SPATTAL CHANGE TN POST -WAR
SOUTHERN REPUBLICAN VOTING
RESPONSES

Dissertation for the Degree of Ph. D.
MICHIGAN STATE UNIVERSITY
GERALD LYNN TNGALIS
1 9 7 3

.v

nuwwnmnm

01087 7

 

 

 

 

 

 

ill '

16

l

 

This is to certify that the
thesis entitled
SPATIAL CHANGE IN POST-WAR SOUTHERN
REPUBLICAN VOTING RESPONSES

presented by

Gerald Lynn Ingalls

has been accepted towards fulfillment
of the requirements for

LIL...— degree in Meography

mnemfl‘dmf‘
LI}; 'i 3 ' Y

Michigm :; Late

University

 

Date Segtember 25: 1.973

  
    

 

 

 

_ , I- 7 fi—w—r—w—W ,, , ,

oz: 1 2 T996

‘LUU'Z

  
   
  
   
  
 
 
  
 
   
 
  

:3 4” 9.9%.? ' t .
'0.” race:
I ~_. This Luvm
‘ h't 91'0“}: g:
. DC “8 cc-ux
' u giver . ». >
IOU-n; 1.25;.1. 3
In»: Luau ‘._« _
Of lift»..- ',I‘..._»v , -. x.‘ l '

““9 “t of mlr;l’t",vyrl Car. ,1 v.2). _ 1
” thattfzvo by ‘M mm at rug. - . fag}.

‘ t “was. The: .Sttict.‘;tlun --4 .- .-.-
' .m 8W6 amuse-s tum .,- . t -- - _
.. M can». on Um ”153;! J. :2

a. _
”( fizg‘wmcis of you ’ n6 n; "h - -

 

 

 

ABSTRACT
SPATIAL CHANGE IN POST-WAR SOUTHERN
REPUBLICAN VOTING RESPONSES
BY

Gerald Lynn Ingalls

The Republican party, long an electoral anathema in the American
South, has recently begun to demonstrate a measure of competitive
potency. This inquiry focuses on the electoral change associated with
the recent growth of Republican electoral support in a contiguous
sample of 448 counties drawn from seven southern states. Particular
attention is given to the identification of spatial regularity in
Republican voting responses at the presidential and senatorial levels of
competition from 1948 to 1972 and to the identification of changes in
the patterns of these responses. Spatial regularity of voting response
is identified by computer techniques which describe the shape and
density of a set of weighted data points. Changes in these voting
responses are identified by the use of ring and sector analysis of the
weighted data values. The description of the spatial structure of the
Republican voting response surfaces forms the basis for the primary
examination which centers on the thesis of an urban-centered electoral
change. The hypothesis of positive relationship between urban size

and level of Republican electoral support is examined.

 

 

Gerald Lynn Ingalls

It appears that electoral change, in the form of increased
levels of Republican electoral support,has produced patterns of elec—
toral response quite unlike those of the years preceding 1948. Elec-
toral support has moved from a spatial concentration in historical
"cores" of traditional Republican support to a distribution of support
that approaches spatial uniformity. This change is not constant
across varying levels of electoral competition, however, since Repub-
lican presidential candidates demonstrate higher levels and more uni-
form patterns of support than senatorial candidates. Thus a strong
evidence of time lag in the growth of support can be identified between
these two levels of competition.

The expected positive association between population size or
urban structure andtfin level of Republican support is verified. The
pattern of this positive association is, however, strongest in earlier
elections (1948-1960) and begins to weaken in later years (1964-1972).
In later years a marked tendency toward decreasing levels of support
at higher echelons of population size is also notable. The coincidence
of more uniform patterns of electoral support, higher levels of sup-
port, and a weakening association between urban size and level of vote
leads to speculation that the grassroots organizational activity of
the Goldwater years are perhaps bearing fruit. And although early
electoral change was closely tied to higher levels of urban structure,
the later growth has achieved a measure of uniformity suggestive of

approaching competitiveness.

 

SPATIAL CHANGE IN POST-WAR SOUTHERN

REPUBLICAN VOTING RESPONSES
BY

Gerald Lynn Ingalls

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR or mrwsom
Depertnent of Geography

1973

’r‘

 

   

x "‘
3 .-
r,
it}
| .
4-4,
‘4 it
”t. 1,33. .

   
  
  
  

M 2!: lh',‘ ' 'f’u‘3-'
Mt has i. - ..
i". ' '
d“. Md LAW: 2'
’ use“ ch Luv r
“Jitter. .‘n It'..;;,’i ‘
fish '

m u]. 'tfi‘:' j
.bhll and tr.

.- W and on m.‘ = To Joy“, who believed
(i Omit-tin 215.,
H‘ “in Chute an: n. )r‘.‘

i »- ‘- Ihe nosey-h gun"

to: finding the ‘Au m... ; ; 3,
pita?» womanlcdqe the might ~.
;ww WOI‘ Center. Hug-
m Import in part. .‘~r“

1‘“ to am the sue: a u.» A.»

&.

.9..1

Mat-

sun ‘

 

ACKNOWLEDGMENTS

 

As in any research effort, assistance in creating the final
product has come from many different sources. For sound ideas, ad-
vice, and criticism, I am indebted in an immediate sense to my
committee chairman, Dr. Stanley Brunn. His patient and helpful dis-
position in reading often unintelligible drafts is greatly appreciated.
And to the remainder of my committee, Dr. Lawrence Sommers, Dr. Robert
Thomas and Dr. Paul Conn, I am especially grateful for the advice,
guidance and encouragement they have provided.

Particular thanks must go to the organizations and individuals
within those organizations who provided assistance at crucial stages
in the research project. One such group is the Inter-University
Consortium for Political Research in Ann Arbor, Michigan, from which
came a large portion of the data used in this analysis. And to people
such as Janet Vavra of the Historical Archives Section of ISR, Janet
Huffman of the Republican National Committee, and Mr. S. L. Kopald and
Ron Rietdorf of the Tennessee Republican party, I wish to express
gratitude for finding the data that others said were unavailable.

I wish to acknowledge the use of the facilities of Michigan
State University Computer Center. Use of these facilities was made
possible through support, in part, from the National Science Founda-

tion. I wish to thank the staff of the Computer Institute for Social

iii

   
 
  
    
   
 
 
  

Finally, but by no means last, comes my family. To Rett the

“I
v| _
.‘ incentive and to Joyce the typist, editor, reviewer, critic and general

’ all-around sounding board, I only hope I can make amends. Your patience
and confidence have not always been verbally acknowledged, but I

'aesure they have been recognized.

All of these people and groups have given invaluable support in the
the bringing to fruition this research effort and I am grateful.

But the ultimate responsibility for error and omission does not rest
with them. Ultimate responsibility for the outcome rests with the

author alone .

 

CHAPTER

 

I

TABLE OF CONTENTS

A CONCEPTUAL FOUNDATION FOR THE RESEARCH PROBLEM

Theoretical Developments in voting Behavior
Research . . . . . . . . . . . . . . . . . . . .

The Geographer' 5 Place in the Study of Electoral
Behavior . . . . . . . . . . . . . . . . . . .

Electoral Change As a Conceptual Framework . .
The South As a Regional Laboratory . . . . . .
The Problem . . . . . . . . . . . . . . . . . .
Specific Statement of the Problem . . . . . .

Elaboration on the Problem . . . . . . . . . .

A LITERATURE FOUNDATION FOR THE RESEARCH PROBLEM

The South in voting Literature . . . . . . . . .
The Question of Electoral Change in the South
The Components of Southern Electoral Change .

The Problem in the General Theoretical Literature
The Spatial Approach to Voting Analysis . . .
Urban size and Electoral Change . . . . . . .

thectives and Hypotheses Derived from the

I . mt.ratw. o a o I e o o e a o e o a e o o I a e

V

Page

viii

ix

12
12

13

14
15
15
23
27

27

 

'5 ."

   
  
  
  

t ,1
,‘o
A!

l.
f,

a;
K N
aficbfi'rm s (Cont'd.)
in , pass
”01:11 me APPROACH TO THE RESEARCH pRosLsM . . . . . . . . . . 44
Aggregate Analysis in Electoral Research . . . . . . . . 44
The Scope of the Problem . . . . . . . . . . . . . . . . 47
The Spatial Context . . . . . . . . . . . . . . . . . 48
Units of analysis . . . . . . . . . . . . . . . . . 48
Delimitation of the study area . . . . . . . . . . . 49
The Electoral Context . . . . . . . . . . . . . . . . 51
The levels of analysis . . . . . . . . . . . . . . . 53
The longitudinal extent of the analysis . . . . . . 56
The Data . . . . . . . . . . . . . . . . . . . . . . . 59
Data sources . . . . . . . . . . . . . . . . . . . . 59
Characteristics of the data . . . . . . . . . . . . 60
The Method of Analysis . . . . . . . . . . . . . . . . . 61
Analysis of Structure . . . . . . . . . . . . . . . . 61
Numerical description of the data set . . . . . . . 62

Numerical description of the spatial characteristics
of the data set . . . . . . . . . . . . . . . . . . 66

Analysis of Process . . . . . . . . . . . . . . . . . 73

ADESCRIPTION 0F STRUCTURE . . . . . . . . . . . . . . . . 79
‘General Patterns of Electoral Response . . . . . . . . . 80
Electoral Parameters . . . . . . . . . . . . . . . . . 81 ‘szi ii
Numerical means . . . . . . . . . . . . . . . . . . 81 A,
Standard deviations . . . . . . . . . . . . . . . . 84

The hypotheses in light of the results . . . . . . . 87

0". V1 '

 

'3' (Cont'd.)

   
 
   
  
  
  
  
  
  
  
  
  

Spatial Parameters . . . . . . . . . . .
Shape of the distribution . . . . . .

Density of the distribution . . . . .

Subregional Patterns of Electoral Response

DESCRIPTION OF THE PROCESS . . . . . . . .

General Patterns of Relationship . . . . .
1 Maps and Ring and Sector Analysis . . .
Rank Order Correlations . . . . . . . .
Specific Patterns of Relationship . . . .

Size of Place Analysis . . . . . . . . .

A PERSPECTIVE OF THE RESULTS . . . . . . .
‘CThe-Results in a Regional Context . . . .

-The Results in a Theoretical Context . . .

“mm. 0'. as a a a so a. as e o o o n

- .Nurs-z at
The kfiia‘.. '

vii

The hypotheses in light of the results .

PAGE
88
88
97

101

102

123
125
125
131
140

142

149
149

159

164

171

175

 

 

 

TABLE

10

11

12

13

14

15

16

17

 

LIST OF TABLES

The South in Presidential Competition . . . . . . . .
Elected Southern Republicans, 1950-1968 . . . . . . .
Elections Used in Analyses . . . . . . . . . . . . .
Election Parameters . . . . . . . . . . . . . . . . .
Coefficients of circularity . . . . . . . . . . . . .

Number of Counties Returning Greater Than 45.0 of the
Total Vote for a Republican Candidate . . . . . . . .

Nearest Neighbor Statistics . . . . . . . . . . . . .
Mean Level of Electoral Support by Ring and Sector .

Percent Change in Presidential Elections by Ring and
Sector . . . . . . . . . . . . . . . . . . . . . . .

Percent Change in Senatorial Elections by Ring and
Sector . . . . . . . . . . . . . . . . . . . . . . .

Percent Urban Population by Ring and Sector . . . . .
Percent Change in Urban Population by Ring and Sector
Spearman's Rank Order Coefficients for Comparisons
Involving the Percentage of Urban Population by
County . . . . . . . . . . . . . . . . . . . . . . .

Core Counties Eliminated from Rank Order Analysis . .

Spearman's Rank Order Coefficients for Comparisons
Involving Total Population by County . . . . . . . .

Number of Counties in Each Population Size Category .

The Relationship between Population Size and Level of
Republican Vote . . . . . . . . . . . . . . . . . . .

viii

11

21

58

82

90

99

100

110

114

115

129

132

134

137

141

143

 

 

LIST OF FIGURES

        
   

Study Area . . . . . . . . . . . . . . . . . . . . . .

Standard Ellipses Describing the Spatial Distribution
of the Republican vote from 1948 to 1972 . . . . . . .

Means Centers of Republican Vote . . . . . . . . . . .
The 1948 and 1952 Presidential Elections . . . . . . .
The 1956 and 1960 Presidential Elections . . . . . . .
‘ The 1964 and 1968 Presidential Elections . . . . . . .
The 1972 Presidential Election . . . . . . . . . . . .

The 1948 and 1954 Senatorial Elections . . . . . . .

Ring and Sector Locations in Study Area . . . . . . .

.;. ix
.. .

The 1960 and 1966 Senatorial Elections . . . . . . . .

‘ The 1972 Senatorial Election . . . . . . . . . . . . .

. Percent Urban Population, 1950 and 1960 . . . . . . .
‘ Percent Urban Population, 1970 . . . . . . . . . . . .

‘Population Size and Level of Republican Vote . . . . .

PAGE

52

92

95

103

104

105

106

107

108

109

112

126

127

144

 

 

 

. at“: ’

CHAPTER I

A CONCEPTUAL FOUNDATION FOR THE RESEARCH PROBLEM

Theoretical Developments in Voting Behavior Research

With the systematic replication of measurements of
significant events, the comparative study of popu-
lations in differing political environments, and the
integration of information from interrelated levels
of the political system, the analysis of political
behavior is entering a new phase.

(Campbell, Converse, Miller and Stokes, 1967, 6)

Voting behavior has long held a special fascination for those
social scientists concerned with political phenomena. The advent of
modern technological advances has served to accentuate this attrac—
tion. Increasingly more sophisticated computer technology has made
available an ever increasing volume of information while improved
measurement and analytical capacities have enhanced the scientists'
capacities to deal with it. In the study of electoral behavior, as
in other arenas of social science inquiry, the consequences of these
technological advances have been significant. Much larger volumes of
voting data, offering greater geographic variety, and extended longi-
tudinal coverage, can now be processed in far less time and with far
greater'accuracy than ever before. But, unlike many other arenas of

social science inquiry, in the study of electoral behavior, this in-

formation and technology explosion has not come at the expense of

 

 

theoretical development. Conceptual and theoretical advances have
virtually kept pace with technological developments.

But the pace of theoretical advances has not been uniform across
all arenas of social science inquiry. The primary impetus for the
continued development of theory relating to voting behavior has come
from political scientists; their counterparts in voting behavior in-
quiry in sociology and geography have provided less substance to general
theory of voting behavior. Yet, the dominance of one discipline does
not, as a consequence, make the search for understanding voting behavior
any less an interdisciplinary endeavor. On the one hand, tangential
disciplines cannot strike forth to develop complementary bodies of
theory revolving about individual disciplinary foci. Such a path leads
only to compartmentalization and duplication. Obviously, cross disci—
plinary cooperation and coordination serve well the cause of substantive
theoretical development in voting behavior research. On the other
hand, care must be taken so as not to carry the spirit of such coopera-
tion and uniformity of interest and scientific pursuit so far as to
ignore the merits of academic divisions. Each discipline certainly has
its own unique focus which can contribute fresh new ideas and approaches
simply by asking questions and stressing research areas that other
disciplines fail to do. The inherent danger of academic uniformity is
the potentially stifling effect it might have on such unique directions
and foci of inquiry. It is obvious that as scientists we cannot afford
to neglect any relevant aspect of voting behavior if we are to obtain
our goal of better understanding of human behavior in an electoral

context.

 

Of course, the apparent dichotomy between the uniqueness of

 

division and the uniformity of cooperation is contrived. In working
to the strength of a particular disciplinary focus or mode of inquiry,
a researcher does not necessarily dilute the impact of the results on
general electoral theory. It is usually the design or the implementa-
tion of a research design that accomplishes this. On the contrary,

as Robert Sack (1972) suggests, a unique mode of inquiry well founded
in general literature can serve to advance general theory by asking
questions that might not have been asked otherwise.

It is this line of thought that has guided the inception of this
research effort. The problem approached in this inquiry is certainly
not unique. The problem centers on accounting for changing patterns
of electoral support for a political party and the results are intended
to provide additional insight into the nature of electoral change.

But the primary objectives, goals, hypotheses and research outlook do
reflect a distinct disciplinary bias. The electoral process is viewed
through a geographic or spatial perspective. The primary emphasis is
upon empirical investigation of the nature of spatial structure of the
voting response surface and upon the spatial process that is reflected
in the creation and change of patterns of support for a political party.
Special attention is given to accurate description of the voting re-
sponse structure, but only as a means of providing a foundation for

the principal investigation which centers upon the processes that created
the structure. It is in these processes that we can expect to find the
key that unlocks the complexities of the voting decision itself. Thus,
the primary interests of this inquiry centers on the spatial mechanisms

operative in the voting decision; however, the end results are intended

 

 

  

to provide insight into the decision—making process itself. This is
the goal of all social science inquiry devoted to explanations of voting
behavior.

Although the question will never be addressed directly, an under-
lying issue throughout this inquiry is the role that the political
geographer can play in the continuing theoretical development in voting
behavior research. Elaboration on the nature of this role is not seen
as a central objective. That has been accomplished by others (Cox,
1969; Reynolds, 1969; Reynolds and Archer, 1969; Prescott, 1959, 1969),
and although some limited discussion of the geographer's place in
electoral research follows, it is only by way of elaboration on the

conceptual foundation for the problem.

The Geographer's Place in the Study of Electoral Behavior

The electoral geographer brings to the study of voting behavior
a unique focus of inquiry--a spatial focus. And to the degree that the
literature is relatively "devoid of models for evaluating the impact
of space upon political process" (Reynolds, 1969, 12), the need for
such a focus becomes apparent. For too long geographic studies of
voting behavior have concentrated on non-spatial explanations. Geog-
raphers interested in the explanation of voting behavior have relied
upon what Cox has described as the "aspatial treatment" where voting
behavior is viewed as a function of within-area unit economic and social
characteristics (Cox, 1969, 113). Such an approach does not operate
from the strength of the discipline of geography since the political
unit is removed from the space in which it is operative. Thus, by

separating the political unit from the spaceixiwhich it rests,

 

 

geographers are seen to minimize the strength of their potential con-
tributions to electoral research. The geographer's greatest potential
strength lies in the examination of electoral behavior within its spa-
tial milieu. Any other approach makes the geographer virtually indis-
tinguishable from any social scientist who does comparative analyses of
areal units. Until the geographer's consideration of electoral behavior
makes contributions to electoral theory, electoral geography's inde-
pendent existence will not have been demonstrated. Cox summarizes this
well when he suggests that:

an approach which emphasizes the space in which areal

units or voters are embedded and the relationships of

these units across space, not only provides electoral

geography with a justification for an existence inde-

pendent of comparative studies in political science; it

also places the systematic field in the spatial mainstream

of current geographical methodology and makes available

the accumulating body of ideas relating to the geometry

and the duality of spatial structure and spatial inter-

action (Cox, 1969, 112).
If such contributions are to be forthcoming, then we must return to
asking the type of questions that make us geographers. And these are
questions about space and location.

The above is not an implication that geography should set about
developing a set of theory exclusive to electoral geography; nor does
this imply that geography should establish a segment of electoral
behavior research quite apart from the remainder of social science.
Rather the argument simply calls for the geographer to do what he does
best. Geographers are best equipped to consider the spatial dynamics

of behavioral phenomena. And in this lies their greatest potential

contribution to voting behavior research. Always the primary goal of

 

 

 

the electoral geographer is the most thorough understanding of electoral
behavior possible. And always the realization must be present that the
body of theory and research findings that will eventually enable scien-

tists to obtain that goal will be multi-disciplinary in character.

Electoral Change As a Conceptual Framework

DYNAMICS--The physical or moral forces that produce motion
or change in any field or system (American
Heritage Dictionary, 1971).

As the definition above suggests, the dynamics of space has to
do with the processes that induce movement or motion across space.
Since the term has an inherently non—static connotation, the concept
of change is a significant part of any consideration of the spatial
dynamics of a phenomena. In electoral behavior a change in the spatial
pattern of voting response is an inherently dynamic process. Thus, for
the geographer wishing to focus on the spatial dynamics of voting be—
havior, one possible avenue is an inquiry into the nature of electoral
change. However, even a most casual perusal of geographic inquiry will
reveal very little research emphasis on the "physical or moral forces"
that induce spatial change or spatial motion in the voting response
system. Geographers have been too much occupied in analysis of static
patterns to delve into the processes that created the patterns. They
have utilized "spatial coincidence“ models to depict the areal covaria-
tion of social, economic and electoral behavior. They have also relied
heavily on pattern analysis of voting response maps to locate spatial
consistency. And such research has proven rewarding by providing

knowledge of spatial regularity in state and national level electoral

 

 

.JL.

behavior patterns. Hence, we are aware of distinctive regional regular-
ity in voting response patterns such as the protracted allegiance of the
American South to a single party. But, as a consequence of the failure
of researchers to focus on the more dynamic spatial components of the
electoral response, we know little about the processes that induce

such regional homogeneity in voting response. And we know even less
about the processes that induce changes in patterns of regional homo-
geneity in voting response or about the mechanics of that change. But
it is just such areas of inquiry that form the core of geography's
potential contributions to electoral theory.

While inquiry into the nature of electoral change is a potentially
useful method of probing spatial dynamics of electoral behavior, it only
becomes so if the inquiry proceeds beyond mere description. As Kevin
Cox (1969) suggests, it is one task to identify spatial regularity in
voting response surfaces,1 and quite another to connect these regular-
ities with relevant elements of the spatial process that lie at the
roots of their inception. A description of the electoral response
patterns of southern voters for Republican candidates may demonstrate
that striking regularity exists in the character of this response.

