
American India ns and Alaska atives, the area
experienced a 4. 1 percent drop in that po pula tion.
If the 2000 combined single-race a nd mixed -race
American Indian and Alaska atives count is used,
the group experienced a 118.6 percent increase .
A bette •· estimate fo r race coums fo r 2000 that
more closely approxima te the single-•·ace coums
of 1990 can be achieved by reallocating the mixedrace responses to single-race categories through a
process called "bridging" (Office of Management
and Budget 2000; National Ce nte r fo•· Health
Statistics 2009). l11e OM B reviewed severa l methods
of assigning the mixed -race counts tO single-race
counts. Generally, methods that fractionally assigned
the mixed-race counts to the groups which were
part of the mixed-race category were prefer red,
though no single bridg ing method was selected .
For Changing Faces, Changing Places we developed
a fractio nal assignment method by co mparing
ancestry responses to the race respo nses in 1990,
using the U.S. Census Public-Use Microdata Sample
(PUM ) fi le for California (Allen and Turner 200 I).
By aggre gating ancestr ies tha t clearly indicated
White , Black, Asia n a nd Pacifi c lslande1; and
American Ind ian, Aleut, and Eskimo races, we
de tenn ined for 1990 those persons who might
have indicated a mixed-race response had the
questionnaire been like that used in 2000. T hat is,
someone who indicated a White race and a J apanese
or Chinese a ncestry could be co nside red a mix of
Asia n a nd Wh ite races. This research suggested
that 67 . I percent of the California Black- Whi te
popula tion identifi ed the ir race as Black and that
32.8 percent of the Californ ia Asian- White population ide ntified their race as Asia n and Pacific
Islander. We used these figures for apportioning
those two mixed-race counts for the Census 2000
data, but for other race co mbinations we assig ned
mixed-race counts in equal fractions to each of
the grou ps that comprised the ir category.
After assigning the mixed-race populations to
the race categories, we reduced d ouble-counting
of persons in race and I Iispanic tOtal counts. To
d o this we chose to count non-H ispanic Whites
and Non-Hispan ic Blacks as the significant race
categories, thus leaving H ispanic Whites and
Hispanic Blacks to be counted as H ispanics. l11is
was because we conside red the H ispanic identity
to be the more important one for people who also
identified racia lly as e ither Wh ite o r Black. In
contrast, we conside red the Hispan ic identity to
/93
1
be secondary among people whose t·ace identity
was American Ind ian, Aleut, or Eskimo ot· Asian
and Pacific Islanders. This meant that Hispanic
members of those group s (primarily Filipinos
a nd Guamanians) were included in the counts of
these race groups but were subtracted from the
Hispanic total count.
T he Census Bureau publishes in its Summary File
I even more detailed race counts, but mixed-race
counts of the various categor ies are not g iven for
these mot·e detailed groups. For this reason, population change maps of specific races like Ch inese,
J apanese, Filipinos, a nd Koreans in Changing Faces,
Changing Places used the si ngle-race counts from
Census 2000, t·esulting in a slight underrepresentation on our maps.
Ch ange in Census Q uesti on s Regardin g
Hispa nic Ide ntity
Sometimes subtle changes in the wording of
questions can cause a change in the response
r ate and e ither an increase or decrease in the
numbers of counted persons. For exa mple, in
Census 2000 the question on Hispanic origin
was cha nged so that a lengthy list of countries of origin did not appear a fter the Othe r
Hispanic write-i n question. This apparently was
the cause of a large drop in the reported counts
of som e Hispanic groups such as Dominicans,
Salvadorans, a nd Ecuadorans, though, the total
count of Hispanics was somewhat higher than
expected possibly due to a better count in 2000
(Suro 2002; Ramirez 2008). For this reason,
maps showing popu lation change for Hispanic
subgroups were not included in Changing Faces,
Changing Places.
Ch an ges in Tract Boundaries
a nd Group Q uarter s Locatio ns
At the metropoli tan level, the census tract serves
as a useful geographic unit for analyzing and
displaying census data. However, between census
dates, the boundaries of many tracts have been
modified a nd, as a result, the popu lation counts
for the non-match ing tracts from one decade
must be reallocated to match boundar ies of the
second decade (Holt et a l. 2004). Particularly
cha llenging for establishing common census
tract areas for Changing Faces, Changing Places
wet·e situations whet·e the 1990 censu s block
boundar ies crossed Census 2000 tract boundaries. This typically occurred in new suburbs
constructed in the 1990s. Prior to 2000 th is
194
was ope n counu·y without the p hysical boundaries norma lly requit·ed by the Census But·eau
to defin e blocks, but with large-scale development new blocks a nd tracts had to be created. In
those cases, the street and park overlays used in
ArcMap and the printed Thomas Guide maps
provided information as to whether or not there
was settlement in those areas.
One unexpected outcome of mapping population
change data with dots in Changing Faces, Changing
Places was the t·evelation of shifts of some group
quarters populations (e.g. no n-housing units such
as p risons, hospitals, dormitories, she lters, etc.)
between adjacent tracts between I 990 a nd 2000 .
This was suggested by the presence of a census
tract fill ed with red dots adjacent to a census tract
filled with blue dots. In most cases these occurrences happe ned whe re an j nstitution such as a
college campus or prison was present and seems
re lated to the way census maili ng addresses were
allocated to blocks. The most notable example
occurred in Los Angeles County in tract 9202 near
th e community of Castaic where the count of 7 11 3
prisoners in the Peter Pitchess Deten tion Center
(the entit·e u·act population) was shifted into a
residential tract to the west. Other shifts occurred
at tract 5500 in the community of Norwalk where
the count of 9 18 patients of the Metropolitan
State Hospital was moved; at tract 2227 where the
count of 2500 college students in dormitories at
the University of Southern California was moved;
and at tract 4024.04 whe re the count of 1200
students attending California State Polytechnic
University, Pomona, was moved. Another e r ror
in tract 1152.02 in 1990 misplaced the coun t of
1400 students attending California State University,
Northridge. It is li kely the students used a campus
ma iling address, which placed the m in a tract
containing no do rmito ries.
Once we discovered a few e r rot·s of this type, we
used table PCTL6, Group Quarters Population by
Group Quarters Type, from California Summary
Fi le l to locate al l tracts with large group quarters
populations and examined them for gains a nd
losses of population between adjacent tracts (U.S.
Census Bureau 200 Ia ). We shifi.ed the group quarters populations of these tracts back to the tract
that contained the institution, which in neady all
cases was the 1990 tract. Visually, the corrections
removed isolated population losses ot· gains within
more general patterns of gain or loss.
Such errors may be common nationally and any
map showing population change at the u·act level
should be checked in areas where group quar ters
are present (Scott 200 I). These shifts a re a real
Cartogra.phy and Geogra.phic l nfonnation Science
problem for those using the census for a statistical
analysi intended to measure change over time
as the abrupt shift in tract population can lead a
researcher to note significant changes when, in
fact, little has occurred. T he problem is part of
the larger issue of defining residence especially for
those in group quarters that are counted as part
of a local population ( ational Research Council
2006). uch u·acts tend to presem atyp ical socioeconomic data, and it may be useful to subtract
them from the total tract po pulations when they
repre ent a large pro portion of the total.
Census 2010
The race and Hispanic counts in Census 2010
essemially will be like those in 2000 so that
measures of change can be mad e conlidemly
for the questions that are mailed to all households on the "short form" questionnaire (U.S.
Census Bureau 2008b). One small change that
will occur is that the Other Asian, Other Pacific
Islander, and Other Hispanic write-in categories included a list of example answers similar
to those presented in 1990. The lack of those
examples in 2000 was thought to be partly
responsible f01· a reduction in write-in responses
previously recorded in 1990, and their inclusion
in 20 I 0 likely could be a factor in increasing the
reported coun ts of the write-in tabulations for
the Iis ted groups. Thus, any calculation of population count differences between 2000 and 20 I 0
for the groups listed under the "Other" categories may be inflated in 20 I0.