Such regularities, once delimited, may even suggest possible spatial
dependencies. But, in the long run, it is still necessary to pose the

question of the usefulness or utility of the regularities. To paraphrase

 

1In this inquiry several terms are used interchangeably in addres-
sing the patterns created when election results are depicted graphically.
Among these are voting response surface (or patterns) and electoral
response surfaces (or patterns). It may be useful to remember that
response is used interchangeably with result and surface interchangeably
with patterns.

Cox, what utility does the recognition of the regularity of the voting

 

response patterns for Republican candidates in southern elections have
in terms of isolating a possible spatial process? Its utility may lie
in the knowledge gained of the spatial dynamics of the process of change
in these patterns. Describing the patterns of electoral response may
provide clues to the spatial process inherent in electoral change.
Describing the process of change itself provides insight into the ma—
chinery of that process. If by definition, change implies process,
then it is difficult to imagine an environment more conducive to the
examination of the spatial dynamics of voting response surfaces than
one in which constituencies have had rather recent exposure to rapid
electoral change. A situation in which such change is occurring within
a spatial context, previously characterized by a historic and rather
distinctive degree of spatial regularity and homogeneity, would appear

to afford excellent conditions for the examination of the spatial dynam-

 

‘ ics of electoral change. Rarely does a social scientist encounter such
: laboratory-like conditions; but, in terms of the elements necessary for
: probing the spatial dynamics of voting response surfaces, few better

L laboratories are to be found at a macro-level than the post-World War II
E American South.1 This inquiry will use this "laboratory" to examine the

spatial dynamics of electoral change.

1Throughout this text the South will be defined as the states of
the old Confederacy: Alabama, Arkansas, Georgia, Florida, Louisiana,
Mississippi, North Carolina, South Carolina, Tennessee, Texas and
Virginia.

 

 

 

4w?—

The South As a Regional Laboratory

Southerners and Non—Southerners alike have become ac-
customed to the image of the Southerner as a bigoted,
uneducated, rural boob.

(Mack, 1970, 2)

It is usually the case that stereotypes ignore reality. But, un-
fortunately even those stereotypes born of honest representation often
outlive the progress of time and reality. Consider the arch stereotype
of the South, which has changed but little from the turn of the century.
Until quite recently it would not be too far amiss to describe a typical
image of the South as a languid, agricultural society, dominated by
the very rich, inhabited by the very poor, ruled by a phalanx of bigots,
yet characterized by a distinct arcadian and rustic simplicity. If any
one factor contributed to the longevity of such distinctive southern

stereotypes, it was the legendary political cohesiveness of the former

Confederacy.

The Democrats are a party of the South . . .; the
Republicans a party of the North, . . . (Rossiter, 1960,
105).

In political context, the terms "Democratic" and "South“ became in-
separately molded in an image of solid electoral support that served
to protect the political hegemony of the white southerner against
encroachment by the Black southerner and his northern political allies.
of course, this image itself ignores the reality of internal complexity
and political dissent within the southern Democratic party (Grantham,
1963: Woodward, 1951; Key, 1949), but it has been, nonetheless, the

dominant stereotype of the political South.

10

The South is now in an accelerated state of change. Economic
development in the form of expanding trade and commerce and industry
have diminished agriculture‘s economic pre-eminence and spawning wide-
spread urban growth has replaced southern rusticity. Concomitant with
the economic and social change is a very real political change. And
just as politics once contributed to a distinctive southern stereotype
so has it now become a vital part of the "new“ image of the South--
the changing South. When one examines the gross statistics, it is not
hard to understand why since the elements of political erosion appear

to have played havoc with Democratic solidity.

 

' From the end of Reconstruction until World War II, the South

3 remained an American political constant by giving almost unswerving

; electoral allegiance to the national Democratic party. In the 17

i presidential elections from 1880 to 1944 the 11 states of the old

I Confederacy deviated from their Democratic allegiance only 6 out of a
total of 187 potential times, or 3.2 percent of the time (see Table l).

i But, in view of what has transpired since 1948, the once-solid Demo-

, cratic southern wall can certainly be said to have been cracked, if not

' splintered asunder. In the seven presidential elections since 1948,

} Republican presidential candidates alone have cracked Democratic soli-

' darity 33 out of a total 77 potential times, or 42.9 percent of the time

(Table 1). Republican candidates, Eisenhower, Nixon and Goldwater,

enjoyed a fair measure of success if compared to their pre-l948 pre-

‘ decessors. Certainly the candidacies of successive Republican and

independent party presidential hopefuls as well as others at lower
levels of electoral competition, are all phenomena that contribute to a

new southern stereotype--the changing South. And they are all

.Hofisu >3
sOwumemfioo use on .m . 3&3 .oofiruom mauwuuwsa Hmoosmmmoumcou ".0 .o .coumcanmma
vmma ou mmea scum moumvflocnu Hmausoofimoum .uuommm Hwaowmm mfiuouumso Hucowmmwumcoo "mumbom

 

.huouow> wumofioomu Hmflusoowwoum caumuooEoo xseam

 

 

 

.huou0fl> oumowosmo adducuoflmoum >uumm phase a
.hu0uoa> oumoflocmo Hewusmoflmoum cuoaansmom x

mm he "masses 0 hmfl "quaoa
ml MI m m m m m 4n. E” nuns—flout,
m n m m m H ea nexus
m h m m m m m N ha oonnoscoa
1 n n m m m a 0 ea mcaaoueo Spoon
1 N a m m H 2 339:6 .332
m n m e m a 0 ea ammwmmammw:
m n m a m m 9 0 ha confluence
m .. m a m o S 3983
m e m m m m m H ha mofiuoam
H e m a 0 5H muucexua
N n m .H. m B o x...” means:

mwauou0fi> mnmalwcma we mo vo om mm mm we mmauouoa> evmauomma mumum
moo macauooam moo mcoauouam
Hence Haves Hmuoa Hmuoa
memanomma

ZOHBHBMQEOU AdHBzmnHmmmm 2H mgOm an.

H mqmda

 

 

0.5

{U

 

 

 

 

 

12

illustrative of successful electoral challenges to Democratic hegemony
in the South. In terms of electoral politics the South may still be
essentially "Democratic" but it is no longer "solid."

At present, the authenticity of either the “solid“ or “changing"
stereotypel'ofthe politics of the South may certainly be brought to
question. But authenticity of stereotypes is not the issue here; the
central issue is the changing southern electoral response patterns.

More specifically the concern is with the "emergence" of the Republican
party as a serious contender for southern votes. And the very coinci-
dence of strong resistance to change and the existence of strong social,
economic and technological pressures to institute this change make the
South an attractive choice for examining the spatial processes involved
in changing traditional patterns of regional electoral behavior. Thus,
in this analysis the conceptual vehicle is electoral change; the elec-
toral response patterns of southern voters for Republican party candi-
dates are the primary foci. The delimitation of spatial and statistical

regularities and modes of electoral change is the primary goal.

The Problem
Specific Statement of the Problem
The problem entails: 1) the spatial and numerical delimitation

and measurement of electoral support for the Republican party in the

 

1In some circles the "changing" South is in as much danger of be-
coming a stereotype as the "old" or "solid" South ever was. Two well-
known books, I'll Take My Stand by Twelve Southerners (1930) and Egg
Can't Eat Maggolias edited by H. Brandt Ayers and Thomas H. Naylor
(1972) are prime examples of a healthy literary and academic concern
with "change" in the South. But I do not wish to dwell long on the
issue of stereotype and their authenticity of life cycle. The stereo-
type is only a grammatical vehicle for introducing the conceptual
vehicle used in this analysis.

 

13

American South during the period 1948-1972; 2) the association of such
spatial regularities with the continued growth of that electoral sup-
port; 3) the description and accurate measurement of the processes that
underlie the change in patterns of electoral support with particular
emphasis on the relationship between urban size and hierarchy on the
growth of Republicanism; and 4) the anticipation of future patterns

of electoral response in the South.

Elaboration on the Problem

This study begins with the premise that the politically solid
South--the Democratic South-—which existed in American electoral poli-
tics for more than a half a century, is now changing. The initial signs
of this electoral change were first notable at the presidential level of
competition but the manifestation of Republican party competiveness at
lower levels of electoral competition, such as the senatorial level, is
now becoming increasingly apparent. The problem becomes one of iden—
tifying and accurately delimiting the specific regions of significant
Republican electoral support and the measurement of changes which can be
labeled Republican growth. Having identified the areas of salient
changes in electoral support, the next problem is to search out and
identify the elements of spatial regularity that are independently
associated with this change. If such relationships can be accurately
identified then perhaps areas of future growth may be anticipated.
The principal focus in such a quest would lead to the spatial processes
associated with electoral change. And in this inquiry particular
attention will be given to the influence of urban size upon electoral

change.

 

 

 

 

CHAPTER II

A LITERATURE FOUNDATION FOR THE RESEARCH PROBLEM

As is frequently the case in geographic inquiry, the literature
which serves as the conceptual and theoretical foundation for this
research problem can be divided into two parts: 1) that literature
pertaining to the geographic setting; and 2) that literature relating
to the theory upon which the problem itself depends. Naturally, the
two parts have no distinct line of demarcation and it is frequently
impossible to decide from which specific objectives or hypotheses are
derived. But, in terms of conceptual and literary organization, it
was found particularly useful in this inquiry to divide the discussion
of the literature into two major categories--the literature pertaining
to the South, and the conceptual literature relating to the general
body of theory of electoral behavior that is applicable.

In the first section of this chapter, the literature pertaining
to the southern political experience is discussed. Specific attention
is devoted to the establishment of the atmosphere for electoral and
political change in the post-World War II South, and to the elements
of the spatial structure linked to this change. The second section
then is devoted to the literature that served as a conceptual founda-
tion for the research problem. Specific attention is given to the

literature relating to changing locational bases of party support and

14

 

“’1‘

15

to the mechanisms inherent in changing voting response patterns in a
subnational or regional context. The primary focus is on literature
that relates to the influence of urban size upon changing regional
patterns of voting response. Following this review a final section
contains the specific objectives and the working hypotheses which

guide the subsequent course of this research.
The South in Voting Literature

The Republican party is aiming an arrow straight at

the heart of the white men's civilization in the South,

and it is distressing to know that we have in our midst

good men and women who are apparently lending aid and

comfort to a common enemy (The Watchman, 1928 and Strong,

1963).
The Question of Electoral Change in the South

The legend of the old South--the romantic cult of the "Lost
Cause"--vague memories, family tales, and grim stories of human suf-
fering in the era of Reconstruction are all elements of an ancient
prejudice against the Republican party in the South. Although the Civil
War gave the GOP a dominant position in American politics that pre-
vailed until the 1930's, its status in the South after Reconstruction
proved shortlived. After the removal of federal forces, the slow evo-
lution of the distinctly unique southern character so directly tied
to the "Lost Cause" legend began. The fortunes of the Republican party
in the South are closely intertwined with the evolution of this dis-
tinct brand of regional character--this southern sectionalism.
In his treatment of southern political sectionalism, Dewey

Grantham (1967) divides its evolution into three major phases. Using

his general outline, it is also possible to trace the evolution of an

 

-—~—v~-—

16

"electoral" sectionalism, even though the concern was certainly less
specific. In the first phase, from 1870 to the turn of the century,
regional unity was still in the process of formation. Negro disen-
franchisement had not yet been accomplished and Republicanism was still
significant in many areas.

In the second phase, from 1900 to 1930, southern political unity
reached its peak and the Democratic party ruled supreme. Democrats
cemented their dominance by controlling voting procedures, apportion-
ment of legislative districts, and by gerrymandering the Republican
party out of contention. The threat of outside intervention was slight
and there was little danger to the one party system which emerged from
the resolve of the white southerner to hold Negroes to a well-defined
economic, social and political place. The one party--the Democratic
party--system was the weapon used to accomplish this resolve (Heard,
1952, 145-146). As V. 0. Key (1949) demonstrated, the race issue
dominated southern politics and served to suppress any meaningful polit-
ical division among southerners. Threats to Democratic party superi-
ority came not from the GOP, but from internal dissension, and fragmen-
tation, such as the movements of the Populists and various agrarian
reform groups. With the exception of historical hard core centers of
Republican support, such as the isolated mountain areas of Virginia,
North Carolina and Tennessee, the South was solid one party during
this phase.

The New Deal ushered in the third phase of the evolution of
sectionalism. The developments of the 1930's and 1940's encouraged
fragmentation in the South. The changing nature of the Democratic

party and the lessened importance of the South in it had their effects

 

17

upon Democratic party domination. Increased federal aid, and with it
increased federal intervention, revived the threat of a renewed inva-
sion from without, and white southerners rallied to the age—old magic
of the defense of the "southern way of life." But unlike before, this
sectional dynamism was couched in defeat—-the southerner did not expect
to win. The changes wrought by encroaching industrialism and urbanism
had taken their toll of sectional fervor. The South was reacting to
historical stimuli as it was simply "playing a role to which it had
long been accustomed" (Grantham, 1967, 49-50).

By then the southerner's political weapon--the one party system--
had begun to disintegrate. The social and economic contextiJIwhich the
system had operated had undergone rapid change and the political system
began to follow suit. The Truman civil rights program and the civil
rights platform of the Democratic national convention in 1948 split
the party. The proponents of status quo and southern political unity
found little reassurance in the Democratic party and they set an ex-
ample of disruption of the system with the 1948 Dixiecrat movement.
The national Democratic party increasingly disassociated itself with
the southern wing on the question of race. Once the power of the
weapon was gone and the system could no longer accomplish its goal--
the subjugation of the southern Black-—then a major obstacle to change
was removed. As Donald Strong argues, "Once the sharpness of the racial
issue was dulled, then the southerner was free to vote his economic
interests." In Strong's opinion, this interest lay with the Republican
party (Strong, 1963).

Given southern dissatisfaction with the Democratic party, is the

only alternative the Republican party? To reason so is to dismiss the

 

18

feasibility of a third-party South. But then in the national political
arena third party movements have proven neither durable nor productive
as a means of effective long-term protest.1 Perhaps it is as Kevin
Phillips suggests:
. . . third parties are not likely to persist long,
they are inevitable casualties of realignment (Phillips,
1970, 287).
But in a southern context it is more likely that, as Phillip Converse
suggests, they are casualties of adjustments in major party philos-
ophies and strategies that come about as a consequence of attending
to the messages of short-term protest.
Thus while these may be spates of Dixiecrat protest,
there seems to be little stomach among southern
politicians for developing a truely independent third
party, there are many forces which are operative in
American politics to counter such a development in the
long run (Converse, 1967, 214).
With remarkable measure of insight Alexander Heard (1952) summed up
the prospects of third party movements in the South thus:
. . . in the long run Southern conservatives will find
neither in a separatist group nor in the Democratic
party, an adequate vehicle of political expression. If
this is true, they must turn to the Republican party
(Heard, 1952, 247).

Using the advantage of almost two decades, Kevin Phillips (1970)

suggested a number of circumstances which would push the South toward

 

1For example, consider the agrarian reform movements such as the
Grange and the Southern Farmers Alliance, the Populists, the Dixie-
crats, and the most recent manifestation of the American Independent
party of George Wallace.

 

19

the Republican party. Among these are the political activity of blacks
within the southern Democratic party, the alien nature of the national
Democratic party to the white South, the failure of the Wallace balance
of power strategy, and the precedent of an "opinion molding upper middle
class of the urban South which is apparently trending Republican."
Phillips concluded:
The gathering Republicanism of the Outer South virtually
dictates the coming alignment of the Deep South. For
national political reasons, the Republican party cannot
go to the Deep South, but for all of the above mentioned
reasons, the Deep South must soon go to the national
GOP (Phillips, 1970, 287).
Thus, when one considers the question of electoral change in the
South, an inevitable consequence is attendance to the prophecies of
growing Republicanism. And, perhaps the evidence of electoral suc-
cesses warrants such attendance. Consider the southern successes of
Eisenhower in 1952 and 1956 and Nixon in 1960, 1968 and 1972. These
are indicative of a change in the patterns of presidential competition.
But perhaps the most spectacular evidence of this change came with the
Republican success in the presidential election of 1964. Of the
overwhelming success of candidate Barry Goldwater in the Deep South,
Bernard Cosman writes:
. . . there is now the real possibility that an enduring
grassroots Republicanism may emerge among the white voters
who live in rural areas . . . In short, from the 1964
presidential election "pockets of realignment" may emerge
within sectors of the Deep South (Cosman, 1966a, 131).

In Cosman's opinion the success of Barry Goldwater in the South, and

the increasingly active role of southern delegates in the Republican

 

20

party conventions suggest that 1964 could have been the beginning of
competitive politics in the Deep South (Cosman, 1966b, 1968).

All of this may, of course, be a manisfestation of what v. 0. Key
(1949) labeled as presidential Republicanism. Under this concept life-
long southern Democrats, due to disaffection with the policies of the
national Democratic party offer support for Republican presidential
candidates while continuing to support Democratic candidates at lower
levels of electoral competition. Yet as Havard (1972) has indicated,
"presidential Republicanism, even where it may have gone beyond the
point of mere protest against the national Democratic party, is no
final indication that the South has abandoned its one partyism in
favor of organized oppositional politics" (Havard, 1972, 721). Havard
points out that the Republican gains, although significant at the level
of national or statewide electoral competition, have been somewhat
sporadic at the level of state legislative competition and below.

As Table 2 indicates, the Republican party has indeed made noteworthy
and rather consistent inroads at other levels of electoral competition
than the presidential level. Perhaps successes at lower levels of
electoral competition will prove just as dramatic in the immediate
future.

Thus, the evidence suggests rather extensive Republican party
gains in the South in comparatively recent times. Republicans are
gaining political offices that 20 years ago were unavailable to, and
in many instances even uncontested by, Republican candidates. This
evidence has for some (Heard, 1952; Phillips, 1970) been suggestive
of a fundamental realignment of party affiliations in the American

South. This question was discussed at length by Phillip Converse

v1

21

.0 .o .couocHnmmz .AchHuHoo oumm
nuamsc mamH-ommH .mmumum woven: or» no possumna Hmufiumeuuum .msmcmo on» no smousm .m.o "momaom

.ooHnmm use» 03» some cH moumum :uonuoom HH on» scum
monuwo women use no some on omuooHo mchHHhsmom mo amass: Hope» on» some comes one mousmHm ommnBH

 

h m v N H H muouecom
HddOHusz
vm mm on mm 0H HH 5 s h h m N wo>Humucuuoumum
HecOHuuz
m N N m muocuo>oo

 

meH osmH mme momH vmmH NmmH oomH mmmH wmmH vmmH ammH ommH

 

momHlommH

Hmz<UquDmmm zummfibom DmBUmHm

N qudfi

.M -

 

22

(1963, 1967), who concludedthat,while the South had indeed undergone
extensive political change in the early 1960's, there was no drastic
realignment of party identification. Rather Converse suggests that a
convergence trend, in which the South will swing more into line with
national patterns of party competition, is a more plausible explanation.
The question of partisan realignment is introduced as a means of
demonstrating the extent to which social scientists have carried their
thinking on the electoral change that has taken place in the South over
the last two decades. The question of realignment of individual
partisan attitudes is, at least for this inquiry, a moot issue, since
the research efforts of this study will focus upon the aggregate level
of analysis. Obviously to deal effectively with party identification,
it becomes necessary to delve into the attitudinal framework of the
individual voters and thus into areas that would entail extensive
survey level analysis. In this inquiry the concern is with changes in
the total or aggregate voting responses surfaces over time. To discern
such changes it is not necessary to have complete knowledge of the
political party with which a voter identifies. At this point the con-
cern is with the aggregate manifestations of the individual voting
decision on election day and the spatial mechanisms operative in the
changing character of the resulting voting response surface. And
these arguments have been presented in support of the thesis of changing
patterns of electoral support in the South; for, in order to examine
the spatial dynamics of electoral change, it must first be established
that change is occurring. In this vein from the literature presented
there is the overwhelming agreement on the existence of electoral change,

if not on the extent of it.

23

The Components of Southern Electoral Change

The last time the Republicans were in Atlanta was 100
years ago. They burned it down (Jimmy Carter, Democratic
Governor of Georgia, as quoted in Murphey and Gulliver,
1971, 173).

Given the general consensus on the existence of electoral change

 

. in the South, and upon the beneficiary of electoral indecision, viz.,
i the Republican party, what then are the components of the process of
; electoral change? Can the medium of electoral change be isolated?
More specifically can the spatial mechanisms inherent in the change be
‘ identified? Again the literature provides useful direction.

A considerable portion of the literature devoted to the growth
of Republican electoral support in the South concentrates on the in-
fluence of the urban and industrial sectors of southern life as
particularly conducive elements in the Republican party growth. As
early as 1949, V. 0. Key was suggesting that the future of Republicanism
in the South would be closely tied to industrial and urban growth. Key's
prophecies have since been echoed by others but more importantly the
particular influence of the urban sector likewise has been demonstrated
' in several empirical investigations.