There is a major change for some variables such
as ancestry, country of birth, or language spoken at
home which were previously collected on a more
detailed "long-form" questionnaire sent to about
one in six household s. The sample questionnaire
will not be distributed at the 20 I0 decennial census;
it has been replaced by an annual survey of 3 mi llion addresses called the Amer ican Community
Survey. Data will be released annually for geographic
areas containi ng over 65,000 persons. T hey will
be released as a three-year average for geographic
areas containing over 20,000 persons and as a
live-year average for areas containing fewer than
20,000 persons. So far, one and three-year counts
have become avai lable, with the live-yem· counts
scheduled to begin release in 20 10.
Although the American Community SUI-vey will
provide popu lation and socioeconomic variables
similar to those of Census 2000, the estimates are
from a smaller sample and will can-y a greater
1-01.37, No.3
sam pling err01; especially fo r small-area tabu lations
such as census tracts ( a tiona) Research Council
2007; U.S. Census Bureau 2009). Other differences
include the faCLthatthe live-year averages reported
for census u·acts represent d ata collected over a
live-year per iod rather than during a brief census
data collection period. T herefore, a longer samp ling time frame is involved . The Census Bureau
suggests using the fiv e-year data to represent
change, and that mixing of data from one, three
and live-year samp les be avoided. Even though
five-year average d ata wi ll be released annuall)•,
measures of change should be reported at no less
than five-year intervals tO avo id using overlapping
years in adjacent time periods, though, statistically, some overlap can be accommodated. Most
reported values include a 90 percent sampling
error or margin of eJTor which will be large•· for
small-area geographic units such as census tracts.
These a ppear in tables along with an e timate
of the data value. Even less precision ex ists for
smaller ubgroups of populations within census
tracts (Wombold 2008a; Wombold 2008b). T hus,
for smaller geographic unit such as censu tracts it
may be reasonable to produce inset maps d isp laying the error associated with census tracts along
with the map of the census variable.
Conclusion
Dot maps have proved a most effective app roach
for capturing detailed va•·iations in population
d ensity. Th is same advantage applies when they
are used to capture change data by means of dots
of two different colors indicating a gain or a loss
in the population. Though, the approach is relatively rare, recent examples of its use have been
presented in East-West Gateway Coordinating
Council (2002), Center fo r Urban and Regional
Affairs (2003), and U.S. De partmen t of
Agriculwre (2007). Dot maps showing change
allow one to see in greater detail the size, location , and trend of changes occur r ing within a
large population. They may also be an effective
visualization tool for •·esearchers analyzing population change within metropolitan areas since
dot maps can suggest where erro•·s in the data
have occurred.
One of the challenges of mapping change is
accounting for d ifferences in the methods of d ata
collection and reporting which have occurred
between census dates. The cartographer needs to
check census reports for modifications that m ight
affect the data of in terest and to be suspicious
195
of u nexpected or unusual changes presented by
the maps. However, we have found that mapping
population change in this way is effective for getting
a feel for what is happening in different parts of
a city and its suburbs, a nd it d oe s provide a view
of population data , which enables greater understanding of metro po litan areas. Mapping change
by d ifferent colored dots as we have illustrated is
easily understood, in add ition to eiTectively portraying trends that are important in the lives of
local people.
REFE RE NCES
Allen, .J.P., and E. 1i trner. 1997. The ethnic quilt:
Population diversity in southem Califomia. Center for
Geographical Studies, Northridge, California.
Allen , J.P., and E. Turner. 200 1. Bridging 1990 and
2000 census race data: Fractional assignment of
multiracial population s. Population Rtsearr;h ami
lblicy Review 20: 5 13-33.
Alle n,J .P., and E. Turner. 2002. Clumgingfaces, clwnging
!>laces: Mapping southem Califomians. Center for
Geographical Studies, Nortlu·idge, California.
Andresen, M.A., and P.L. Brantingham. 2006.
Visualizing ambient population data '"ithin census
boundaries: A dasymetric mapping p rocedu re.