In assessing the results of the 1952 presidential election,
Donald Strong (1955a, 1955b) demonstrated the greater electoral strength

i of candidate Eisenhower in larger cities. Strong suggested that the
I apparent defection of wealthier city dwellers to Republican ranks
. could mean that future GOP successes in the South could prove greater
in areas with larger urban populations. In a comparative study of the

1952 and 1956 results, in which he traced the existence of urban

24

Republicanism from the late 1930's, Strong suggested that an enduring
brand of urban-centered presidential Republicanism had indeed been
cemented in the 1952 and 1956 contests (Strong, 1960). This brand of
Republicanism was centered upon an "ever increasing number of prosperous
southern urbanites who are reacting as unfavorably to the economic poli—
cies of the Democratic party as do their counterparts in northern
cities" (Strong, 1960, 49).
In a similar analysis of the 1960 presidential election results

I for southern states, Bernard Cosman found that Richard Nixon came close
to Ike's showing in the cities of the non-Deep South, and improved upon
Ike's showing in the Deep South cities, "which have large nonwhite

( populations, small percentages of Negro registrants, and few Catholics"
(Cosman, 1962, 320). To Cosman, Nixon's 1960 showing indicated that,

i "enduring presidential Republicanism has developed more rapidly in the

metropolitan South" (Cosman, 1962, 321), while no comparable develop—

ment could be pinpointed elsewhere. Although Cosman suggested that GOP

, success below the presidential level was also an urban centered
phenomenon, he could see no large scale Republican successes beyond the
presidential level of competition in the immediate future.

Since neither Strong nor Cosman found comparable development of

| Republican support outside of the urban South, indications were that

. this was a metropolitan Republicanism. This conclusion was reinforced
by Cosman's examination of the 1964 election results in which he found

i continuing development of non-Deep South Republicanism along urban-
industrial lines, and possible development of grass roots rural

Republicanism in the Deep South (Cosman, 1966a).

 

 

25

In an analysis of the Republican vote in 57 urban counties for
six presidential elections, Stanley Brunn and Gerald Ingalls (1972)
addressed the question of the relationship between urban size and sup-
port for GOP presidential candidates. It was discovered that larger
urban counties throughout the South gave heavier support to Republican
candidates, while urban counties of the Rim South generally offered
greater electoral support for GOP candidates than did their counterparts
in Deep South states.

In concentrating upon the relationship between urban structure and
the growth of electoral support for a political party, the impact of
additional components of the electoral change occurring in the South
is thus overshadowed. This by no means is meant to deny the existence
nor the impact of factors such as socioeconomic class (SEC) in the de-
velopment of support for the Republican party in the South. The role
that SEC plays in changing electoral allegiances was demonstrated by
Strong (1955a; 1955b; 1960) and Cosman (1962; 1966a; 1966b). In addi-
tion one need only consider such findings as those of Prothro, Campbell
and Gregg (1958), Matthews and Prothro (1964; 1966) and Drum and
McCranie (1970) to verify the impact of social class on attitudinal
change. But such components are best left to inquiry based on survey
design. As previously argued, explanation of such components from the
realm of aggregate data analysis involves a considerable degree of risk
in transferring inferences made on basis of aggregate data to individual
attitudes or behavior. To examine the impact of social economic
class on electoral behavior as represented by the aggregate level

voting response surface, it would be necessary to resort to a spatial

 

. flfi.‘—'

26

coincidence model.1 Space is not the crucial element in such an ex-
amination. It is only an area defined by definite boundaries (usually
county) where a given level of electoral support coincides with some
given level of socioeconomic class. It would appear more within the
geographers' domain to examine behavior within its spatial milieu.
Thus, the focus of this study is on the process of change in successive
electoral response surfaces and how that change is influenced by a
distinct spatial structure--the urban hierarchy.

Already there exists abundant empirical evidence on the existence
of a strong relationship between the urban spatial structure in the
South and the spatial structure of the Republican voting response
surfaces. While such evidence lends credence to the notion of an urban
centered concentration for the development of Republican electoral sup-
port, it fails to depict accurately the exact nature of this relation-
ship. For example, Brunn and Ingalls (1972) have demonstrated that the
support given to GOP candidates at the presidential level is greater in
larger cities than in smaller; but does this relationship hold true
at other levels of electoral competition? Is the support given to
GOP candidates subject to a time differential or time lag between
levels of the urban hierarchy? Is the support for Republican candidates
confined to more urban areas or does such support diffuse outward from
these centers to less urban areas? Questions such as these relate
to the spatial process of changing electoral response patterns. As

such they form the primary focus of this inquiry. Before we answer

 

11h such a model the political unit is essentially abstracted
from the space around it. It becomes only a convenient devise for
examining the relationship between the level of the vote and SEC.

 

27

these queries, we must first examine the second major section of lit-

erature upon which this examination is based.

The Problem in the General Theoretical Literature

Thus far an attempt has been made to provide a foundation from
which the research problem can be examined in light of the uniqueness
that is inherent in any empirical inquiry. Thus, the South as a unique
portion of political space has been examined in terms of previous re-
search efforts which have focused on the southern political experience.
However, while the South may possess certain unique qualities that set
it apart, this does not preclude utilization of the southern political
example to expand the general body of electoral theory. Thus, while
it is important to remain alert tothenore unique qualities of the
South as one example of electoral change, it is essential that the
research problem focus on the aspects of this inquiry which affords the
best opportunity to expand the general theoretical base. Thus, we now
turn to that portion of the literature which provides the conceptual
or theoretical foundation. Since this inquiry has a spatial focus,
this literature is concerned primarily with the spatial approach to
voting analysis. However, the literature which concerns the relation-
ship between urban structure and the spatial dynamics of changing

party preferences is also discussed.

The Spatial Approach to Voting Analysis

For the most part, students of electoral behavior

have examined the areal covariation of socio-economic
characteristics of constituencies and their voting be-
havior, but have contributed little to a further under-
standing of the spatial dynamics of changing voting pref-
erences (Barnett, 1972, 9).

 

 

28

The examination of the spatial dynamics of voting preferences is
a comparatively recent phenomenon in electoral geography. It is only
the decades of the 1960's and 1970's that have witnessed a concern for
the less static aspects of the voting response. In the decades that
preceded 1960, geographers tended to rely heavily on maps of voting
response patterns and upon verbal description of these patterns. In
fact, until comparatively recently, the processes that produced these
voting responses were often only alluded to via subjective implication
of possible causal linkages. Under such a framework the basic model of
geographic inquiry into electoral behavior normally took the form of:
l) the preparation of maps of election responses; 2) the description
of the resulting map patterns; and 3) "a search for areally covariant
constituency attributes" (Reynolds and Archer, 1971, 1).

Consider one of the earliest efforts of geographers to probe the
behavior of an electorate. In a search for national factors that con-
tribute to the shaping of regional-territorial political opinion, E.
Krehbiel (1916) analyzed British parliamentary election returns between
1885 and 1910. Krehbiel's work is significant not so much for his re-
sults--he discovered that liberal representatives were elected from
industrial and poor farming areas, while conservative representatives
were elected from the more fertile agriculture areas—-as for his ap-
proach to and conceptualization of the problem. His use of maps as a
basis for description of voting response structures and for subjective
determination of social and economic causal linkages, and his concen-
tration on national elections are indicative of the pattern of inquiry
used by electoral geographers for five decades. Under this pattern

of inquiry the dominant interest was in justification or verification

 

29

of national, regional or state level cultural or political boundaries,
and not necessarily in the elections of distinctly spatial processes
that influenced the resulting response patterns.

This concept of electoral analysis and this approach proved to
be quite long-lived. So much so that a half of century later a French
geographer, J. Billet (1958) could suggest that indeed the formulation
of laws in the cataloging or the explanation of election results was
not the milieu of the geographer. Rather the geographer's role was to
assess the "economic, historical, sociological, political, psychological,
and demographic factors, which together shape public opinion" (Prescott,
1969, 297). Thus, the basic requirement of an electoral study, if it
was to provide a profitable flow of ideas back to geography, was that
the "election issues should concern the raison d'étre of the state"
(Prescott, 1969, 300). Elections not fulfilling this basic requirement
were, at best, useful only as a means of crosschecking other methods
of delimiting the integral political regions of the state. Of course,
such assessments placed severe restrictions on electoral studies done
by geographers since if electoral research did not benefit geography,
it was not really a profitable venture. Such an introverted attitude,
while highly restrictive of the role that geography could play in
electoral research, represents an accurate estimation of the approach
to electoral research in geography from Krehbiel to the 1960's.

The latter part of the 1960's brought to geography the revolution
of method and technology characteristic of social science in the
past two decades. Social scientists began to discover the value of
computers and mathematical and statistical techniques to research,
particularly in electoral research with its virtually unlimited

quantities of election data.

 

 

 

3O

Coincident with this revolution of technology and method in
geography was a reconsideration of the concept of the geographer's
role in electoral inquiry and of his approach to electoral analysis.
Geographers broke with the tradition of limiting their inquiry to state,
regional or national elections and began to probe the complexities of
lower levels of analysis. Particular attention was given to electoral
behavior within the urban environment (Lewis, 1965; Kasperson, 1969a,
1969b) and to sub-sectors of the urban scene (Cox, 1968, 1970, 1971).
Although geographic inquiry was still predominately descriptive, the
description became increasingly more numeric (Haring, 1959; Roberts and
Rumage, 1965; Lewis and Skipworth, 1966). Geographers also began to
explore other areas of electoral inquiry, as witnessed by the examina-
tion of non-partisan elections (Brunn, Hoffman and Romsa, 1969a, 1969b;
Hoffman, 1970).
All of these "new" directions in electoral research were indica-
tive of the changed attitude towards geographic research of election
responses. Geographers began to question their approach to electoral
inquiry and their role in electoral research. As discussed in Chapter
I, the conventional areal analysis of elections that had so long
dominated geography came under criticism as "aspatial." Under the
areal analysis approach, geographers were seen as not working from the
strength of their training in spatial analysis since the
. . . areal units under consideration are abstracted from
the "space" in which they are imbedded . . . (Reynolds
and Archer, 1969, 2).

The importance of such criticism is that it introduced the notion that

the explanation of voting behavior lay not in description of structure

 

31

and static patterns of voting response but rather in description of
processes that create the structure. But perhaps more importantly
such criticism served to expand the limited research horizons of
electoral geography. By stressing the facets of electoral inquiry with
which they are most qualified to deal, geographers are at once retain-
ing their identification while offering up a potentially significant
new dimension to electoral research.

It is apparent that in the pursuit of understanding electoral
behavior, geographers have chosen to concentrate on the static and
speculatetMIthedynamic, map the response and imply the existence of
process linkages, and describe the pattern and conjecture cause. While
such inquiry has provided a wealth of empirical evidence of state,
regional and national response habits, it has given little indication
of processes behind the more dynamic aspects of the voting response.
Such inquiry does, however, provide a means of conceptualizing the
problem under consideration in this study. From the development of
electoral geography described above, it becomes important that electoral
inquiry seek a balance between description of pattern and description
of process. Thus, patterns of voting response in the South must be
accurately described numerically, cartographically and verbally; how-
ever, this description must be rooted in an awareness of the spatial
system in which the voting responses occur. In this case the spatial
system can be limited since the literature suggests a distinct urban
bias in the changing patterns of electoral support for the Republican

party.

32

Urban Size and Electoral Change

Support for a political party has distinct spatial as well as
social and economic dimensions. Hence, changes in the spatial patterns
of electoral response are indicative of the movement of a new political
idea across space. With its strong tradition of diffusion research
one might expect geography to have made substantial contributions to
the understanding of the spatial dynamics of changing patterns of
electoral response. Yet surprisingly few examples exist in the
geographic literature where consideration is given to the spatial
dynamics of electoral change and even less where the relationship of
change to urban structure is examined.

The geographical research that does exist on the diffusion of
support for political movements has focused on the concept of spatial
contagion or the interpersonal transfer of political information over
space (Gould, 1961; Cox, 1968; Reynolds, 1969; Reynolds and Archer,
1969; Barnett, 1972, 1973). For the most part contagion research has
concentrated on the processes pertaining to the transfer of political
information and the influence of this information on the electoral de-
cision. For example, Reynolds and Archer (1969) discovered that the
spatial form of the voting response surface in Indianapolis mayoral
elections could not be explained solely on the basis of socioeconomic
indicators. Apparently the "within-precinct" flow of political informa-
tion was also an active factor. In subsequent attempts to model this
transfer or flow of information, Reynolds (1969) examined the signifi-
cance of relative candidate locations, while Cox turned to the influence
of within constituency interaction or "neighborhood" effect (Cox, 1971)

and to residential location (Cox, 1970). And in two examples of

33

inquiry into the diffusion of political support for a new political move-
ment, J. R. Barnett examined the electoral change associated with the
growth of the Social Credit League in New Zealand (1972) and the spatial
growth and subsidence of the Danish communist party between 1920 and

1964 (1973).

Such inquiries into the spatial dynamics of electoral change
provide a basis for grasping the significance of the contagion or
transfer effect on the decision-making process. But what of the
effect that specific components of the spatial structure have on elec-
toral response patterns? Consider one such component--the urban struc-
ture. What is the nature of the relationship between urban structure
and changing patterns of electoral response? In the example of the
Social Credit League in New Zealand, Barnett (1972) discovered that as
urban size increased the support for the party decreased, a factor he
attributed to the mobilizing and communication effects of small town
social systems and to the contagion effect. Cox (1971) has suggested
that the interpersonal transfer of political information is accentuated
in smaller urban places since the transfer of information and the iden-
tification with the community is heightened in smaller towns. In a
larger community the neighborhood effect is accentuated and inter-
personal contact suffers a corresponding overall decrease due to the
constraints of residential segregation on personal mobility. These
findings appear to indicate that urban size is inversely related to
the flow of political infbrmation.

Additional support for this hypothesis can be derived from

the findings of two additional inquiries into southern electoral

34

behavior.1 In an analysis of the electoral support given third party
candidate George Wallace in the 1968 presidential election, S. Birdsall
(1969) identified a pattern of "lower leve1(s) of Wallace vote given

by counties with large cities." In an analysis which concentrated on
the Wallace vote in cities of the southeast, 8. D. Brunn (1970) pointed
out that although the wallace vote was generally lower in larger cities,
there was a great deal of variation depending on whether the city was
in the "core" or peripheral region of the southeast. Thus, in each
instance we note a general negative relationship between size and level
of support for wallace. If we view the wallace movement as a form of
change or at least as a significant deviation from established (two
party) behavior, we gain additional support for our thesis that a nega-
tive relationship exists between change and urban size at a regional
level.

In an earlier discussion of the growth of Republican electoral
support in the South, it was demonstrated that the medium of Republi-
can growth was the urban or metropolitan structure of the South. Yet
the examples of the Social Credit League in New Zealand (Barnett, 1973),
Democratic voters in Ohio (Cox, 1971) and even the two examples of
Wallace support in the southeast (Birdsall, 1969; Brunn, 1970) suggest
the opposite--a negative relationship. If the two sets of literature

appear at odds, they are not. As Barnett discovered in the example

 

1Obviously these two studies are not indicative of electoral
change in a longitudinal connotation since only one election is con-
sidered. However, even though the focus is not on electoral change,
such studies nonetheless offer valuable insight into the nature of the
relationship between urban size and patterns of electoral response.

35

of the social credit League in New Zealand (1972) and again in the
example of the Communist party in Denmark:
. . . political forces, operating at a number of levels,

influenced the spatial pattern of adoptions in the dif-
fusion of a new political movement (Barnett, 1973, 42).

In the case of the Danish Communist party a

within-constituency effect, possibly a manisfestation of
the hierarchical diffusion and spatial concentration of
Communist support in larger urban areas "accounted for"
most of the variation in the early election series
(Barnett, 1973, 42).
Thus, the pattern of electoral change must be viewed as a function of
the level and spatial context in which the political movement occurs.
Once we consider the nature of the South as the spatial context in
which we are to assess changes in the structure of regional patterns
of partisan alignment, more of the apparent dichotomy can be resolved.
In terms of the relationship between partisan alignment and urban
structure, the preponderance of Republican sentiment in villages, towns
and smaller cities outside the South is certainly a well-documented
phenomenon in voting statistics and literature. Consider two specific
inquiries into the nature of this relationship. In an examination of
the relationship between the size of a place and partisan alignment,
L. D. Epstein (1956) noted that Democratic strength in Wisconsin
gubernatorial elections diminished sharply as the size of the city
declined. After a failure to substantiate this pattern in Michigan
(Masters and wright, 1958), Epstein's analysis was duplicated by

D. Adamany (1964) using subsequent Wisconsin elections. Adamany dis-

covered that "urbanized areas continued to vote strongly

36

Democratic . . ." while ". . . small cities and villages provided over-
whelming Republican margins . . ." (Adamany, 1964, 486). As in
Epstein's analysis, Adamany discovered the major breaking point between
Republican and Democratic support was at the 50,000 population size
level. But most importantly, at least in terms of its value to this
inquiry, Adamany discovered no apparent change in the relationship
between size of place and vote response. Thus, despite a number of
significant Democratic party victories at both the senatorial and
gubernatorial level during the 1958-1962 period, the relationship had
remained the same as in the 1948-1954 period of Republican dominance.

These two studies serve as an empirical verification of the
nature of the relationship between partisan alignment and urban struc-
ture in an American political context. In Wisconsin at least the
Republican proportion of the two party vote diminished as city size in-
creased. Such a relationship coincides with established precepts of
partisan alignment in the United States.

Quite obviously it is not possible to transfer the notion of
partisan alignment derived in these two Wisconsin studies to a southern
electoral inquiry today. The South does not fit the national pattern
of partisan alignment. It may eventually achieve a national "norm,"
but at present it is predominately Democratic in partisan alignment
at any level of the urban spectrum. But despite the inappropriateness
of national partisan alignment concepts, it is still possible to come
pare the patterns of changing electoral response to the urban structure
in both Wisconsin and in our southern example. In the South the
Republican party is gaining electoral support and is apparently doing

so more rapidly in larger urban places than in smaller. In Wisconsin

37

the Republican party dominated electoral politics. The medium of
electoral change was also the urban structure and greater success comes
in larger urban places. It can be argued that this is a natural en-
vironment for the Democratic party and that the Republican party in

the South is operating in an unnatural environment. Electoral change
should be rural centered in the South since it is there that Republicans
are traditionally strongest and it is also there that the interpersonal
transfer of information is accentuated. But the South does not fit

the national norm. And it might be argued equally well that small
towns and villages are also the more recalcitrant when it comes to non—
agrarian political change. If the presence of Republican support in an
urban environment still seems at odds with the accepted impressions of
locational bases of support for the Republican party, perhaps it would
be well to recall that the Republican party is seen as a logical al-
ternative to the traditional southern Democratic conservatism (Heard,
1952; Strong, 1963; Cosman, 1966b; Phillips, 1970). In light of the
failure of third party movements, it is the ggly_alternative just as
was the case in Wisconsin. wallace's successes in the rural environ-
ment and his diminished support in cities can be interpreted as both

a last gasp at a third party conservative alternative and also as a
portent of future partisan alignment in urban areas. Finally, it is
possible to argue that the metropolitan environment is a natural focus
of new ideas and new information. Although it may not be transmitted
as quickly, the natural anonymity of the neighborhood and the greater
diversity and sophistication of urban life make new ideas more readily

acceptable.

38

Thus, although one set of literature may speak of urban related
change, while another links such change to more rural environments, the
two do not appear in reality to be at odds. The political levels and
environments vary; thus, the results must be interpreted in different
context. And finally, in dealing with the southern political experience,
it would be well to remember that if the South does not conform to an
established theoretical norm, the norm need not necessarily be brought
to question. For the South has failed to conform to any norm but

its own for over a century.

Objectives and Hypotheses1 Derived from the Literature

The literature has provided a foundation from which the problem
might be more concisely defined. Several possible avenues of inquiry
are suggested by inconsistencies, omissions or simply from questions
left unanswered in previous research. But, before any avenue of in-
quiry can be pursued, we must first precisely and carefully delimit
the nature of the electoral support for the Republican party. Thus,
the overall objective of this study becomes:

Objective 1: The spatial and numerical delimitation and
measurement of the electoral support for the
Republican party at the presidential and

senatorial levels of competition in a sample

 

1Although the term hypotheses is used throughout this section, no
specific connotation of statistical testing can be assigned to most of
these hypotheses. They are more akin to "expectations" than hypotheses.
Although descriptive statistics are utilized in Objectives 1 and 2 and
some nonparametric statistical tests are applied in Objective 3, the
general nature of the data and research design did not dictate, nor in
some cases permit, the implementation of strict probabilistic criteria
for acception or rejection of "hypotheses."

Rationale:

Hypothesis 1 :

39

region of counties in the American South

from 1948 to 1972.

In order to treat changing patterns of elec—
toral support, it is first essential that both
the location and the intensity of that support
be accurately delimited for all elections
under consideration. Only by accomplishing
this can we hope to describe the pattern of
support and the nature of the change that
occurs. Two levels of electoral competition
are considered since the literature suggests
ticket splitting may be operative within the
electorate. It is expected that there will

be a time lag evident between the level of
electoral support for the GOP at the presi-
dential level and the level of support at

the senatorial level. Thus, it is hypothesized
that:

For each set of elections under considera-
tion--l948 presidential and senatorial; 1952
and 1956 presidential and 1954 senatorial;
1960 presidential and senatorial; 1964 and
1968 presidential and 1966 senatorial; and
the 1972 presidential and senatorial--the
electoral support for the Republican party
will be greater at the presidential level of

competition. Although the level of electoral

Hypothesis 2:

40

support for the Republican party will be con-
sistently higher at the presidential level,
the senatorial support will demonstrate more
stable patterns of response. Thus, it is
further hypothesized that:

The senatorial level of competition will
produce more stable electoral response pat-
terns than the presidential level of compe-

tition.

Having described as precisely as is feasible the spatial and

numerical character of electoral support, it is then possible to turn

to the identification of such spatial and numerical regularities as

exist in the voting response surfaces under consideration. Thus, a

second objective becomes:

Objective 2:

Rationale:

The determination of such spatial regularities
as are associated with the increase in elec-
toral support for the Republican party.
Accurate description of the structure of
voting response surfaces provides a foundation
for consideration of shifts in locational
bases of support for the Republican party.

And in describing the change that occurs in
patterns of support the regularity in these
shifts can be explored for process linkages.
Through the use of specific computer programs
which summarize the spatial properties of

areal data, the distributional characteristics

41

of the electoral support for the Republican
party will be traced from one election to the

next .

It is expected that the numerical, statistical and spatial de-

scription of successive electoral response patterns will reveal a strong

indication of less areal concentration in patterns of electoral support

in each succeeding election. Thus, it is hypothesized that:

Hypothesis 3:

The numerical description of the spatial
distribution of electoral support will reveal
a strong tendency toward more evenly distrib-
uted spatial patterns of support, i.e., less
geographical concentration in areas such as
Appalachia and more widespread locational

bases of electoral suppert.

Again it is expected that such a widening of geographical bases

of electoral support for the Republican party will vary by level of

electoral competition. Thus, it is hypothesized that:

Hypothesis 4:

The tendency toward a more widespread dis-
tribution of electoral support will vary by
level of electoral competition with the
broadening base of locational support be-
coming more apparent at the presidential
level of competition at an earlier date than

at the senatorial level.

Having accomplished the first two objectives, it is then possible

to turn now to the principal objective of this inquiry:

42

Objective 3: The description and accurate measurement of
the nature of the relationship between urban
size and the growth of support for a political
party.

Rationale: Once we have described the structure of the
voting response surfaces (Objectives 1 and 2),
it becomes possible to attempt a description
of the process by which change is occurring.
Since the literature strongly suggests that
this change is an urban centered phenomenon,
this is the thesis that will be examined.

In doing so, specific comments on the nature
of the relationship between urban size and
changing patterns of voting response can be
made. It is expected that the higher levels
of Republican support will be found first in
counties with larger urban populations. Thus,
it is hypothesized:

Hypothesis 5: A strong positive relationship exists between
the level or percent of urban population and
intensity of Republican support.

In connection with Objective 3, the notion of a cutoff level or
breaking point in the size of population of a county that supports the
Republican party will be examined. Specifically the following questions
will be explored:

1) Can a breaking point in the level of Republican support

be identified?

2)

3)

43

If so, is the breaking point consistent at both levels

of electoral competition?

With more recent elections does the breaking point

move down the scale of population size?

CHAPTER III

THE APPROACH TO THE RESEARCH PROBLEM

In the first two chapters the prdblem has been discussed in
terms of its conceptual and literature foundations. These chapters
have provided the limits, goals and specific expectations and directions
of the problem. It remains now to chart the manner in which these
objectives can be accomplished. For, as in any research effort, certain
fundamental decisions concerning the nature, scope and level of analysis
are made, which will affect the end product. Understanding why and
how these decisions are affected is crucial to grasping the nature of
the results. This is the focus of this chapter. Specific attention is
given to the role of aggregate analysis in electoral research, to the
rationale and procedure employed in limiting the electoral and spatial
context of the problem, to the nature and form of the data, and to the
specific statistical, numerical, cartographic and spatial techniques

employed in the analysis.

Aggregate Analysis in Electoral Research

Aggregative analysis has severe limitations. . . . and
those who attempt to explain the flow of the vote solely
on the basis of the statistics which the election apparatus
makes available labor at a great disadvantage (Campbell,
Converse, Miller and Stokes, 1967, l).

44

45

The question of survey versus aggregate level analysis in research
is one of the more enduring debates in social science. And, as the
quotation above indicates, electoral research is not exempt from such
discourse. The question is introduced here primarily because the level
of analysis in this inquiry would appear to be resisting the tide of
survey research so much in increasing evidence in electoral research.

The arguments against aggregate level analysis are well known
and appreciated. Campbell, Converse, Miller and Stokes (1967) have
summarized some of the disadvantages succinctly: aggregative analysis
permits only an indirect approach to analysis; the data cannot be re-
lated directly to individual electors; surrogate measures must be
utilized to relate explanation to behavior; alternative hypotheses
often are not testable since additional data sets are not readily
available; and finally perhaps the most severe criticism, that being
the temptation to misinterpret aggregate level behavior as an indica-
tion of individual behavior.

The arguments in favor of aggregate analysis generally assume a
position somewhat akin to one an efficiency analyst might take--aggre-
gate data cost less to collect, are more plentiful and are thus more
readily accessible. Even though such arguments were important to this
study, aggregate level analysis is employed more for reasons of dis-
ciplinary orientation and context of the problem. That is, the problem
involves the spatial analytic approach, and as such, requires the use of
a spatially or locationally referenced data set. Such data are in-
frequent and difficult to obtain especially if the inquiry involves
longitudinal rather than temporally specific analysis. For the geog-

rapher one limitation on individual level analysis is the virtual absence

46

of longitudinal data sets for purposes of cross-temporal comparison.
With regards survey data, geographers are in much the same position as
political scientists in the pre-Survey Research Center era. There exists
no substantive body of survey data which provides adequate spatial
information for longitudinal analysis.
The argument of non-availability does not deny the need for
survey level analyses of spatial electoral behavior. For as Cox has
suggested, it is possible to relate the voting decision of an individual
to a location in an information flow network by:
l) the identification of spatial regularities in voting
response surfaces, or
2) the verification of spatio-temporal processes which will
one day be synthesized into a spatial simulation model
(Cox, 1969, 113).
If spatial elements, such as distance,direction, relative locations
and density functions, are components of the overall model of the voting
decision in its spatial milieu, then it is certainly feasible that one
might identify the dynamic elements of the voting decision in its
spatial context. This can be accomplished by working at the level of
individual voting decisions and the relation to the flow of information
within a spatial information network. But Cox also suggests that the
aggregate level of analysis should not be ignored in our search for
spatial regularity in voting response. As long as we remain satisfied
with general statements of spatial regularity in voting patterns and no
attempt is made to transfer the results to explanations of individual
behavior, aggregate level analysis can prove very productive. For

example, statements pertaining to the nature of voting responses of the

47

overall study area lie within the constraints of the aggregate level
data employed in this inquiry; yet statements relating to individual
voter responses exceed these data constraints. Thus, in dealing with
data pertaining to aggregates, any attempt to transfer relationships
between those aggregates to the individual level invites misinterpre-
tation. But in light of the theoretical and developmental status of
electoral research in geography, perhaps at this point it is not neces-
sarily the individual who need be our primary focus. Attitudinal changes
are important in the explanation of individual voting responses. To
explain change at the aggregate level and to deal with changes in the
spatial regularity of voting response surfaces, we need not cope with
individual attitudinal change so long as the generalizations made are
confined to aggregates. For geographers there may come a time when
knowledge of individual attitudinal change of southern voters may be-
come essential to the continued development of electoral theory but in
terms of addressing the question of spatial dynamics of voting responses,
that time has not arrived. we know little of the spatial dynamic
elements of voting response and additional empirical evidence at the
aggregate level can well provide the foundation for continued inquiry

at the individual level.

The Scope of the Problem
Consideration of electoral change in the South does not neces-
sarily imply that the entire region serves as the spatial context, nor
that all levels of electoral competition be represented. Quite the
contrary, the procedural nightmare that would ensue from attempting

to manage such volumes of data would offset any advantages that might

48

accrue. Through the implementation of sound sampling procedure, it is
quite possible to limit the spatial and electoral context, and still
accomplish an effective description of the regional patterns of elec-
toral behavior, 229 an elaboration on the nature of electoral change and
the attendant spatial mechanisms inherent in such change. In limiting
the size of the study area and the number of elections, the scope of

the research problem is being tailored to meet meaningful expectations
and goals. The rationale for limiting the scope of the problem is

provided below.

The Spatial Context
Units of analysis

The basic unit of analysis employed in this study is the county.
For purposes of spatial analysis covering several contiguous states
the county represents the most logical selection. The state is unac-
ceptable since the degree of aggregation usually masks significant
internal variations. Congressional districts are inadequate for
longitudinal analysis since their boundaries fluctuate with reappor-
tionment. Smaller political divisions, such as wards, or precincts,
are far too cumbersome for use in regional level analysis. Thus, in
terms of the cost involved in both time and effort, the county level
of analysis represents an adequate compromise since there is more elec-
toral detail than at the state level and far less data manipulation
than at the lower levels of analysis.

The county does offer one important advantage--the availability
of census information, economic indicators and a host of other types

of data. While such information may be available for more detailed units

49

than county, the collection units employed at lower levels often do not
correspond to the ward and precinct boundaries employed in election ad-
ministration. Thus, in county level analysis no problems of data com-

parability are encountered should additional social or economic data

be needed.

Delimitation of the study area

In the 11 states of the former Confederacy1 there were 1,139
county or county equivalents in 1970.2 If all the counties in the South
were employed in this analysis, there would be an enormous, and un-
necessary, expenditure of time, effort and money in sheer data manipu-
lation involved in satisfying the objectives of this study. Given the
detailed level of analysis intended for the inquiry,such volume would
prove vastly restrictive since it would effectively preclude the incor-
poration of some numerical and spatial techniques deemed useful.
Additional problems would also arise in the case of Virginia where in-
dependent cities have a tendency to appear and disappear in the course
of the 24 year study period.

With such restrictions in mind, it was decided to employ only a
portion or a representative study area in the actual analysis. Al-
though a sample study area may limit the comments which can be directed
toward regional electoral behavior, the corresponding advantages ac-
crued in analysis of the spatial dynamics of electoral change more than

compensates. The larger the number of observations the greater the

 

1The definition of the South employed in this analysis.

2Virginia has independent cities which are the equivalent of

counties.

50

effort necessary to describe the resulting electoral response patterns.
By curtailing the number of observations, more effort can be devoted
to detailed analysis and less to manipulation of data.

In order to assure that the areal sample of counties utilized
was a valid representation of the entire region, specific selection
criteria were established. In order of their importance, these criteria
are:

l) The size of the sample must be such that adequate
coverage of the region (the South) is possible yet of
sufficient size that the volume of data can be managed.
Although no specific areal size is necessary, a minimum
of 25 percent, or 285 counties, is employed.

2) The sample must be drawn from at least six of the
eleven states of the South.

3) The sample must present a reasonable balance between
Deep South and Rim South states.1

4) The sample counties must be contiguous.

5) The shape of resulting study area must be as regular2

as feasible.

 

1The Deep South states-~Louisiana, Mississippi, Alabama, Georgia
and South Carolina--have been found to have a slightly different polit-
ical and social base than the Rim South counterparts (Cosman, 1966b;
Birdsall, 1969; Phillips, 1969; Brunn, 1970; Brunn and Ingalls, 1972)
of Virginia, Tennessee, North Carolina, Texas, Arkansas and Florida.

2Regular implies that the study area contains no outstanding
profusions, such as the peninsular of Florida might create if the entire
South were employed.

51

The first three criteria have to do with reducing the size of
the study area to a manageable areal sample while maintaining its
regional validity. The last two criteria are incorporated for use of
specific spatial analytical techniques that will extend the descriptive
capacity beyond mere visual interpretation of voting patterns. In
these techniques, described in the following section, contiguity and an
approximately rectangular study area are beneficial. By using such
techniques, it is possible to pinpoint more specifically the spatial
variation in electoral response patterns over time.

It was found that a sample size of 448 counties (approximately
40 percent) satisfied all five criteria. The outline of the study
area is shown in Figure 1 and the 448 counties are listed in Appendix 1.
The study area includes counties from three Rim South states and four
Deep South states. It provides valid representation of the South since
it includes: the historical core of Republican support--Appalachia;
ample diversity of county population size; cultural and economic dif-
ferences--Black belt agricultural area to industrial zones; and finally
a reasonable number of various sized cities. Thus, the sample appears
to offer a viable compromise of adequate coverage, adequate size and

the desired shape and contiguity.

The Electoral Context

Having established the size and areal extent of the study area,
we can now turn to the variables employed in the analysis of electoral
change. The nature of this inquiry dictates that most of the variables
will be data depicting the outcome of elections, or more accurately the

figures representing voting response for Republican party candidates.

52

 

 

FIGURE I.

STUDY AREA

  
   
 

  

\ ‘ .
guesthIIs\\ \ '\ .

\ :- ’I‘k’i“ ‘ \ xvi) '
.\ \\ \E’; . \.\‘x~K"-° #225,,
‘ \l‘T NN , .2 X _.

MISS.

 

“\V Study area
m S.M.S.A.

9 .00 29° Boo
MILE.

 

 

53

Hence, decisions on the specific variables employed involve delimiting

the electoral context of the research problem.

The levels of analysis

Since this study is concerned with an entire region, it is nec-
essary to limit the analysis to those elections which involve at the
very minimum statewide electoral response patterns. Electoral patterns
derived from competition for non-statewide political offices are likely
to be subject to increased interference from very localized influences.
These influences represent "noise" in the overall patterns of electoral
response, and as such, increase the number of explanatory factors which
must be considered. As Reynolds (1969) and V. 0. Key (1949) discovered,
such factors as candidate location, home town and distance to nearest
competing candidate can play important roles even in state level elec-
tions and their effect at lower levels of competition is likely to
increase. Even though statewide elections are by no means immune to
such "localized noise" they are at least less subject to significant
variation. In statewide elections, we are dealing with only one response
surface not several. Thus, in order to reduce the number of extraneous
or uncontrollable factors involved in the regional level analysis, it
was decided to employ only those elections which involve statewide
voting. Of course elected offices such as states attorney and secretary
of state involve statewide voting response. But such offices do not
reflect very well the feeling of regional identity with which we are
concerned,since they can and often do involve candidates of little
regional,much less national prominence. Thus, in this analysis we shall

consider only those levels of electoral competition which retain a

54

measure of regional 22g_national identity. Given this additional re-
striction the selection of elections for analysis must be made from
presidential, senatorial and gubernatorial levels of competition.

Why not all three levels? First, because of the very nature of
gubernatorial elections in the South, there are far more "no data"1
elections than at the other two levels. Second, senatorial candidates
are in a sense national as well as state level candidates. With
gubernatorial candidates the element of national appeal is reduced.

But then why consider presidential elections at all? Why not
consider only senatorial elections? One reason lies in the references
that can be made to the literature. The literature on Republican elec-
toral inroads in the South has focused at the presidential level. By
employing presidential elections, the results can be cross checked
with past research effbrts even though they are non-geographic in
orientation. In addition, the presidential level is the only competi-
tive office where one is assured of a Republican candidate being on the
ballot in every election. Thus, it is possible to obtain excellent
longitudinal data coverage.

But the overriding reason for including two levels of elections

lies in the nature of presidential election response patterns in both a

 

In a southern electoral context one becomes readily familiar
with the limitation of an absence of electoral statistics. Such absence
comes less as a consequence of non-availability than of non-competition
since in a one party system, such as existed in the South until com-
paratively recently, second party challenges were, and still are, not
always a certainty. Georgia, for example,had no Republican candidate
for senator or governor from 1948 to 1966. It proves exceedingly
difficult to analyze voting response when there is no stimulus.

55

southern and national context. If we accept the theory of a normal
partisan alignment in the United States, in any presidential election
there is likely to be a healthy Democrat proportion of 53 percent of the
total two party vote. Thus, any normal election should be a Democratic
victory (Campbell, Converse, Miller and Stokes, 1966). Consider the
last seven presidential elections. In two, 1948 and 1968, there were
Ehggg_significant parties. In four of these elections there were

Republican not Democratic victories. And in two, 1964 and 1972, there

 

were such landslide victories as to set aside any notion of a "normal"
election. The important point here is the recent volatility of the
presidential level of competition as far as the voting analyst is
concerned. The results are so fluid that extreme caution must be ex-
ercised in analysis conducted at this level.

To such criticism of presidential election data one must also
add the particularly southern phenomenon of recent presidential Repub-
licanism. This refers to the practice of ticket-splitting practiced by
southern Democrats in which a voter may choose to support the GOP at
only one level, the presidential level. At all other levels the voter
may, and often does, remain Democratic (V. 0. Key, 1949; Cosman, 1962).
Although the recent inroads made by Republicans at lower levels of
electoral competition suggest this phenomenon may be ebbing, it is still
necessary to provide a means of testing for its presence.

Given both the volatility and ticket-splitting aspects inherent
in recent presidential elections, consider the ramifications of not
using additional levels of analysis. It is feasible that the successes

of Republican party candidates at the presidential level may give an

56

inflated assessment of the true status of the party at other levels of
competition. Conversely, a poor showing at the presidential level,

or a third party candidate siphoning off potential voters may mask
more successful competition at lower levels.1 Thus, by expanding

the analysis from presidential level competition to lower levels of
competition the impact of such volatility is reduced. And perhaps a

truer assessment of partisan preference is made possible.

The longitudinal extent of the analysis

Once the decision on the level of analysis is made, it only re-
mains to determine which actual elections will be used. To accomplish
this task a tentative time period was first established with 1948
as the beginning of the period of analysis and 1972 as the closing date.
The beginning date, 1948, is based on the literature as the generally
acknowledged beginning of the erosion of Democratic solidarity. It
marks the first major break of the South with the Democratic party,
this coming as a result of the walkout of southern delegates from the
1948 convention due to a dissatisfaction with Truman and his civil
rights legislation. In addition this date marks the first major third
party movement in the South since the populists. For these reasons,
1948 is taken as the beginning of electoral change in the South, and
hence as the beginning date for analysis of the study area.

Deciding which elections to use at the presidential level presents

no problem since all elections between and including 1948 and 1972

 

1Throughout this inquiry senatorial elections will be referred
to as lower levels of electoral competition.

57

are used. Simplicity is not the rule for the senatorial level, however,
since not all states elect senators in the same six-year sequence. If
all senatorial elections held between 1948 and 1972 are used, the result
is a two-year interval sequence which creates a number of "no data"
situations. For example, in 1952 only one of the seven states under
consideration had a senatorial election; while in 1968 five of seven
had senatorial elections. Thus, an attempt was made to find the time
span interval which was most efficient in eliminating "no data" situa-
tions. It was discovered that if six year intervals were used between
the years for which data would be collected, the "no data" situations
were minimized. Thus, if we begin at 1948, the dates for which sena-
torial data would be collected would be 1948, 1954, 1960, 1966 and 1972.
But even this interval still produces situations in which no elections
were held in some states. In an effort to further reduce the "no data"
situations, it was decided that data from elections held within two years
of an analysis date (1948, 1954, 1960, 1966 and 1972) could be substi-
tuted where a state had held no election during one of the five years
chosen. Thus, if South Carolina had no Republican candidate or held no
election in 1960, but did have a GOP candidate in the 1962 senatorial
election, then the results of the 1962 election would be substituted
for the "no data" situation in 1960. The result of employing this pro-
cedure was that in only 5 of 35 times1 were no data available. In
these five cases, no data are treated as zero level of support for the

Republican party. Table 3 provides a listing of the elections employed

 

1Three of these five no data situations develop in Georgia elec-
tions, 1948-1964, and the two others develop in Florida (1954) and
Mississippi (1948) elections.

58

.uocuse an coHusHHmsou "mumaom

.ousowocso cmoHHosmom 02:;

 

 

 

 

 

Name on: News

some was some

23 2 2 2 2 2 mum 2 3.2 was $3

33 so 8 so 3 Mm mm mm 83 $2 $2

83 8 Mm om 8 I W 8 83 one 33

«m2 em elm v... 3 I .i am 22 one 23

I

meme we as me u- .. om we mva «up meme

H33 Hueuouocom E. 8 oz ms 8 «a .2 29 um oz ms 5 mm .2 .253
um nHmmHsso HuHuouscom HMHucoonoum HeHucoonoum
cH oomonEm mHo>oq N uunanumHmcd cH owns mounuum coHuoon now Hm>uousH

3635 we? «see

 

mmqudz< ZH Gums mZOHBUMHm

m mqmda

S9

in this analysis. Appendix 2 provides a list of the candidates in
each of these elections.

Such a procedure creates some problems of comparability, since
all senatorial voting response surfaces are composed of state level
elections, some of which were held two years apart. And in the case
of the 1948, 1960 and 1972 dates, additional difficulties arise since
the effects of presidential candidacies must be allowed for some, but
not all, parts of the voting response surface. But any attempt at
regional or national level electoral analysis, at any level except
presidential competition, encounters similar difficulties. The alter-
native--no analysis--is too high a price to pay. Rather this procedure
is adopted and special care is taken to note any variation which might

result from bandwagon influences or from the short time lag.

The Data
Data sources

Once the spatial and electoral context of the inquiry were de-
limited, the data were collected. Data for elections held from 1952
to 1970 were obtained from two sources: 1) Richard Scammon's America

at the Polls (1956-1972), a nine volume set of county election sta-

 

tistics, and 2) the Institute of Social Research in Ann Arbor, Michi-
gan. Data for the periods 1948 to 1952 and 1970 to 1972 were obtained
from a variety of sources including the two mentioned above, various
state Republican committees, the national Republican committee, and
various secretaries of state. Wherever possible data were gathered
from two sources and selectively cross checked for validity and re-

liability. Data pertaining to urban population were taken from the

60

County and City Data Book (1956 and 1962) and from the 1970 Census of

Population for respective states.

Characteristics of the data

The election data used in the analysis were coded in the form of
a percentage of the total vote for the Republican candidate in each
election under consideration and in each county in the study area. The
percentages values are rounded to the nearest one-tenth percentage
point. Data for population size were coded by county for both the
absolute population size and the percentage of the total population that
is classified as urban by the Bureau of Census.1 Absolute figures are
given to the nearest whole number; percentage urban data are rounded
to the nearest one-tenth percent.