Cartographica 43(4): 267-75.
13ak, P. , M. Schaefer, A. Stoffel, D.A. Keim, a nd I.
Omer. 2009 . Density equa lizing distortion o f large
geographic point sets. Cartography mul Geographic
lnfomwlion Science 36(3): 237-50.
13ar rett, L. 2008. Soil taxonomic order. [h ltp://blogs.
esri.com/Su p p or llphotos/ ma pping_center _ may_
2008/ images/632/o •·iginal.aspx; accessed November
15, 2009.].
Center for Urban and Regional Affairs (CU RA).
2003. Population change 1990-2000: Minnesota.
[http://www. u tra. umn.ed u/reporter/03-Summ/ PopChange-Maps.pdf; accessed November 15, 2009.].
Dent, B., J .S. Torguson , and T.W. Hodler. 2009.
Cartography: Thematic lflitp design, 6th ed. 1\ew York,
New York: McCraw-ll ill.
Eas1-West Gateway Coordinatin g Cou ncil. 2002.
Geographic profile of p opulation change: 19902000.St. Louis, Missouri. [hup://www.ewgateway.org/
pdffi les/ libra ry/d atacen ter/geop rolileOrPopChg2002.pdf; accessed ovember 15, 2009.).
Egwuekwe, L. 20 I 0. ·n,e decline: T he geogra phy of a
recession: Unemployment rates by coun ty. [http:l/
coho rt l l .amcricanobserver. net/latoyaegwuekwe/
multimediafinal.html; accessed Fe bruary 28, 2010J.
Ethington , P. 200 I. Census 2000-race contours,
Los Angeles County. animated racial d istribution s
by census tracts, 1940-1 990. [hup:/l"'''w. usc.ed u/
sch ools/spp d/rcscarch/ census20 00/race_census/
racecontours/la.htm ; accessed March 4, 20 10].
Fawccu , C.B. 1935. Population map s. Thf Gfogmphical
.Joum al 85(2): 142-6.
196
Holt, j. 13., C.P. Lo., a nd T W. Hodler. 2004.
Dasymeu·ic estinwtio n of pop ulation d en sity a nd
aerial interpolation of cen sus data. Cartography and
Geographic lnfomwtion Scie~~u 3 1(2): 103-2 1.
lnstitut d' i\.menegeme nt et d' Urban isme de Ia Regio n
d ' lle-d e-France: lnstilllt natio nal de Ia stat istique
et des etudes economiques (IAU RIF; I SEE).
2000. Atlas de> Fn mciliens: To me I: Territoi•·e et
Population. P:tris: IAU RIF; INSEE.
J enks, G.F. 1953. "l'oimillism " as a cartographic
technique . The J>roj~ssionfli Geographer 5(5): 4-8.
Kimerling, A.J. 2009. Do tti ng the dot map. revisited.
Cal1ography and Geographic lnfomwtion Somce 36(2):
165-82.
Lodge, F. E. 20 I0 . Atlas of world histOry. [hnp:h'-"''"·
a t Iasofwo rld h isto •·y.com/ E N G LIS H/ B2 5 50. hIm I;
accessed Februa•)' 28, 20 I0.].
Mackay, JR. 1949. Dotting the dot map: An analysis of
d ot size, num ber and visual tOne density. Su111eying
and Mapping 9( 1): 3-10.
Martin, D .• D. Dorling, and R. Mitchell. 2002. Linking
censuses through time: Problems and solutio ns. Area
34( I): 82-9 1.
Martin , E., M. de Ia Puente, and C. 13ennclt. 2005 . The
effects of questionnaire and content change on race
data: Results o f a re plication of 1990 race and origin
questions. Research Report Series Survey Methodology
#2005-05. \ V..shin!,'ton: U. S. Census 13urcau. [hupJ/
"'"v.census.gov/srdlp apers/pdUrsm2005-05.pdf].
Mennis,J . 2003. General ing surface models of population
using d asymetric ma pping . The J>rofessumal GeogmjJher
55( 1): 3 1-42.