The frequency distribution of each variable employed was tested
for correspondence with a normal distribution via the Kolmogorov-
Smirnov test for normality.2 Of the 15 variables (the percentage of
Republican vote in 12 elections and the percentage of urban population
in 1950, 1960 and 1970) employed in this analysis in only two cases,
the 1964 presidential and the 1972 senatorial results, were the dis-
tributions normal at the .15 level of significance. For each variable
not found to be normally distributed, two transformations, squaring the

data and log transformation, were performed on each variable in an

 

lAny population grouping larger than 2,500 is considered urban
by the Bureau of Census.

2A program entitled NORM (Wittick, 1971) written by Theodore
Miller, Department of Geography, University of Iowa, was employed in
these tests. The program employs a Kolmogorov-Smirnov Test, two
tailed, to test for normality.

61

attempt to render it normal. But in only two cases, the 1952 and 1968
presidential election returns, did such transformation produce a normal
distribution. The data were thus left in the original percentage

form and the only revision made was the computation of z scores for
map pattern comparisons. Thus, for all numerical and statistical
purposes the data must be considered as deviating from a normal dis-

tribution.

The Method of Analysis

Basically the analysis of the data involves devising a procedure
for testing the validity of the hypothesis (or expectations) estab-
lished in Chapter II. The procedure used in this inquiry can be divided
into two stages-~ana1ysis of structure and analysis of process. The
first stage--structural analysis--consists mainly of a description of
the structure of successive voting response surfaces. In this case
both numerical and visual forms of description are employed. Numerical
description, that is means and standard deviations, is employed since
it permits a more precise delimitation of the nature and spatial ex-
tent of electoral support for the Republican party. Such accuracy is
extremely useful in the second phase of the analysis--the description
of the process of electoral change. Since, in this case, process is
hypothesized to be linked to urban structure, the analytical procedure

employs techniques which test this thesis.

Analysis of Structure
Describing the spatial structure of voting response surfaces
traditionally has consisted of verbal description and visual impres-

sions of electoral response surfaces. While such descriptions are

62

unavoidable in structural analysis, there are some serious limitations
involved. Descriptions thus derived are essentially personal impres-
sions of the analyst, and, as such, they may be subject to as many
varying interpretations as there are analysts. In addition, the patterns
of the voting responses themselves can vary depending on the choice

of map scale, style and, most importantly, the data intervals used

to prepare maps. Such deficiencies dictate the necessity of incorpo-

rating additional descriptive aids in the structural analysis.

Numerical description of the data set

In this inquiry visual pattern analysis is employed. However, it
is supplemented wherever feasible with numerical description of the data.
Such evidence lends additional support to structural analysis since the
interpretation of voting response patterns is based not only on visual
impression and personal expertise of the analyst, but also on evidence
derived from consideration of the numerical and statistical character-
istics of the data set. For example, consider two such statistics which
are derived from the distribution of the data and which permit the
analyst to assess the performance of individual constituencies or
counties against others. These statistics--the mean and the standard
deviation--measure the average (mean) performance and the extent to
which each constituency deviates from the average (standard deviation).
Thus, if we wish to assess the performance of the Republican party over
successive elections, we can obtain a convenient summary of the sta-
tistical distribution of votes by deriving these two statistics for each
election. As Lewis and Skipworth (1966) indicated in their analysis

of British parliamentary elections, the mean vote and the standard

63

deviation provide a measure of compactness in the distribution of the
vote over successive elections. The more uniform the distribution,
i.e., the less the tendency toward high or low percentages, the smaller
the standard deviation and interestingly enough the more successful

the party (Lewis and Skipworth, 1966, 3).

Adapting this notion of uniformity in distribution to the elec-
toral performance of the Republican party, it might thus be applied as
an indication of growth of support. Since the data represent the pro-
portion of the total vote that went to Republican candidates, increases
in these percentages with increasingly more recent elections would
point to greater success. Furthermore, a corresponding decrease in
the standard deviation of each election with the passage of time would
indicate more uniform or compact distributions of vote. Such uniformity
would indicate that the areas of very high and very low support indica-
tive of earlier electoral competition were disappearing as the party
became increasingly more competitive and more successful. In other
words, the distribution of the vote is expected to become more compact
from 1948 to 1972; and as indication of this increasing uniformity
the standard deviation computed for each election will decrease. Of
course, as hypothesized in Chapter II, the means and standard deviations
of senatorial vote distributions are expected to demonstrate greater
stability than those of presidential vote distribution.

Just as such statistics readily provide a means of describing
numerical distribution, they can be employed in spatial descriptions
as well. Consider a pattern of voting response represented through the
use of the percentage of the total vote gained by the Republican candi-

date in each county or constituency in each election. If we were to

64

transfer these figures (percentages) to maps of the study area, we
would obtain a series of maps of the spatial distribution of voting
responses. However, these percentage maps do not provide any indication
of how the performance of an individual constituency measures up to
the overall performance or the performance of all others. But, if the
standard deviation of each constituency was the basis of the map pat-
terns, the map would demonstrate the performance of each unit based on
the mean of the total. Maps based on standard deviation units would
thus show the relative strength of the Republican party in each county
based upon the showing of the party in the total sample study area.
As such, the maps would not reflect constituencies won or lost (though
this may be computed rapidly from the mean) but they would allow state-
ments to be made about the strength or allegiance of individual con-
stituencies to the party. Those counties nearest the mean would be
considered "typical" in their response patterns for that election.
Those furtherest from the mean would be either heavy supporters of the
Republican party (positive standard deviations) or greatest opposition
(negative standard deviations).

Often when choropleth maps1 are drawn from such data as percentage

votes cast for one party or another, the categories used to portray

 

An alternative form of map presentation would have been isarithmic
mapping. But isarithmic mapping involves a considerable degree of
interpolation and generalization of data. Consequently, this form of
presentation would restrict comparison of maps to general areas of the
voting surface rather than specific political units. The advantages
gained in comparability and cross-constituency comparison via the use
of central tendency measures would be lost to generalization. Choro-
pleth mapping permits full utilization of individual performances,
if so desired.

65

patterns of response are chosen quite arbitrarily. Standard deviation
units can be used to define map categories (intervals) more objectively.
For example, a middle or average category can be established by one-
half standard deviation units above and below the mean (average) value
of the distribution. Subsequent categories can be defined in one
standard deviation unit on either side of the mean value. Categories
thus derived are based upon parameters that accurately and meaningfully
describe the statistical data to be mapped.

But even standard deviation units can be improved upon as a basis
for mapping statistical data. By using the mean as an origin and the
standard deviation as a means of standardization, it is possible to
calculate another descriptive statistic for each observation called a
z score. The Z score is simply another method of comparing how far
the various observations deviate from the mean; yet the Z score pos-
sesses the advantages of locating the observation by taking into account
the Eggal variability of the distribution, as well as being a standard-
ized measure. Hence, it was decided to employ Z score values in mapping
the county results of each election. The same criteria for selection
of interval or category size can be applied to Z score values as were
applied to standard deviation units. With Z score values the mean is 0
and a standard deviation is measured as +1.0 or -1.0. Allowing for a
mean range of 1.0 Z scores ($0.5 Z scores on each side of 0) about the
mean of 0 and increments of 1.0 Z scores the interval for mapping be-

comes :

66

Heavy <-2.500
REPUBLICAN OPPOSITION Moderate -2.500 to -l.501
Light -l.500 to -0.501
Mean Range -O.500 to +0.500
Light +0.501 to +1.500
REPUBLICAN SUPPORT Moderate +1.501 to +2.500
Heavy >+2.500

To facilitate discussion of map patterns negative Z score values will
be discussed as opposition and positive values as support, with degrees

of each as represented above.1

Numerical description of the spatial characteristics of the data set
Maps prepared from statistics descriptive of the total distribu-
tion of the data set permit the analyst to make visual assessment of the
location and distribution of electoral support for a political party.
But interpretation of such maps involves both visual perusal and sub-
jective assessment. As previously argued this procedure is highly
variable and rather imprecise. Greater accuracy, reproducibility and

objectivity can be achieved through the utilization of numerical

 

1Since the data, except for one variable, do not conform to a
normal distribution, probabilistic statements and assumptions cannot be
applied in the use of these statistics. Thus, it cannot be assumed that
68 percent of the Observations fall within :1 standard deviation range
of the §'(mean), 95.5 percent in +2 standard deviation of the i'range and
so forth. It is possible to make use of i, standard deviations and Z
scores in spite of deviation of the distribution from normality. If
the mean is 10.0 and the standard deviation is 20.0 and the Republican
party receives +50 percent of the two party vote in a county then that
performance lies +2 standard deviations from the mean. However, be-
cause of the normalcy question, it cannot be assured that 95 percent
of the other counties do as well.

67

description of the spatial characteristics of a data set. Such
numerical description of voting response surfaces can be achieved by
attending to the spatial and locational network in which the response
is generated.

The intensity of Republican electoral support in each county in
each election is measured by one value-~a percentage of the two party
vote. This percentage is an average for the entire constituency. If
the constituency is viewed as a single point in space with this same
percentage value assigned to this point, it is possible to visualize
a distribution of n points in space, where n equals the total number
of counties. In order to distinguish between these n points a grid
referencing system of X and Y axes might be utilized to determine where
each county (constituency) point is located in space. Each county is
defined in terms of an X coordinate, a Y coordinate and a weighted 2
value (percentage). Once each point is so defined, it becomes possible
to derive a set of descriptive statistics similar to those utilized to
describe the numerical distribution of the data set; only in this case
these statistics describe the distribution of points in space weighted
by the voting response at those points. For example, a central location
or geographic mean can be calculated and as before the deviation of all
other points in space from this mean location can be derived. In addi-
tion, once the nature of the dispersion of points about the mean loca-
tion is known, the overall distributional orientation can be described.
If these statistics were calculated by weighting each with the percent-
age of Republican vote, they would effectively describe a voting re-
sponse surface spatially as well as numerically. By continuing such

measurement over successive elections, it is possible to utilize these

68

statistics to trace the nature of spatial changes in these point dis-
tributions or voting surfaces.

To enable such analysis of spatial distribution to be performed,
the distribution of county centers was described via a three digit set
of X and Y grid coordinates. Using this location grid system and the
percentage value for each election a set of descriptive statistics which
numerically describe the spatial characteristics of the data was cal-
culated for each election. The spatial characteristics of the point
distributions were described in terms of their dispersion, shape and
density.

The dispersion of the weighted points was described using a system
of concentric rings, defined in reference to a stationary base point,
to describe distance from that point. Direction from the center is
defined by dividing the study area into sectors using lines radiating
outward from the base point. By counting the points in each ring and
sector and summing the value of the weights in each, the average
weighted value for each ring and sector can be derived for an election.
In this fashion intensity of support can be pinpointed and change over
successive elections is traced, by direction and distance from a point
common to all voting response surfaces. This procedure used in con-
junction with the Z-score maps presents a more accurate measure of the
variation in the levels of Republican electoral support than simple
visual description and inspection.

Fortunately, this ring and sector count need not be attempted by
hand drafting and hand calculation procedures. The speed and accuracy

of a high speed digital computer can be utilized. A computer program

69

entitled LOCATEI (Wittick, 1973, 16-17) was utilized which employs

user-defined rings and sectors to describe geographic data.

In addition to describing the dispersion in the weighted point
distribution, it would also be useful to obtain a measure numerically
describing the shape of the distribution. To accomplish this a computer
program entitled CENTRO2 (Wittick, 1973, 18-19) was employed to cal-
culate centrographic measures which describes: l) the point distri-
bution in terms of its relationship to a linear or circular pattern;

2) the mean and deviation from that mean; and 3) the dispersion in
terms of standard reference axis calculated from the mean and standard
distance. The descriptive measures calculated by CENTRO that are
of specific use are:
1) the Mean Center . . . which is the equivalent of an
arithmetic mean of a univariate distribution.
2) the Standard distance . . . which describes the disper-
sion along a line passing through the mean center.
3) the Principal axes (major and minor) of the distribution
. . . describe the points at which the standard dis-
tance is at a minimum and maximum respectively.

4) the Angle of Rotation . . . the degree of rotation

necessary to minimize the standard distance.

 

1The original programmer of LOCATE was Duane F. Marble, Department
of Geography, Northwestern University. The LOCATE program used in this
analysis has been modified by Robert I. Wittick, Department of Geography,
Michigan State University for inclusion in GEOSYS, an information system
for the description and analysis of spatial data (Wittick, 1973, 16-17).

2 . .

The original programmer of CENTRO was John F. Hultquist, Depart-
ment of Geography, University of Iowa. The CENTRO program used in this
analysis was modified by Robert I. Wittick (1973, 18—19).

70

5) the Coefficient of Circularity . . . which measures

the degree of roundness of the distribution (Hultquist,

Holmes and Brown, n.d., 2-7).
The first four are useful in plotting a standard ellipse which describes
the path of standard distance-~one standard distance from the mean--
values as the axes are rotated 3600 about the mean center. The coeffi-
cient of circularity describes the shape from linear (0.0) to circular
(1.0) values.

The coefficient of circularity, mean center and the values which
describe the standard ellipse are all centrographic measures which are
employed to describe electoral change. Since these values describe
the shape of a distribution of points and since the distributions of
points depict voting response patterns, the results of successive
measurement by CENTRO are employed to chart change in voting response
surfaces over time. If, for example, the shape of the distribution of
weighted points which depicts the voting response surface for 1948 is
described using the centrographic measures produced by CENTRO, it is
expected the mean center would be situated in the Appalachian region of
western North Carolina and eastern Tennessee. In addition, the coeffi-
cient of circularity likely would approximate a more linear pattern and
the standard ellipse would be highly elongated in a northeast to south-
west direction. This pattern is to be expected since in the earlier
years of the examination period the Republican party core (locational
base of electoral support) was centered almost entirely in the Appala-
chian area of Tennessee and North Carolina. However, with more recent
elections, if as hypothesized, the GOP becomes increasingly more suc-

cessful at the polls, its locational base of support should demonstrate

71

less regional concentration. The traditional core should diminish in
relative importance as electoral support increases in traditional
Democratic bases of support to the south and west of the traditional
core. Thus, it is expected that the mean center computed for successively
more recent elections will advance generally south and west. Corre-
spondingly it is expected that the coefficients of circularity would
demonstrate increasingly less linearity and will begin to approach
circularity. Finally, the ellipses describing each election should
demonstrate less ellipticity and more compactness since the standard
distances along the major and minor axis are expected to approach
equality. Of course, for any given date during the period under
scrutiny, it is expected that the senatorial parameters will demonstrate
less indication of change than their presidential level counterparts.
Thus, in 1964 for example, the mean centers of the senatorial elections
should be situated further north and east, the ellipses should be less
circular and the coefficients of circularity should be nearer 0.0 than
the presidential level counterparts.

These expected results, if verified, are indicative of changing
patterns of electoral response in the study area. This change involves
a gradual spread of electoral support for the Republican party from an
early concentration in Appalachia in the late 1940's and virtual absence
of support elsewhere in the study area, to a more uniform pattern of
support throughout the South.

However, even though the numerical measures of spatial distribu-
tion described thus far may hint at a trend toward more uniform patterns
of voting response, these measures cannot provide reliable indices of

either concentration or uniformity. To accomplish these types of

72

measurement we must turn to another type of measure. We must describe
the density of the point distribution.

Density is most commonly measured by dividing the number of
occurrences of any given phenomenon by the area in which the phenomenon
occurs. For the purposes of this inquiry, this is insufficient since
simple density provides no indication of the relative position of the
individual points. If, for example, 50 contiguous counties vote heavily
in favor of a Republican presidential candidate, the relative compact-
ness or clustering of this support can be overwhelmed by the total area
of all the counties where support may be relatively low. In order to
account for such interval variation nearest neighbor analysis is ems
ployed. This technique provides a measure of the degree to which a
pattern of points departs from randomness. This is accomplished by com-
paring the actual straight line distance between each point and its
nearest neighbor with the expected distance if the nearest neighbor were
distributed randomly. Thus, the observed mean distance between each
point and its nearest neighbor is divided by the expected mean distance
to achieve a statistic R which can vary from 0.000 to 2.1491. This R

statistic provides a measure of clustering, randomness, or uniformity

since:
when R = 0, maximum clustering exists;
when R = l, randomness exists; and
when R = 2.1491, maximum uniformity exists.

In this inquiry a R statistic will be computed for each of the
12 elections or distribution of points. The size of the point distri-
bution will be limited to only those counties that provide greater

than 45.0 percent of the total vote to Republican candidates in each

73

election.1 Computation of the R values will be accomplished by utilizing
a computer program entitled NABOR (Rhynsburger and Wittick, 1973, 21-22).
Since in earlier election years (1948-1956) Republican electoral
support was highly concentrated in specific geographic locales, it is
expected that these point distributions will produce R statistics which
approach clustering (R nearer 0.0). In the middle election years these
R values should approach randomness (R approximating 1.0) as support
develops in areas outside the traditional cores of Republican support.
Finally, in later years, as support becomes rather evenly spread across
the study area, the R statistic should approach uniformity (R approach-
ing 2.1491). Again in each case the R values of senatorial level elec-
tions should lag behind (be nearer 0.0) their presidential level counter-

parts.

Analysis of Process

Although it has been suggested that the changes in the locational
base of support for the Republican party occur uniformly over the sur-
face of the study area, this is not meant to infer that the change is
spatially random. Quite the contrary, a great deal of evidence has been
introduced earlier to support the thesis of a spatially specific pattern
of change centered on cities.

The second stage of this analysis is devoted to testing the thesis
of an urban related pattern of electoral change. Where the first stage
centered on detailed description of the structure of electoral response

surfaces, the second stage involves detailed description of the process

 

l . .
Forty-five percent 18 taken as a measure of at least a competi-
tive range of electoral support.

74

underlying changes in those electoral response surfaces. The analysis
does not attempt to describe all those processes linked to change;
rather the analysis focuses on one particular element of the spatial
structure which the literature has strongly indicated facilitates this
process of change, viz., the urban structure. Thus, the thesis that the
growth of electoral support for the Republican party occurs in larger
urban population centers is tested. To expedite this testing procedure,
an analysis routine was devised involving map comparison, non-parametric
tests of statistical relationship and comparison of mean levels of
support at varying categories of population size.

One method of comparing voting surfaces to urban structure is by
simple visual perusal of maps depicting each phenomena. Since maps of
voting surfaces will have been prepared previously for the description
of the voting structure, there is no difficulty in preparing maps using
the same county units, map intervals and Z score base, for the urban
structure. Three such maps were prepared from data for the three
census periods used. Thus, the original variables (before computation
of central tendency measures) were percentage of urban population by
county in 1950, 1960 and 1970. Visual inspection of these election and
urban sets of maps was accomplished in a search for such coincidence of
spatial regularities as might exist.

To explore the relationship between urban structure and voting
structure in greater detail, it was determined that a more precise
measure of association was needed. Unfortunately, the nature of the
data employed in this analysis restricts somewhat the possibilities
for determining the exact nature of this relationship. It would be use-

ful, for example, to turn to the traditional tests of significance, such

75

as the t-test, the F-test and correlation coefficients, to test the re-
lationship between urban size and change. But these tests require cer-
tain assumptions about the statistical characteristics of the population
under study if the results are to be valid. One of the principal assump-
tions is normality. But, as discussed earlier, the data in this analy-
sis deviate from a normal distribution. Thus to explore the urban/vote
response relationship we must rely on tests which require fewer assump-
tions about the population parameters than the traditional inferential
tests. We must turn to non-parametric statistical tests of association.
The specific test chosen for this analysis is Spearman's rs which
is derived from a group of routines that employ rankings of data to
measure association. The measure utilizes rankings of two variables to
achieve a statistic (rho) which is somewhat analogous to a product-
moment correlation. Thus, the rho statistic will vary between -l.0,
where the ranks are in perfect disagreement, and +1.0, where the ranks
are in perfect agreement. A rho of 0 indicates no relationship at all.
Again principal computations will be made by employing the high speed
digital computer and a program for deriving rho correlation coefficients.1
Since this program computes exact probability levels in testing
for the level of significance of rho, it is possible to test the hypoth-
esis that urban structure and voting structure are related. To accom—
plish this, it is assumed that there is no significant relationship be-

tween percentage of urban population by county and the percentage of

 

1The program used in the computation of Spearman's rS was written
by John Morris of Michigan State University (Morris, 1967a and 1967b).

76

Republican vote. This reverse approach is adopted since we cannot
actually "prove" that the two are related. We can only demonstrate
that the observed distributions could or could not have occurred by
chance. The reverse hypothesis (the null hypothesis, H0) is thus that
the observed data are the results of random variations. Thus, the level
of significance is the probability that the data might have been gener-
ated by a random process. If this probability is low, the Ho may be
safely rejected; if this probability is high, the null is accepted and
the research hypothesis--that urban structure and voting structure are
related-~is rejected. By eliminating the false hypothesis the research
hypothesis can be accepted since we know the distributions actually
could not have occurred by chance.

Therefore, we hypothesize that there is no relationship between
percentage of urban population by county and percentage of Republican
vote. Since there are 12 elections (7 presidential and 5 senatorial
level elections) and 3 separate measures of urban population (1950,
1960 and 1970) we must devise numerous null hypotheses (one for each
possible comparison of urban structure and vote structure ranks).

But generally all null hypotheses (Ho) will assume the following form:
Ho there is no significant difference between the level of
support for the Republican party in the 1948 presiden-
tial election and the level of urban population in 1950.
Testing will be accomplished at the .05 level significance. Thus, if
the probability figures are less than the .05 level, the values of rho
can be said to be significant at that level. The rho values will be

computed for the rank order of each of the following variables:

77

Percent
urban population by‘county in: Republican vote by county in the:

 

1948 presidential election

1952
1950
1948 senatorial election

1954

 

 

1956 presidential election
1960

1960
1964

1960 senatorial election

 

 

1968 presidential election
1972
1970
1966 senatorial election
1972
While the visual and statistical correlation procedures provide a

means of gauging the overall relationship between urban structure and
voting structure, these methods cannot accurately depict the variation
in voting preference that occurs within the urban hierarchy. It is
difficult for example to measure the effect of population size on voting
response. To satisfy this goal, tables depicting the relationship be-
tween urban population size and level of electoral support were pre-
pared. The level of electoral support is depicted by the mean level of
support in each election for each category of population size. Popula-

tion size categories were computed by relying on census figures for

total population by county. Seven categories were utilized as follows:

78

>250,000
100,000 - 250,000
50,000 - 99,999
25,000 - 49,999
10,000 - 24,999

< 10,000

In every case the population data from the census year nearest an elec-
tion year was employed to determine population size of a constituency
(county). For example, in the case of the presidential election of
1948 the 1950 census figures were used; and for 1956 presidential re-
turns, the 1960 census figures were employed.

In addition to exploring the relationship between population size
and level of electoral support for the Republican party, these tables
will also facilitate examination of the breaking point concept (Epstein,
1956; Adamany, 1964) discussed in Chapter II. Variation of electoral
support due to population size should be reflected in a variation of
the mean values for each category of population size.1 Any marked
variation in the pattern of these mean values will identify the break-

ing point or perhaps points.

 

1Although the difference of means tests or some other appropriate
test of significant difference in computed means would be useful in this
particular section of the analysis, the data do not fit the assumptions
of normality. However, visual inspection of the computed tables suffices.

CHAPTER IV

DESCRIPTION OF STRUCTURE

Although the context within which geographers work and
view the world has changed recently, we have not abandoned
traditional questions; "Where?" and "What is where?"

are basic in any geography and they still occupy much of
our time (Abler, Adams, and Gould, 1971, 82).

Geography has long been a discipline inclined toward asking
"Where?" And understanding just "what" existed at "where" has always
seemed a natural order of circumstance. Though these questions have
assumed a new dimension with the change of context in which they are
considered from the absolute space of miles and kilometers to the rela-
tive space of times and costs, they are nonetheless an essential part
of geographic inquiry. It is hardly conceivable, at least at an em-
pirical level of inquiry, to move to questions of "Why where?" and
"How?" without first having established "Where?" Thus, it is a principal
goal of this study and the primary objective of this chapter to de-
scribe as accurately and comprehensively as is feasible the structure
of the voting response patterns of GOP support from 1948 to 1972.

The description will consist of a general analysis of overall
patterns of electoral response as well as an examination of more
detailed or specific patterns at the subregional level. The section
devoted to general patterns of electoral response considers first the

nature of the elections under scrutiny and then places these elections

79

 

80

into the proper spatial framework. In both the electoral context and
the spatial context the discussion centers on the numerical parameters
which describe the vote distributions. In the discussion of specific
areas of Republican growth, the primary Objective is the identification

of spatial consistencies in the changing patterns of electoral response.

General Patterns of Electoral Response

Electoral theory has demonstrated the stability of partisan
alignment of a constituency or individual voter. But the presence of
electoral change presupposes the existence of disruptions or fluctua-
tions in the normally stable patterns of electoral response. Thus, we
can move forward in this analysis in the expectation of encountering a
measure of volatility in voting response patterns. But how then do we
distinguish between the long term disruption of electoral response pat-
terns that comes as a consequence of electoral change,and the relatively
more volatile short term fluctuations that may occur due to deviant elec-
tions or special nature of a particular election? For example, we may
discover that the South has indeed changed from the solid Democratic
position of pre-1948 to heavier Republican electoral support in 1972.
But how then do we distinguish between this 24 year or long term change
and the temporary disruptionsixivoting response that Barry Goldwater's
strategy or a third party candidacy of obvious southern appeal may
have wrought? We cannot unless we understand the character of each
election under consideration. Thus, initial consideration is given to
the electoral context and to the specific characteristics and the spa-

tial properties of the elections under scrutiny.

81

Electoral Parameters

In this section the goal is simply to outline the main features
of the elections under scrutiny in terms of the numerical distribution
of votes. This is accomplished by discussing these elections in terms
of the numerical parameters of the vote distribution, such as the mean
and the standard deviation. A summary of these parameters for each of

the 12 elections is presented in Table 4.

Numerical means

In general the patterns of the mean levels of support suggest
that in the 448 counties of the study area the electoral fortunes of
the Republican party are improving with each election. Although it has
been a rather volatile rise, the mean level of support has increased
considerably from 1948 to 1972 at both levels of competition. Con-
sidering now only 1948 and 1972, the mean level of Republican vote in-
creased over three fold from 13.83 percent to 44.63 percent at the
senatorial level and from 20.08 percent to 73.71 percent at the presi-
dential level. For the most part, this increase has demonstrated a
greater measure of stability at the senatorial level, with only one
deviation from a pattern of steady increase in the mean--the 1954
senatorial elections. But even at the presidential level, the pattern
of a steady increase in the mean level of support is broken only twice.
In his second attempt at the presidency, the appeal of candidate
Eisenhower to voters in the study area seems to have decreased slightly;
and not unexpectedly, in 1968 the appeal of third party candidate
George Wallace sent the mean level of support for the GOP plummeting

back down to near the 1948 levels (26.85 percent level).

82

TABLE 4

ELECTION PARAMETERS

 

Level of Electoral Competition1

 

 

 

Presidential Senatorial
Standard Standard
Year Mean Deviation Mean Deviation Year
1948 20.08 18.78 13.83 19.22 1948
1952 36.72 17.37
13.12 17.47 1954
1956 34.36 17.97
1960 38.98 17.04 18.99 19.15 1960
1964 56.01 16.62
36.89 21.28 1966
1968 26.852 16.452
1972 73.71 10.72 44.63 14.24 1972

 

1Figures based on percentage of total vote that went to Repub-
lican candidate.

21f the Wallace percentage of the total vote is added to the
percentage voting for Richard Nixon, the mean value is 75.88
and the standard deviation is 9.81.

SOURCE: Computed by author.

83

One of the more notable patterns in these mean levels of support
is the sharp difference between the means of elections before and after
1960. At the presidential level, despite the popularity of Dwight
Eisenhower, the mean never rises above 40 percent until 1964 when it
increases 150 percent over the 1960 level! Only the effects of a three-
way split in the vote disrupts the pattern of generally higher levels
in post-1960 elections. And at the senatorial level this pattern is
duplicated with 100 percent increase in the mean level of support from
1960 to 1966. Indeed up until the 1966 senatorial level elections,
there was general support for presidential Republicanism. The dis-
crepancy between the magnitudes of the means at the two levels of elec-
toral competition in the pre-l960 period suggests that Republicans were
indeed faring considerably better at the presidential level of compe-
tition. Thus, the relative successes1 of the Republican party at the
presidential level in 1952, 1956 and 1960 might well have come as a con-
sequence of the appeal of presidential candidates or the disenchantment
with Democratic presidential candidates, and not as a result of overall
party gains; especially since this success was not duplicated at the
senatorial level in 1948, 1954 and 1960. But the argument of strictly
presidential Republicanism begins to lose validity once the results of
the post-1960 elections are introduced. For although the major dif-
ferences in levels of support between levels still remain after 1960,
there is a dramatic increase in the mean level of support in the sena-

torial competition which matches or surpasses that of the presidential

 

1Relative to the near vacuum of pre-1948 times.

84

level from 1960 to 1964. This increase, coming as it does at both
levels, offers evidence that the inroads made by the Republican party
beginning with 1964 were far more than simply a form of single level
(presidential) Republicanism. The occurrence of this break at both
levels of competition introduces the possibility that the dramatic in-
crease may have come as a consequence of something more than simple
candidate appeal. Some of the senatorial level gains in the post-1960
period can be attributed to the fact that the Republican party began
offering candidates for competition in Georgia after 1960.1 But the
mere fact that Georgia Republicans felt they had progressed far enough
to offer competition to the Democrats is noteworthy in itself. At any
rate the influence of the 1960-1964 period in the fortunes of the

southern GOP is one point to bear in mind as the analysis unfolds.

Standard deviations

Although the mean value represents one avenue of describing
electoral support for the Republican party in the sample of counties
under consideration, the standard deviation is perhaps a more accurate
measure of party performance. Standard deviations provide a reliable
measure of compactness in the distribution of votes. In short, the
lower the standard deviation the more uniform the distribution. Thus,
in 1948 and 1952 when some counties of eastern Tennessee and western
North Carolina were providing Republican candidates from 50 to 70

percent of their vote, while Deep South counties were giving less than

 

1Since Georgia accounts for 159 of the 448 counties in the sample,
any increase from the 0.0 level of support of a no candidate election
would most significantly alter the mean values.

85

1 percent to these same candidates, one might logically expect the dis-
tribution to become more compact as electoral support becomes more
uniformly distributed. That is, as the distribution becomes more com-
pact, the standard deviation should decrease. Since to a large degree
more uniform distributions reflect an improvement in Republican fortunes,
the standard deviations should provide a valuable measure of their
success.

The standard deviations for the 12 elections under scrutiny are
provided in Table 4. In general these values provide a measure of
support for the thesis of electoral change characterized by a growth of
electoral support for the Republican party. Consider first the presi-
dential level of competition. Except for the election of 1956, the
standard deviations demonstrate a remarkably stable and continuous
decline. Even in 1968 when three parties split the vote, the standard
deviation is below the value for 1964.1 The values of the standard
deviations at the presidential level of competition indicate considerable
improvement in the competitive position of the GOP in the study area.

Republican fortunes at the senatorial level are somewhat less

promising. Over the entire period the standard deviation values decline

 

1One common assumption is that the vote for George wallace would
have gone primarily to the Republican candidate, Richard Nixon, in a
normal two party contest. At the very least Wallace, so the argument
goes, out into Nixon potential support (Converse, Miller, Rusk, WOlfe,
1969). To examine this notion the wallace vote and the Nixon vote
were summed and considered as one total. Wherever statistics for the
1968 election are provided, this two party summation will also be given
as a means of comparing the actual results with the theoretical poten-
tial. It is interesting that the mean values at the presidential
level lose some of that characteristic volatility if the mean of 75
percent for this total two party vote is considered in lieu of the 27
percent for Nixon alone.

86

to a point where the 1972 value of 14.24, approximates the 1972 presi-
dential value of 10.72. But the pattern of standard deviations at the
senatorial level is far less stable than those at the presidential
level. In fact, from 1954 to 1966, the Republican party fortunes appear
to decline rather than improve as the standard deviations actually rise!
One likely explanation for the volatility of this parameter may lie

in the manner of measurement used at the senatorial level. Since this
parameter is highly susceptible to the range of the distribution values
and since it measures deviation for the entire sample area, the influ-
ence of individual state contests would be considerable. For example,
the very high value in 1966 (21.28) could be a function of considerable
variation in level of support across state lines. In South Carolina

in 1966 the Republican candidate Strom E. Thurmond was rolling up a
mean level of 62.2 percent of the total vote against his hapless
Democratic opponent, Bradley Morrah; while in Georgia, Earl E. Patten,
the first Republican candidate for senator since reconstruction, could
hardly be expected to beat unsurmountable psychological odds, historical
precedent, 22g two term incumbent Herman Talmadge.1 Such a wide diver-
gence in the level of support is expected to produce large standard
deviation values; and these larger values are directly attributable to
a voting surface composed of seven different electoral situations
instead of one as in the years of presidential competition. Such in-

ternal variation may dilute the value of the standard deviation in the

 

1These candidates were actually involved in an election taking
place in 1968, but the 1968 senatorial election in Georgia was sub-
stituted for a no data year in 1966.

87

consideration of senatorial election responses since this parameter
may not be as reliable as the presidential level. At any rate the low
standard deviation for 1972 senatorial (14.24) and the steady (and
stable) increase of mean level of support at this level, do provide a
degree of support for the notion of improved Republican fortunes at the

senatorial level of electoral competition.

The hypotheses in light of the results

In summary the expectations of general improvementiJ1the competi-
tive position of the Republican party vis a vis its pre-1948 electoral
situation is fulfilled. In light of the increase in mean levels of
support at both the senatorial and gubernatorial levels of competition
and in light of the overall decrease in standard deviation values at
both levels, the political fortunes of the GOP appear on the rise.
The hypothesis of time lag in the level and intensity of support between
the presidential and senatorial levels of competition is also fulfilled.
Republican candidates have achieved greater overall success in the study
area at the presidential level. Indications of improved competitive
position at the senatorial level in the post-1960 period do not, how-
ever, support the notions of presidential Republicanism. From 1960 on,
electoral change appears to be prominent at both levels. In addition,
the hypothesis of greater stability in the patterns at the senatorial
level of competition is supported by the values of the mean levels of
support. However, the need for caution in the interpretation of sena-
torial patterns is made apparent by the highly irregular nature of
standard deviation values. Particular caution is called for in the in-

terpretation of patterns of response for the following elections:

88

presidential, 1956 and 1968; and senatorial, 1954 and 1966. It is these

elections that deviate most from the overall patterns.

Spatial Parameters

Given these words of caution on the nature of the elections under
consideration, and the insights into the overall electoral context,
we may now proceed to place the election in a more spatial framework.
This is accomplished by utilizing the same numerical description given
for the elections, i.e., by providing parameters such as the mean and
the standard deviation of each election. But in this phase of the
analysis these parameters provide descriptions of point distributions.
Thus, the level of Republican electoral support is considered only as a
value assigned to a particular point in space. And the total of these
weighted points is characterized by the shape, density, and dispersion
of the distribution. These parameters thus provide a measure of the

spatial variation in voting response over time.

Shape of the distribution

Three parameters which provide a measure of the shape of a dis-
tribution of points are the mean center, the coefficient of circularity,
and ellipses computed from measures calculated to represent one standard
deviation from the mean center. Usually these parameters provide
reliable indices of variation in the shape of a point distribution over
time. However, in this instance it was discovered that these parameters
were subject to marked influence by comparatively small portions of the
total point distribution, especially if extreme values of either weights
or locations were involved. Thus, the shape of the ellipses and the

value of the coefficients of circularity would fluctuate considerably

89

if peripheral counties1 in the areal sample were eliminated from the
analysis. Consider the values of the coefficients of circularity
provided in Table 5. The coefficients on the left represent the values
obtained by utilizing the entire (448 observation) sample. The values
on the right are the coefficients obtained if 10 percent of the extreme
points are cleaned from the distribution.2 Although the difference is
slight in some cases, for example, the 1952 presidential election, the
cleaning operation has the potential of markedly affecting the coeffi-
cient derived such as in the case of the 1948 senatorial election. Thus,
caution must be employed when discussing the values for the coefficient
of circularity or when examining the structure of the ellipses, since
it is difficult to account for the influence of such highly deviant
values.

It was, however, noted that the mean centers of the distributions
considered in this inquiry were apparently subject to less influence
by extreme values. For example, altering the composition of the dis-
tribution by eliminating the most extreme 10 percent of the points
from the analysis actually caused little variation in the location of
the mean centers. The only noticeable affect was a slight shift of
the mean location to the north. This northward shift appeared to be

rather consistent for all elections. Since the mean centers appear to

 

1Peripheral is used in terms of both extreme locational values and
extreme data or weighted values.

2Data cleaning or elimination of the most extreme data or loca-
tion values is an option provided with the CENTRO program. It is a
commonly employed procedure and in this case proved a useful method of
detecting the reliability of the spatial parameters employed in this
analysis.

90

TABLE 5

COEFFICIENTS OF CIRCULARITY

 

Level of Electoral Competition

 

 

 

 

 

Presidential Senatorial
Total Distribution Total Distribution

Year Distribution Minus 10%1 Distribution Minus 10%1 Year
1948 .931 .836 .723 .481 1948
1952 .903 .898

.553 .543 1954
1956 .891 .899
1960 .937 .917 .707 .623 1960
1964 .858 .939

.927 .899 1966
1968 .9832 .906
1972 .953 .958 .941 .988 1972

.UNWEIGHTED POINT DISTRIBUTION COEFFICIENT = .962

1Distribution after elimination of 10 percent of data extremes.

2The coefficient derived by using point distribution weighted with

sum of Wallace and Nixon vote was .957.

SOURCE: Calculated by author.

91

be subject to less fluctuation than coefficients of circularity, the
primary discussion of the shapes of these 12 elections or point dis-
tributions will focus on this parameter.

Neither the coefficients of circularity nor the ellipses computed
from standard distance measures in themselves offer much support to the
thesis of gradual electoral change. The results of computations for
both parameters suggest high values approaching circularity. For ex-
ample, the expected low values for the coefficients of circularity in
earlier elections were not confirmed (Table 5). The coefficients of
.931 and .723 for the 1948 presidential and senatorial elections re-
spectively reveal a nearly circular distribution of points, not the
linear pattern expected. The high values are consistent throughout the
study period, although, as postulated, somewhat lower coefficients are
obtained for senatorial elections. This tendency toward a gradual in-
crease in this parameter does not permit confirmation of the hypothesis
of gradual improvement in the competitive position of the GOP in the
study area. The coefficients of circularity are subject to too great
an influence from extreme cases.

The nearly circular structure of the ellipses for these elections
(Figure 2) confirms the circular shape suggested by the coefficients of
circularity. Thus, again the expected gradual increase from low coeffi-
cients and elongated ellipses to high coefficients and less elliptical
shapes is marred by almost circular patterns throughout the period.
And, again, on the basis of the configuration of the standard ellipses,
the hypothesis of gradual regularity in the shape of the point distri-

butions cannot be confirmed.

92

 

 

FIGURE 2.
STANDARD ELLIPSES DESCRIBING THE SPATIAL DISTRIBUTION
OF THE REPUBLICAN VOTE FROM I948 TO I972

PRESIDENTIAL

 

SENATORIAL
l948

 

 

 

 

93

If the extreme data elimination option is employed in computing
these parameters the progression from low to high coefficients and from
ellipticity to circularity is more closely approximated. However, no
logical justification can be advanced for such data cleaning at this
point in the analysis. In the analysis of the relationship between
electoral change and urban structure such elimination is utilized since
the core counties are not involved in any electoral change. Thus,
eliminating them from an analysis of electoral change over the entire
study area can be justified. But at this point a description of the
total structure of electoral support is under consideration. Thus, no
part of the spatial structure can be eliminated since this would result
in altering the structure itself. Thus, the uncleaned data are employed
in the analysis. However, an examination of the relative position of
the ellipses (Figure 2) reveals some hint of changing patterns of sup-
port. The gradual, if somewhat irregular shift, of the ellipses south-
ward with reference to the outlines of the study area is suggestive of a
movement of the locational base of party support to more southernly
locations. But this phenomenon is best approached by considering
another parameter, the mean center of Republican support.

The mean center is somewhat akin to an arithmetic mean, and is
frequently employed in the computation of the geographic center of
population for the United States. In this study the mean center be-
comes a geographic center of Republican support. The mean centers
represent the geographic center of the electoral support for the Repub-
lican party in the study area for each of the presidential and sena-

torial elections. The mean centers for each of these 12 elections are

94

plotted in Figure 3. To facilitate interpretation, presidential elec-
tions are plotted separately from senatorial elections.

These centers are viewed as indicative of the geographical loca-
tions of electoral strength and weakness. For example, consider the
mean centers of the 1948 elections. As expected the center of support
in 1948 lies in the traditional Appalachian core of Republicanism with
the senatorial mean in a slightly more northern position than the
presidential mean which lies at the southern edge of the core. The
strength of electoral support for the GOP in 1948 was Appalachia;
its weakness lay almost everywhere outside that core1 but in particular
electoral support for Republican candidates diminished as one moved
south from this core. However, as electoral response patterns in the
study area changed, the support for Republican candidates in the Deep
South increased. Since a large number of the sample points lay in this
area, and since their weighted values (percentages) increased, the mean
center was expected to move southward in subsequent elections. Applying
this interpretation to the mean centers of our electoral sample, we
might thus expect a north to south migration of mean centers much as
the population centers of the United States have steadily moved from
east to west.

In general the performance of the Republican party that is mani-
fested in the mean locations for these 12-point distributions conforms
to expectations. As expected, the general pattern for the means is a

north to south alignment with earlier elections situated in more

 

1As will be noted later in the analysis, in 1948 a measure of sup-

port for the GOP could also be found in western Tennessee, the Piedmont
region of North Carolina and in the area surrounding Winston county,
Alabama.

95

 

 

FIGURE 3.
MEAN CENTERS OF REPUBLICAN VOTE

 

g—J—T'ENN.

 

““53 ALA.

_72
all-0‘...-I.ofieoona

 

 

PRESIDENTIAL LEVEL

 

f-J—‘FENN.

I946
0 I954

 

MmL-A- GA- 196%

I972
I966 o‘

 

 

 

 

  

SENATORIAL LEVEL

Neg

ac.