Myers, D. 2002 . Demographic and housing transitions in
South Central Los Angeles, 1990-2000. Special Report ,
Population Dynamics Research Gmup, University of
Sonthem Califomia. (hupJ/""w.usc.edu/schoolslsppd!
resc;•rch/census2000!].
National Center for Health Statistics. 2009. l'ostcensal
estimates of the resident population of the United
tates for July I, 2000:J uly I, 2008, by year, coun ty, age,
bridged race, Hispanic origin, and sex (Vin tage 2008).
["'"v.cdc.gov/nchs/nvsVbridged_race.htm; accessed
November 9, 2009).
ational Geog•·aphic Society (U.S.). 1988. Historical
atlas of the United Slates. Wash ington, D.C.: Nation al
Geograph ic Society.
National Research Council. 2006. Once, only once, an d
in the right p lace: Residence ru les in th e decen nial
census. In: D.L. Cork and P.R. Voss (cds), ftmel on
rtsidence mles in tlu decamial census. Washin gton,
D.C.: National Acadamies Press. [http://books. nap.
edu/o pe nboo k.php?record _id = 11 727&page= RI ;
accessed FebrWII)' 25, 20 I 0].
National Research Council. 2007. Using the American
Community Survey: & nefits and challenges. In: C.F.
Citro and G. Kelton (eels.), ftuul 011 th~ fimctionality
and usability of dala from the American comm1mity SIIYVe)'·
\\f,JShington, D.C.: l11e National Academies Press.
[http:h\"""·nap.edu/catalog.php?record_id= 11 90 I ;
accessed Februat)' 28, 20 I 0).
Cartography 011d Geographic Information Science
Oak Ridge ational Laboratory. 2003. Land scan
global population database. Oak Ridge, Tennessee:
Oak Ridge National Laboratory. [http:/Amw.ornl.
gov/sci/landsc.~n/; accessed March 4, 201 0].
Budget.
1997.
Office of Management and
Recommendations from the Interagency Committee
for the review of the racial and ethnic standards to the
OfficeofManagementand Budget concerning changes
to the standards for the classification of federal data
on race and ethnicity. Federal RPgi.ster 7/9(97, Pan II):
36873-946. [httpJAvww.whitehOitse.gov/omb/fedreg_
directive_15/; accessed February 26, 20 I0].
Office of Management and Budget. 2000. Pnroisionnl
guidana on the implementation of the 1997 standanis for
frdtral tlnta 011 raa rmd etlmidJy. Washington, D.C.:
Prepared by theT:~bula tion Working Group Interagency
Committee for the Review of Standards for Data on
Race and Ethnicity. [http:/AvMv.whitehouse.gov/omb/
inforeglr&e_guidance2000upclate.pclf;
accessed
Feb11.1a1)' 25, 20 l 0].
Ramirez, R. 2008. Analysis of multiple origin
reporting to the Hispanic origin question in Census
2000. Working paper 'o. 77. . . Census Bureau,
v\'.tshington, D.C.
colt, J. 200 I. Census said to misplace many prisons
and donns. New lbrk Times, November 28. [httpJA''" "·
nytimes.com/200 l/ l l/28/us/census-saicl-to-misplacemany- pr isons-and-dor ms.hun l?pagewanted = I;
acce sed Feb11.1a1)' 25, 20 I0].
uchan, TA., MJ. Pen)', J. D. l'itzsimmons, A. E. J uhn,
A.M . l a it, and CA. Brewe.: 2007. Cemus atlas of the
United Stales, series CEMSR-29. U.S. Census Bureau,
Washington, D.C.
Suro, R. 2002. Counting the "other Hispanics." How
many Colombians,
Dominicans, Ecuadorians,
Guatemalans and Salvadorans are there in the United
States? Pew l lispanic center study. (http://pewhispanic.
orglliles/re ports/8.pdf; accessed November 12, 2009].
Turner, E., and .J. I~ Allen. 199 1. All atlas ofjJopulation
jJatlems in metropolitan Los Angell's Cmmty 1990.