No.3

5.0

0 I00 MI

 

+ Geographic
center of
study area

 

 

96

northerly locations and the later elections farther south. In general
this pattern holds for both levels of competition but the expected lag
of senatorial elections is confirmed since their mean centers are
situated in more northerly locations than the presidential counterparts.
The pattern of progressive southern advancement of means is especially
noticeable if the means for each of the elections at both levels are
referenced to the unweighted mean center (depicted by a + on each map).
In general the early elections (1948-1954) are furtherest away from the
unweighted mean center, the latest elections (1964-1972) tend to be
closest;1 and the middle elections (1955-1963) somewhere in between.
The early elections demonstrate the locational characteristics attrib-
utable to traditional patterns of GOP support since the mean centers of
these elections are situated virtually in the Appalachian core. The
latest elections demonstrate the type of response patterns expected if
a competitive electoral situation existed in the study area. Although
there are specific areas of concentration of Republican electoral
support (and non-support), there exists little of the virtual non-
competitiveness or zero level support that so characterized earlier
electoral periods. This is not meant to imply that the electoral sup-
port patterns of 1964 and 1972 were spatially consistent or equally
intensive throughout the study area; rather it suggests that the elec-
toral support was coming from virtually all sectors of the study area.
Thus, it is possible to utilize the relative position of these

three groups of presidential elections to characterize the nature of

 

1The presidential election of 1968, of course, deviates from this
pattern. But if the total vote for wallace and Nixon is used the mean
center for 1968 would be almost directly over the 1972 presidential elec-
tion mean center.

97

the electoral change occurring in the study area. Consider the charac-
ter of this change in terms of the periods of growth of Republican com-
petitiveness. In the earlier elections, such as 1948, there was a dis-
tinct absence of electoral support for the GOP outside of the traditional
cores of Republicanism. With succeeding elections, 1952-1960, the

party gained some electoral strength in other sectors of the study area
and the mean center is drawn southward from the mountain stronghold.
Finally, in the later elections, 1964 and 1972 in particular, the Re-
publicans gained enough strength to become competitive. Thus, except

for the presidential election of 1968, the positions of geographic

means for both presidential and senatorial elections from 1948-1972
indicate a steady increase of electoral support for the Republican

party in portions of the study area where little or no support existed
before. This would suggest that the electoral support is gradually be-
coming less geographically concentrated in traditional core areas as sup-
port has developed in more southern areas. Apparently the overall dis-
tribution of support is gradually approaching a situation in which no
sector of the study area is lacking some measure of support. Electoral
support is gradually becoming more uniformly distributed throughout the
study area. But to examine this conclusion further, we must turn to a

description of the density of the point distributions.

Density of the distribution

Descriptions of the density of the point distributions were ob-
tained by isolating those counties that returned over 45 percent of
their total vote to Republican candidates in each election under con-

sideration. The total number of counties in this category in each

98

election are provided in Table 6. These distributions were subjected to
nearest neighbor analysis to derive an R statistic descriptive of the
density of each distribution. The results of the 12 nearest neighbor
computations are provided in Table 7.

Given the R values obtained there can be little doubt that the
pattern of electoral support (over 45%) is moving from a more clustered
to a more uniform distribution. Since the highest possible R value in
this case is the 1.328 value obtained for the total study area, the
point distributions of later elections are approaching uniformity. The
R values move from .525 and .514 for presidential and senatorial, 1948,
respectively, to 1.321 and 1.156 values in the 1972 presidential and
senatorial elections. And only the presidential 1956 and 1968 and the
senatorial 1954 break a pattern of steady increase.

The pattern of these R values offers strong evidence of a tendency
toward more uniformity in the distribution of Republican electoral sup-
port.1 If these statistics are used in combination with the shift in
the location of mean centers, the notion of electoral change receives
additional support. The low R values for early elections combined with
mean centers in Appalachia depict concentration of electoral support
in the traditional cores. Progressively higher R statistics combined
with progressively more southerly locations for mean centers are indica-
tive of the gradual increase of electoral support in more southern parts
of the study area. Electoral support for the GOP is indeed becoming

more uniformly distributed and apparently it is occurring at both levels

 

1 . . . . .
Support, in this instance, 18 defined as more than 45 percent of
the total vote.

99

TABLE 6

NUMBER OF COUNTIES RETURNING GREATER
THAN 45.0 OF THE TOTAL VOTE FOR
A REPUBLICAN CANDIDATE
(Total Number of Counties is 448)

 

Level of Electoral Competition

 

Year Presidential Senatorial Year
1948 62 41 1948
1952 138
35 1954
1956 130
1960 158 56 1960
1964 340
248 1966
1968 76 (445)1
1972 443 252 1972

 

1Wallace vote plus Nixon vote.

SOURCE: Calculated by author.

100

TABLE 7

NEAREST NEIGHBOR STATISTICS

 

Level of Competition

 

 

Year Presidential Senatorial Year
1948 .525 .514 1948
1952 .826
.491 1954
1956 .819
1960 .908 .685 1960
1964 1.214
1.136 1966
1968 .578 (1.324)1
1972 1.321 1.156 1972

 

TOTAL STUDY AREA R STATISTIC = 1.328

1The R statistic derived by using the wallace
vote plus Nixon vote as a percentage of the

total vote.

SOURCE: Calculated by author.

101

of competition. However, the relative location of the mean centers and
the comparatively lower values of R at the senatorial levels suggests
again that the senatorial level of competition is lagging behind the

presidential level.

The hypotheses in light of the results

If taken together the evidence of electoral change favoring the
Republican party is certainly worthy of note. The migration of the
mean centers of location suggests a change in the locational base of
party support from the Appalachian core to a point approaching
the unweighted geographic center of the study area. This suggests that
support for the party has significantly increased in former Democratic
strongholds of southern and central Georgia, Alabama and Mississippi.
These results offer strong evidence in favor of accepting the hypoth-
esis of decreasing geographical concentration of electoral support.
For, as hypothesized, electoral support does appear to become more uni-
formly distributed throughout the study. And, as hypothesized, this
tendency toward less geographic concentration is not occurring in equal
proportions at both levels of electoral competition. Presidential can-
didates apparently invoke more spatially uniform response patterns than
senatorial candidates, if the entire study area is employed.

Determination of specific locations for electoral support and for
change in patterns of electoral response is the goal of the final stage
of the description of point distributions, an analysis of the dispersion
of the points. In this case the dispersion is described by ring and
sector counts of the intensity of electoral support. As a supplement

to this means of description, the maps of electoral response are

102

utilized to determine intensity at specific locations. In describing
such dispersion and relating support to specific locations, the scale
of analysis moves from general electoral response to subregional pat-

terns .

Subregional Patterns of Electoral Response

The electoral response patterns in the seven presidential and
five senatorial elections under consideration are summarized in Figures
4 through 10. These maps reflect the pattern of electoral support for
the Republican party based on the mean level of support in each election.
In interpreting the resulting patterns care must be taken to avoid as-
suming that high positive patterns depict high levels of support. This
can be true but only in reference to the mean level of support. Thus,
in the 1948 senatorial election (Figure 4) a county with a relatively
high 2 score value of +0.5 to +1.5 might still be lost to the
Democratic party since the mean level of support in this election was
in the study area only 13.83 percent. Even a +1.5 2 score may indicate
the Republican gained only approximately 42 percent of the total vote.
If the mean value is low even heavy support (>+2.5 Z scores) may be
relatively low in relationship to Democratic percentage of the two
party vote. Thus, the term support must be interpreted relative to the
mean given at the base of each map.

To aid in the interpretation of these maps, a ring and sector
analysis breakdown for each election is provided in Table 8. The values
given in this table represent the mean value of electoral support in
each ring and sector for each of the 12 elections under consideration.

The location of each ring and sector is shown in Figure 11.

 

 

FIGURE 4.

 

    
  
 

     
   
     

 

l

7
g

PPPPPPPP
RRRRRRRRRR

 

 

 

 

 
   
  
  
  

   

       
 
   
     
        
    
  

FIGURE 5.
/ " 3:121:1- - PRESIDENTIAL
REPUBTLéCAN
,., ///, :iiliiz-z-z‘ ----------------- / '
/w ’ g” CW'
‘fi ’s§€?%’€%'
¢A/ ’ /
y .. 9_P_P_0_SJI'_QI!__
' +121 scoznes
_- , Heavy V/A _
‘ ' ’// ¢Wl 1‘. Moderate WA _2::
. v/ L____LIghI WA #50
M€EEP_°"9° % +015.
LIghf
grzanZLz4'vaiouon; 17.97 Mwer‘m 1'“; :2:
Heavy

 

 

       

      
     
 

‘Rgfiflb

w-g 1;, ,
Mir/fig?
/%§4'@;

 
  

f

;%%%‘;%%V-e

  
 
 
 

 
 

    
 
 

    
  
   
 

l960

PRESIDENTIAL
REPUBLICAN

;
A."
r/ ‘0‘"

‘87
5%,
'Z/l9
/,

Mean: 38.98 1""1:i""""‘131
S’tandard Deviation: I7.o4

 

 

 

 

FIGURE 6.

    
   
 

    
   
  
   
   
 
  
 
   

  

' '/ // 4”” I964
/ 4///
I. ,, ,
II“ /1.
"2:223 ‘ @ V/‘M
23:: $532233. ӎ
REPRSJILQ'L-
Heavy .’////
Moderate , _
L____Light ' /
“2'15“” I +0..
Mean: 56.0! Ugh, % + 50

 

Standard Deviation: l6.62

 

I.______.__'_-_

l968

PRESIDENTIAL
REPUBLICAN

 

 

106

 

 

FIGURE 7.

l972 PRESIDENTIAL REPUBLICAN VOTE

 

 
  
  
    

 

e%///
r ........

fl?

.....
eco-
......
......
nnnnnnnnnnn
.......

oe-

000000000000000

eeeeeeeeeeeeee
..........

 
 

aaaaa
eeeeeeeeeeeeeee
IIIIIIIIIIIIIIII
IIIIIIIIIIIIIIIIIIIII
ooooooooooooooooooooo
eeeeeeeeeeeeeeeeeee
eeeeeeeeeeeeeeee
oooooooooooooooo
eeeeeeeeeeeeeeeee
oooooooooooooooooo
oooooooooooooooo
IOIICIOQOOO/ can.

 

 

 

 

Mean: 73.7l
Standard Deviation: IO.72

 

 

 

 

 

REPRSJ I198“
I z
I SCORES
E Heavy
-2.
I Moderate / 5°
E.___':Ighf / I.50
Mean Range -°'5°
I mom ”5°
+I.so
I
: Moderate % +2.50 ‘L____'39 Mi.
: Heavy
I

SUPPORT

 

 

 

 

 

 

 

 

108

 

 

FIGURE 9.

   
 

.;.:.:.;.:.:.;
'2'. -;Z;I;.;.;.- .,..:,:_:_::: . . . .-
.:§::.-.;..:.:.:.- .:-.-.::. ”$55. . . If .,._ I 6
. ='-:-:-.:-:-:-:-. 9 O
'.-.-.;.-.'.;.
Z'Z'.'.:.. .

AAAAAAAAAA

       
  
 
  

  

    

e; :53

   

  

., ,1 .........

:

 

 

 

109

 

 

FIGURE IO.

I972 SENATORIAL REPUBLICAN VOTE

 

 

 

 

 

 

 

MOO": 44.63
Standard Deviation: I424

,QEF’QSJ ILQI! _ _

2
SCORES

 

-z.so
Moderate V ,

-I.eo
___I_.Ight /// -050

Mean Range

 

I
I
I
I Heavy
I
I
I
I

 

m +0.50

LI III
g ..= +I.sa

'I' 2.50

 

 

0 I00 MI

 

SUPPORT

 

 

TABLE 8

MEAN LEVEL OF ELECTORAL SUPPORT BY RING AND SECTOR1

 

Senator

President

 

 

72

66

60

54

48

72

68

64

60

56

52

48

 

42.8

11.4

9.8 15.9 20.6 21.2 42.2 26.1 78.8

0.1
11.8

R151

37.3

7.1
13.2
14.9

18.9 23.5 52.5 18.5 72.5

17.8

R152

34.6

21.0 29.2 59.3 22.5 74.6

18.9

R153

27.8

9.5 17.3 17.1 23.5 58.3 19.2 74.7

11.4

R154

24.2 27.8 63.0 17.0 76.1 0.2 0.8 2.3 14.1 38.7
17.9

25.5

R155

36.9

26.5

13.6 14.6 26.8 29.1 60.9 15.8 79.0 9.3 6.3

R156

23.1 30.1 30.1 56.9 22.4 80.8 5.8 3.6 7.6 19.1 47.0

16.5

R157

44.8

19.2

33.1 39.6 37.9 48.3 34.1 78.4

26.1

R158

110

52.2
56.5

41.9
49.1

42.2 45.6 42.8 41.9 74.6 22.7 24.7 28.7
44.8 23.7 45.2 60.8 31.7 70.4 0.7 7.1 28.

42.2

30.5
2.8

 

R251*

R252

15.6 41.7

3.5
0.8

26.6 11.5 33.1 59.5 19.1 79.4 0.1 1.0
11.0

7.7
10.5

R253

39.6

18.1

20.0 24.8 64.4 16.8 80.1 0.4

23.9

R254

28.0 32.3

2.8
15.3

3.6
9.9
21.4

76.6

5.3
8.7

8.3 23.7 22.0 24.1 73.9
17.8

12.3

R255

38.0 29.1

28.8

66.5

76.5

40.9

40.4

38.1

R256

26.1 37.6 35.7

16.5

72.8

30.5 34.5 35.8 51.9

23.4

R257

38.8 47.3 50.5 52.1 44.4 39.4 71.2 27.2 28.0 22.0 45.7 55.0

R258*
R351

 

60.6 62.6 61.5 52.1 54.1 72.7 48.1 45.4 47.1 58.1 63.8

51.7

53.3 27.8 47.9 54.7 34.4 65.6 8.8 19.0 40.3 53.4 56.4
46.4

10.0

R352

3.3 3.7 13.4 35.7 50.7
7.6

7.6
2.6

17.7 44.1 44.9 54.6 26.6 70.9
13
21.8

R353

61.6 30.3

10.7

34.8 32.4 34.3 53.1 17.9 77.3

.8

R354

62.9 29.9

30.0 32.2 32.1 62.8 14.5 83.7 2.1 16.7
6.2
14.3

12.2

R355

29.7 32.8

15.3

3.2 35.3 28.3 29.6 83.2 5.9 68.8
23.1

23.4

R356

39.2 40.5

20.3

69.5

46.1

37.7

33.8

35.1

R357

35.1 45.7 44.3 49.1 36.1 37.2 63.7 33.7 24.9 21.4 47.5 55.0

R358*
R451*

 

46.4 56.4 53.4 46.2 44.2 68.0 29.8 33.8 46.2 46.2 50.4

.6

36

111

.muou deconuneuna.

.mmoa no H and; mucsoo aucoauoum an» ocean poww>oum van and m was mIm muouoom .c mcwm

 

 

 

 

H
o.mv m.h¢ o.mH o.mH h.NN n.0m 0.5N h.mm o.mm b.vm h.mm o.m~ hmvm
N.mm o.mm v.HH 0.0 o «.50 o.HH o.mm m.HN 0.Nm 0.5m N.m mmvm
Nb mm om vm mv Nb mo v0 00 mm mm mv
HOHMGOm UGGQHmOHm

 

~.v.ucouv m mqmda

112

 

FIGURE II.
RING AND SECTOR LOCATIONS

IN THE STUDY AREA

3 _..——-——— TENN-

 

 

 

 

 

3 Sector numbers
4 Ring numbers

9 I03 209 MI.

 

 

113

Interpretation of the mean values presented in Table 8 proves to be
exceedingly tedious since the interpretation procedure involves digesting
a 27 x 12 or 324 item table of mean values (one for each ring and sector
combination). In order to simplify this procedure it was decided to
eliminate the necessity of dealing with actual mean values and rely on

a surrogate indicator of change. Since the main focus of this inquiry
is electoral change and the improvement of Republican electoral for-
tunes,1 the surrogate measure will be the percentage of change between
each of the seven presidential and five senatorial elections for each

of the 27 ring and sector combinations. To further simplify the in-
terpretation procedure, these percent change values are then categorized

by intensity of change as measured by the following interval scale:

X = No change or negative change
'1 = 0.1 to 5.0%
2 = 5.1 to 10.0%
positive
change 3 = 10.1 to 15.0%
4 = 15.1 to 20.0%
5 = >20.0%

The results of this simplification procedure for presidential level
elections are provided in Table 9 and for senatorial level elections
in Table 10.

Possibly the most consistent pattern of voting response depicted
in these 12 maps and 3 tables is the traditional Appalachian core of

Republicanism. The only deviation from a consistent pattern of heavy

 

1Erosion of Republican electoral position would entail preserva-
tion of status quo, i.e., continued Democratic party electoral dominance.

114

TABLE 9

PERCENT CHANGE IN PRESIDENTIAL ELECTIONS BY RING AND SECTOR

 

48 to 52 52 to 56 56 to 60 60 to 64 64 to 68 68 to 72

 

Ring 1
$1 2 1 l 5 X 5
$2 4 l 1 5 X 5
S3 2 l 2 5 X 5
S4 2 X 2 5 X 5
SS 3 X l 5 X 5
S6 1 3 1 5 X 5
S7 2 2 X 5 X 5
88 2 2 1 3 X 5
Ring 2
81* 3 X l X X 5
$2 5 X 5 4 X 5
S3 4 X 3 5 X 5
S4 3 X 1 5 X 5
SS 4 X 1 5 X 5
$6 5 1 1 5 X 5
S7 2 1 1 4 X 5
S8* 2 l 1 X X 5
Ring 3
81* 2 1 X X 1 4
$2 5 X 5 2 X 5
S3 5 X 1 2 X 5
S4 5 X 1 4 X 5
SS 4 l X 5 X 5
$6 5 X l 5 X 5
S7 3 X 1 2 X 5
88* 3 X 1 X l 5
Ring 4
81* 2 2 X X X 5
S6 5 X X 5 X 5
S7 3 X l X X 5

 

*Traditional Core.

LEGEND
X = No change or negative change
1 = 0.1 to 5.0%
Positive 2 = 5.1 to 10.0%
Change 3 = 10.1 to 15.0%
4 = 15.1 to 20.0%
5 = >20.0%

SOURCE: Calculated by author.

115

TABLE 10

PERCENT CHANGE IN SENATORIAL ELECTIONS BY RING AND SECTOR

 

48 to 54 54 to 60 60 to 66 66 to 72

 

Ring 1
81
82
S3
S4
85
86
87
88

Ring 2
81*
82
83
84
85
86
S7
88*

Ring 3
81*
82
83
S4
SS
86
S7
88*

Ring 4
81*
S6
87

xxxI-oxxxx
><I~IUJH><><><><
b»uIn>quIUIR>uI
UIUIUIUIUIUIUIUI

I-‘XNxxI-‘NH
waNI-‘I-‘UIH
wwwhbwmw
NXXHWU‘INN

xxHxwax
XNNUUNUIH
mbummmww
NHHXNWI—‘N

h)h‘
Pow
LII-bx
RIP:

)4
H
X

 

*Traditional Core.

LEGEND
X = No change or negative change
1 = 0.1 to 5.0%
Positive 2 = 5.1 to 10.0%
Change 3 = 10.1 to 15.0%
4 = 15.1 to 20.0%
5 = >20.0%

SOURCE: Calculated by author.

116

Republican electoral support in this area is the 1964 presidential
election (Figure 10).1 This core is equally apparent on both map and
ring and sector analysis. In the ring and sector analysis the rela-
tively heavy support became apparent from the regularly reoccurring
high mean values (Table 8, R2, 81; R2, 88; R3, 88: and R4, 81).2

These five areas have been identified on all ring and sector tables by
an asterisk (*). Since these five ring and sector combinations enter
the analysis period (in 1948) with relatively high mean levels of sup-
port and maintain consistently high (at least relative to the remainder
of the study area) levels, the amount of change involved is often slight.
Thus, in the case of these core counties the summation of percent change
by ring and sector can be rather misleading, since the pattern would
appear to be one of erosion of support rather than positive change.

But reference to Table 8 verifies the negative change generally involves
rather small drops in mean levels. For example, R2, 81 has 42.2 percent
in 1952 and the same value in 1956.

Closer scrutiny of the ring and sector count offers one addi-
tional center of traditional Republican electoral support. Situated in
southwestern Tennessee, this lesser core of electoral support offers
mean levels of support somewhat below the levels of the Appalachian core
(Table 8, R3, 87). Yet reference to V. 0. Key (1949, 76-78) confirms

the historical pattern of support offered by this area. From

 

1Although the 1972 electoral response patterns (Figure 7) would
appear to be lower than normal, the dominant pattern lies within the
mean range category. The mean for 1972 is 73.7 percent.

2These notations are read as Ring 2, Sector 1 (R2, 81) and Ring
2, Sector 8 (R2, 88), etc.

117

anti-secessionist sentiment in 1861 to Republican majorities in 1896,
1916 and 1944 elections, this area has been a seat of Republican party
support.

Closer examination of successive maps of voting response reveals
one additional small node of historical Republican electoral support
in the area of Winston county, Alabama (Key, 1949, 281). This latter
example is not detectable by ring and sector count, but it is identifi-
able on virtually every map (1964 and 1972 presidential excluded) as an
island of GOP support.