Center for Geographical Studies,
Northridge,
California.
U.S. Census Bureau. 1985. Evaluating censuses
of population and housing, Statistical Training
Document, ISP-TR-5. \\'.tshington, D.C. [hup :/A,~vw.
census.gov/srd/paper pdf/rr85-24 .pdf].
U.S. Census Bureau. 200 I a. Census 2000 summary l~ l e
I Cali forni a[ mach ine-readabledata files]/pre pared
by the U.S. Census Bureau. [hup://faCLfinder.
ce ns us.gov/se rv le t/ Dataset Ma i n PageSe r vlet ?_
prog ra m = DEC&_subm e nu l d = data se ts_ I &_
lang=en; accessed Februa•)' 25, 20 I0].
11ll. 37, No. 3
U.S. Census Bureau. 200 I b. Census 2000 redistricting
data. (Public Law 94-17 1) Summary File- Technical
Documenttion/ Prepared by the U.S. Census Bureau.
[lmp :/A'~'w.ce nsu s .gov/p rod/ccn 2000/dodpl 94-l 7 1.
pelf; accessed July 13, 20 I0].
U.. Census Bureau. 2002. Measuring America: The
decennial censuses from 1790 to 2000. [http://"""'·
census.gov/ procl/2002pubslpol02marv.pdf; accessed
March 4, 20 10].
U.S. Census Bureau. 2003. Seierted appmdi.xes: 2000:
Swmnmy social, economic, and hou.sing characteristics.
PHC-2-A . Washington, D.C.: U.S.Governme nt
Printing Office. [http:/A"""·census.gov/ population/
cen2000/ phc-2-a-B.pdf; accessed ~l a •-ch 6, 20 l 0].
U.S. Census Bureau. 2008a. Mi\ior diffe rences in
subject-matter conte nt between the 1990 and
2000 census questionnaires. [http://"'' ""·census.
gov/popula tionh""'''/cen 2000/90vs00/index. htm I;
accessed March 4, 20 I 0].
U.S. Census Bureau. 2008b. Q uestions planned for
the 20 I0 census and Ameri can Communicy Swvey.
[hupJA,ww.census.gov/acsA"""''/Downloads/Questions_
Planned_for_the_20 I 0_ Census_and_Amer ican
Communicy_Sun·ey.pdf; accessed March 4, 20 10].
U.. Census Bureau. 2009. A comj){ISS for understrmding
and using American Community urvey data: What
researchers need to know. Washington, D.C.:
U.S.Governme nt Printing Office. (http:/A"""·
ce n sus .gov/a cs/'"''~v/Dow iii O<ld s/ACS Research. pelf;
accessed February 27, 20 10].
U. . De partment of Agriculture. 2007. T he
census of agriculture. [wMv.agcensus.usda.gov/
Publications/2007/0nl ine_ l l ig hligh ts/Ag_ Atlas_
Ma ps/index.asp; accessed ovember 12, 2009].
University of Sydney Ar-cha eological Comput ing
l .aboralOI)'. 2004. l l meMa p map animations.
(hu p ://mvw.time map.netlindex.php?option =corn_
co nte nt &ta s k= view& icl = 124& 1temid= 130 ;
accessed ~ebma•)' 28, 20 I 0].
William-Oisson, W. 1963. TI1e commission on a
world population map: HistOI)', activities, and
recommendations. Geografiska Annaler 15(4): 243-50.
Wombolcl, L. 2008a. Sample sit:e matters: Caveats for
users of ACS tabulations. ArcUser 8( I ): 9- 1 I. [http://
m"v.esri.com/library/reprint · pd~ arcuser_samplesize.pdf; accessed Ma1-ch 8, 20 I 0].
\Vombold, L. 2008b. Examining er mr: Considering
the effect of sample siLe and ermr source when
using census data. ArcUser 8(2): 8-11. [hup://wMv.
esri .com/ news/arc user/0 70 8/clc moa rticl e . h t m I;
accessed March 8, 20 I 0].
•
197