The major core of Appalachia, and the lesser "cores" of south-
western Tennessee and Winston county, Alabama appear as consistent areas
of light to heavy support on the maps of voting response. These tradi-
tional "cores" have consistently offered heavier than average support
to Republican party candidates. As such, these areas will hardly
be subject to the same degree of electoral change which more traditionally
Democratic areas underwent in the same time period. Since the objec-
tive of this research is identification of changing patterns of elec-
total response, these traditional bases of Republican support will
receive less attention in subsequent analysis. However, the appearance
of these areas in this phase of the analysis does offer a measure of
confidence in the analysis procedure. Had these areas not been so
readily apparent in the description §2§_the literature, there would have
been cause to doubt the reliability of the descriptive procedure em-
ployed in the inquiry.

Aside from these cores of Republicanism, few long term consis-
tencies in voting response are readily apparent. The map patterns

outside core areas demonstrate a rather high measure of volatility,

118

with patterns seldom maintaining themselves longer than two elections.
In short, dynamics appear to be the dominant feature of the maps and
tables.

Close examination does provide some consistency, however. One
of the more noteworthy trends is the reoccurrence of a pattern of sup-
port for the GOP in the northern fringes of the study area and non-
support in the southern areas. Since the behavior of southern counties
(Sectors 3-6) generally assumes a pattern virtually opposite that of
their northern counterparts (Sectors 1-2 and 7-8), this north-south
division is readily distinguishable. For example, in 1964 southern
counties offered Goldwater heavy support while northern counties pro-
vided only average to light support.1 Yet, even these regional dif-
ferences erode in more recent elections. In earlier elections the north-
south division coincides with a general Rim versus Border South divi-
sion, and even some state boundaries can be traced between patterns of
support and patterns of non-support; but in the more recent elections
the influence of boundaries becomes less apparent and the pattern of
support/non-support becomes less concentrated as support becomes more
uniformly distributed. In fact by 1972 the pattern displayed on the map
of response patterns strongly hints at a random pattern of support
(Figures 7 and 10). The ring and sector analysis confirms the uni-
formly high levels of support throughout the study area in these

elections.

 

1Interestingly enough, it is only the five ring and sector com-
binations designated as Appalachian core (Table 9) which lost electoral
strength from 1960 to 1964.

119

This north-south division becomes more apparent if comparisons
are drawn between the levels of competition. The same general tenden-
cies are present at both levels as electoral support moves from a concen-
tration in northern fringes in 1948 (Figures 4 and 8) to more uniform
patterns of support in 1972 (Figures 7 and 10). The major difference,
however, is in the patterns of earlier elections. As might be expected
given the nature of the senatorial surfaces, the differences between
north and south in the elections of 1948, 1954 and 1960 are striking.
The pattern of north-south division is far more visible from senatorial
maps of the elections (Figures 8 and 9) than for the presidential
counterparts (Figures 4 and 5). In addition, the ring and sector
analysis demonstrates that the mean values of the southern sectors
(Table 8, Sectors 3-6) are generally well below those of the north
(Table 8, Sectors 1-2 and 7—8). However, as the north-south division
erodes in subsequent elections, the difference in levels of competition
becomes less apparent.

Aside from this general north-south division, it is, however,
possible to detect more subtle areas of apparent change in electoral
response patterns. Less obvious regularities in the response surfaces
are also noted in: central Alabama (R2, 86; R2, 87); virtually all of
South Carolina (R2, 82; R3, 82); the Atlantic coastline of Georgia and
Florida (R3, 83): the panhandle of Florida (R3, 84; R3, 85); and in the
area centered on Atlanta, especially those counties to the east towards
South Carolina (R1, 81 and 82) and towards the southwest towards
Columbus (R1, 85 and 86). In these examples some measure of increased
support for the GOP can be isolated by close examination of map pat-

terns in conjunction with ring and sector Tables 9 and 10.

120

It is interesting that in every case mentioned, except for
South Carolina, these consistent growth areas are either urban centered
or lie at the fringes of the study area in higher population density
coastal zones. The central Alabama example centers on the Birmingham
and Montgomery corridor. The Atlanta example contains several of
Georgia's larger cities, such as Athens, Mason, Columbus and Augusta.
And the coastal Georgia and Florida exampls are areas where increased
in-migration has resulted in high population density coastal settlement
as well as larger cities such as Savannah, Jacksonville and Pensacola.
Such coincidence of consistent patterns of growth of GOP support and
urban population clusters speaks well for the analysis of process to
follow in the succeeding chapter.

Except for the area between Atlanta and Columbus (R1, 85 and R1,
86), there are few areas where growth patterns remain consistent across
both levels of competition. The exceptions are the areas of South
Carolina (R2, 81 and 82) and coastal Georgia and Florida (R3, 82 and 83).
Of course, the South Carolina example is subject to the distortion of
candidate orientation since the "Thurmonism" of South Carolina voters
may be more widespread than the Republicanism. And the Georgia-
Florida example of the presidential level change is broken by dis-
tortions such as the Wallace movement in 1968. In fact in every case
except the Atlanta-Columbus example the pattern of change is subject
to a greater measure of volatility at the presidential level of com-
petition. Such volatility does not merit attaching any partisan labels
to the change since the next election may bring a complete 1800 change.
However, since these areas formerly were Democratic strongholds, per-

haps the volatility itself is worthy of note. If such areas cannot be

121

labeled Republican bases of support at least the volatility is sugges-
tive of electoral change. And the electoral change taking place in
these areas would appear to be no single competitive level phenomenon;
for although it may be less intense, it is as prevalent and more
stable at the senatorial level than at the presidential level of com-
petition.

Relying on the maps of voting response to detect any major dif-
ferences in the intensity of electoral support for the GOP at the two
levels of competition is difficult. Examination of the senatorial
maps (Figures 8-10) did, however, reveal major changes in state level
support patterns. Consider the change in Georgia patterns from the
senatorial 1948 (Figure 8) to the senatorial 1972 (Figure 10). Con-
sider also the change in the response patterns of South Carolina and
northern Florida. In each of the cases the dominant pattern in earlier
years is opposition. Yet in later elections the states display pat-
terns of support not unlike more traditional areas of Republicanism.
But for the most part, the major indications of time lags in levels of
support are revealed in the ring and sector analysis. The differences
between senatorial and presidential means for the same period conform to
expectation. The means of the senatorial elections are lower in earlier
years but tend to equalize in later years. This pattern of lower means
in earlier years proved useful particularly in identifying areas of
change. The examples of electoral change in central Alabama, South
Carolina, coastal Georgia and Florida's panhandle, became more Obvious
when the mean level of support at the senatorial level of competition
was considered. In these geographic areas the change appears to be

occurring at both senatorial and presidential levels. In the other

122

areas less intense bi-level support is noted in later elections
(1966-1972).

Thus, the map and ring sector analysis in general verifies the
impressions of overall electoral change expected and observed earlier.
But the identification of specific areas of change is difficult to iso-
late since the electoral change has apparently not stabilized enough
to impart any definite partisan labels to specific areas. And in those
areas, such as central Tennessee, where this quite possibly could have
occurred, this growth may well be masked by the relative position of
ring and sector boundaries versus county boundaries. The pattern
analysis, while illustrating traditional areas of support, does not
allow strict delimitation of "emergent" supportive areas of Republi-
canism. However, if the electoral change is focused on nodes, es-
pecially urban foci, then there is an explanation for the lack of
general and persistent patterns of change. For in such a case the
change is characterized as nodal or point (county) centered not areal.
This facet is examined in the subsequent analysis on the nature of

change.

CHAPTER V

DESCRIPTION OF THE PROCESS

Aside from this "tradition . . . Republicanism," found

in seven of the southern states, probably the best known
feature of presidential Republicanism in the South has been
the greater willingness of metropolitan electorates, as
compared to voters in the non-metropolitan sector, to
support Republican nominees (Cosman, 1966b, 53).

Description of the voting response structure provides three basic
prerequisites for analysis of electoral change. We now possess:

1) knowledge of the character of the elections under study; 2) a
description of general patterns of electoral support of specific points
during years since 1948; and 3) a detailed image of locational bases
of support for the Republican party. Given this pattern foundation, it
is now possible to continue the investigation of southern electoral
change into less static arenas, that is, to investigate the process of
electoral change in a spatial context.

In turning to an analysis of process we do not abandon description.
Rather the description simply shifts from area-wide voting structure to
urban structure, as it is here that the literature intimates the change
is first likely to occur. Thus, in attempting to affect an analysis of
changing patterns of electoral response not all possible variables or
inputs into the electoral change process are considered. In fact this

inquiry centers on only one segment of the process, that which relates

to the urban environment: for it is the hypothesis of this inquiry that

123

124

a strong positive relationship exists between the level of urban popu-
lation and the level of Republican support. Consequently, the descrip-
tion focuses on the urban structure of the study area or, more
accurately, on the hypothesized relationship between the urban structure
and the voting structure.

In considering the electoral change it is the Republican vote that
is the medium for analysis, since increased Republican vote response is
equated to change in the once "Democratic" South. The principal ques-
tions then become: Is there a relationship between the vote for Repub-
lican candidates and the urban structure or urban population? Does
increasing population size or increased urban population facilitate the
increase of the GOP vote? Or perhaps more appropriately, does a rela-
tionship exist at all? If so, what is the direction and strength of
the relationship? Does the relationship vary between competitive elec-
toral levels? The purpose of this chapter is to provide answers to
such questions.

The general relationship between the Republican vote and the urban
structure is examined through the use of visual correlation and non-
parametric statistical correlation techniques. As in the previous
chapter, map analysis and ring and sector counts again form the basis
for visual interpretations of relationship. Statistical relationship
is obtained by rank order correlations. A more detailed examination
of the relationship between vote and urban structure is affected by

considering the vote levels for varying population size categories.

125

General Patterns of Relationship

Maps and Ring and Sector Analysis

Utilizing the procedure employed in creating the voting response
structures described in Chapter IV, similar surfaces depicting the
levels of urban population in each of three census periods were devised.
These are provided in Figures 12 and 13. These maps depict the distri-
bution of urban population by county in each of three census periods--
1950, 1960 and 1970. Urban population is measured by the percentage of
the total population of a county living in a city over 2,500 population.
To facilitate comparison of these urban structures to their voting re-
sponse counterparts, the same 2 score intervals are employed for both
the voting and the urban structure. As a supplement to the visual
comparisons made between the urban structure and the voting structure,
ring and sector counts were derived for the three urban structures using
the same procedure employed above. The results of these ring and sector
counts are summarized in Table 11. As a further supplement to analysis,
the ring and sector results were once again simplified by calculating
the percentage and categories of percent change in urban population.
The same procedure was employed as that used for the voting structure.
The results are given in Table 12.

Although the maps of urban structure provide useful background
on the nature of the urban population, any concrete gains from visual
correlation are negated by the imprecision of such an analysis. It is,
for example, useful to know the nodes of population growth and stability
in that population. Also, it is useful to be able to locate specific
areas of high and low urban population. Yet a visual linkage between

three individual maps of urban population and 12 voting response surfaces

126

 

 

FIGURE I2.
PERCENT URBAN
POPULATION

       
 
         
   
     
    

 

  
   
  

 
 

 

a; I “ I: 50
' f"!£8 . ,4, '9’:
fife,

. ll/flg}/I/g’§I////e’/l 1”

’ " /w’a4//A7///

 

    

........

 

PERCENT URBAN
POPULI‘TION

Mean: 25.98
Standard Deviation: 23.86

 

 

127

 

 

FIGURE I3.
PERCENT URBAN POPULATION
|97O

 

Mean: 29.7?
Standard Deviation: 24.62

 

Z
SCORES

- 2.50
‘ -I.50
-o.5o
+0.50
.--:-:-:-:I +I.5o

+2.50

0 I00 Mi
|———l .

 

 

 

 

128

TABLE 11

PERCENT URBAN POPULATION BY RING AND SECTOR

 

Census Year

 

1950 1960 1970
R181 21.4 27.4 26.6
R182 7.7 7.7 16.2
R183 29.7 29.7 39.9
R184 25.8 25.8 30.8
R185 18.9 18.9 27.5
R186 28.4 28.4 34.8
R187 24.1 24.1 34.8
R188 27.5 27.5 37.6
R281* 14.5 17.8 18.6
R282 23.5 29.8 32.6
R283 9.8 17.2 25.7
R284 28.2 36.0 40.2
R285 19.8 27.7 33.4
R286 22.3 27.7 32.1
R287 22.6 29.5 32.3
R288* 20.3 23.5 26.4
R381* 19.5 21.0 22.8
R382 29.0 30.2 33.4
R383 32.6 38.4 41.4
R384 20.5 26.3 27.5
R385 27.7 35.5 39.2
R386 15.3 22.3 26.4
R387 15.6 23.0 29.0
R388* 13.2 16.1 22.0
R481* 39.4 41.2 42.6
R486 27.6 27.6 38.2
R487 16.7 18.7 24.3

 

*Traditional Core

SOURCE:

Calculated by author.

129

TABLE 12

PERCENT CHANGE IN URBAN POPULATION
BY RING AND SECTOR

 

 

 

1950 to 1960 1960 to 1970
Ring 1
81 2 X
82 X 2
83 X 3
84 X l
85 X 2
86 X 2
87 X 3
88 X 3
Ring 2
81 l l
82 2 1
83 2 2
84 2 l
85 2 2
S6 2 1
S7 2 1
88 1 1
Ring 3
81 1 1
82 l 1
83 2 l
84 2 l
85 2 l
86 2 1
87 2 2
88 l 2
Ring 4
81 l 1
86 X 3
87 l 2
LEGEND
X = No change or negative change
1 = 0.1 to 5.0%
Positive 2 = 5.1 to 10.0%
Change 3 = 10.1 to 15.0%
4 = 15.1 to 20.0%
5 = >20.0%

SOURCE: Calculated by author.

130

proves to be a nearly impossible task. Further, the details provided
by the ring and sector counts add little to such an analysis.

Interpretation of relationship between urban and vote structures
by ring and sector counts involved visual comparison of two tables--
one 27 x 12 items and another 27 x 3 items. Thus, only the more
readily apparent coincidences of relationship between high vote levels
and high urban levels could be isolated. For example, it is compara-
tively simple to note the coincidence of high urban levels in the coastal
areas of the Atlantic and Gulf regions of South Carolina, Georgia and
Florida (Table 11, R3, 83 and R3, 84) and the correspondingly high mean
levels of GOP support (Table 8, R3, 83 and R3, 84). Also the high urban
values of the Atlanta, Georgia area (Table 11, R1, 88) are matched by
moderately high levels of GOP support (Table 8, R1, 88). But any
attempt at drawing more detailed and well based relationships involves
considerable measures of subjectivity.

As earlier criticism indicated, the imprecision of map comparison
analysis is to be expected from virtually any map pattern analysis.
But the detail of the ring and sector count were designed in part to
overcome the imprecision of this method.1 And the success obtained in
isolating Republican strength areas (cores) and areas of consistent
change in the description of structure were sufficient cause for high
expectations in the analysis of relationship. But in general, the map

comparison and ring and sector count procedure is more revealing for its

 

1Perhaps less detailed ring and sector counts, i.e., 2 rings and
4 sectors, might alleviate the unwieldy nature of the procedure. But
the fewer the number of rings and circles the closer to the original
surface, and thus to simple map comparison.

131

omissions rather than its insights into the relationships under scrutiny
here. The procedure proved too subjective, too cumbersome, and too
imprecise for the detailed analysis needed to ascertain the nature of
relationship between urban structure and voting structure, much less
more detailed relationships such as the variations that might exist at
different levels of competition. Thus, we must proceed with more

accurate methods of delimiting this relationship.

Rank Order Correlations

To gain insight into the relationship between the urban structure
and the level of Republican support, Spearman's rs statistics were com-
puted for each possible comparison of electoral and urban population
ranks (Table 13). In the initial comparisons the values representing
percentage of urban population were ranked and compared to the ranks
derived from the level of Republican vote in each election. In only
three presidential level elections, 1948, 1952 and 1960, and two
senatorial elections, 1948 and 1954, was the null hypothesis of no
significant relationship rejected. Apparently if the entire range of
sample counties is employed in the rank order comparisons, the existence
of a significant relationship between urban structure and voting struc-
ture can only be confirmed in earlier elections from 1948 to 1960.
This is, however, reason to question the utilization of the entire
range of sample counties in such comparisons.

Consider at this juncture the primary thesis of this inquiry
as it questions the existence of the process of electoral change in the
study area from 1948 to 1972. The thesis here involves the existence
of a strong positive relationship between the urban structure of the

study area and the change from low levels of Republican electoral

132

.uonusm an ooumaooamu "mumDOm

.Ha>aa mo. an acmonmacmam.

 

 

 

 

 

 

 

ammo. «vma. who. amma. onma
moo. boo. «ema. «hma. ooma
who. ooo. «vma. «maa. omma
nouoo mocaz oameomv can“: w
In.
Nvo. moo. omo. ano. onma
nvo.I amo.I ammo. voo. ooma
«awn. eomm. cmma. «mma. omma
imamsnm anuoee cane: a
mnma ooma ooma vmma mama mnma ooma voma ooma omma mmma mvma
anauoumcom aaaucooamoum

 

EOauauomEou mo ao>oa

 

4

MBZDOU rm ZOHedaomOm ZdflMD ho mwflezmummm mm? OZH>AO>ZH
mZOmHmdmZOU mow mBZmHUHhhmOU mmflmo ¥z¢m n.242m<mmm

ma manta

133

support to competitive levels of support. But then it must be realized
that not all counties within the study area are changing in this fashion.
The question then becomes the logic of including such counties in the
study area when they are not involved in electoral change. Some counties
are remaining staunchly Democratic. These counties apparently resist

the opportunity to alter traditional response patterns. In rejecting
change to increased levels of Republican support, these counties present
a valuable insight into the character of resistance to change. As such,
reason dictates their inclusion in such an analysis.

But what of their opposites? What of the staunch Republican
counties? Such counties as identified in Chapter IV are now and have
virtually always been Republican "cores." Since we shall focus on only
one aspect of the process of change--the relationship to urban struc-
ture-~there is little apparent merit in including in the analysis,
counties that have virtually no opportunity to "change" to competitive
Republicanism. These are and have virtually always been at that stage.

Thus, on the basis of their traditionally high levels of electoral
support for the GOP at virtually all levels of competition, it was de-
termined that certain "core" counties did not fulfill the basic pre-
requisite for inclusion in an analysis of electoral change. For this
reason, these counties were purged from the sample of 448 previously
selected counties. A list of these 24 "core" counties so eliminated is
provided in Table 14. Decisions on the authenticity of core counties
were made by reference to the literature, and to the analysis done in
Chapter IV.

Having thus eliminated the "core" counties from the ranks of both

the urban and voting structure, the rho values were recomputed and the

134

TABLE 14

CORE COUNTIES ELIMINATED FROM RANK ORDER ANALYSIS

 

 

Number of
Counties
Eliminated State County Names
1 Alabama Winston
0 Florida None
1 Georgia Fannin
0 Mississippi None
8 North Carolina Ashe Graham
Avery Mitchell
Cherokee Wilkes
Clay Yadkin
0 South Carolina None
14 Tennessee Carter Jefferson
Cocke Johnson
Grainger Macon
Hamblin Sevier
Hancock Scott
Hardin Unicoi
Union Wayne

 

Twenty-four total counties eliminated as traditional areas

of Republican electoral support.

SOURCE:

Compiled by author.

135

results are given in Table 138. The elimination of these core counties
affected the strengths of the relationship between urban and voting
structure. It was now possible to reject the null hypothesis of no
significant relationship for five presidential level elections (1948,
1952, 1956, 1960 and 1968) and two most recent senatorial elections
(1966 and 1972). Interestingly enough the only two cases of presiden-
tial elections where it was necessary to accept the null of no rela-
tionship are the examples of the most successful inroads made by the
GOP in the study area--1964 and 1972.

It is also interesting to note the effect that elimination of the
core counties had upon the relationships between urban structure and
senatorial level elections. With the total sample only the 1948 and
1954 null hypotheses could be rejected at the senatorial level. Elimi-
nation of the "core" counties allows rejection of only the 1966 and
1972 null hypotheses. Such a complete reversal of the relationship at
this level offers some basis for suspicion of the approach used. Thus,
an alternative approach is examined.

Given the different results achieved by altering the sample em-
ployed in the rho comparisons, the difficulty lies in deciding which
of the two sets of rho values is the more valid expression of the nature
of the relationship between urban structure and changing voting struc-
ture. The rho values computed by eliminating the core counties appear
to offer a more valid summation of electoral change and urban structure
relationship, and a more accurate reflection of reality. But even in
this instance it was necessary to accept the null hypothesis of no
significant relationship in 5 of 12 comparisons. It would involve

considerable risk to speak of any conclusive relationship based on such

136

a record. Thus, in order to further clarify the nature of the urban-
vote relationship, it was decided additional comparisons of urban
structure and vote structure were needed as a means of cross-checking
previous results.

In theory the percentage of urban population provides a valid
measure of urban structure or amount of the total population residing
in an urban environment. But, the level of 2,500 assigned by the
Bureau of Census as urban does perhaps stretch the true character of
large size associated with urban or, more accurately, metropolitan
character. And, although the literature discussed previously related
vote change to urban structure, the implication throughout was that
urban was synonomous with metropolitan or at least larger size. An
urban place of 2,500 certainly does not meet such criteria. Theoreti-
cally, at least, it is feasible to have a majority of a county popula-
tion of 20,000 residing in a few small towns of 2,500 to 5,000 people
and still be highly urban according to Census definition. Yet the
county itself with only 20,000 population may be essentially non-urban
in character. With these shortcomings in mind, it was decided to employ
a second measure of urban structure to re-examine the nature of the
relationship between urban structure and electoral change. In this
case the measure of urban structure was the actual population size of
the county.

Another series of Spearman's rank order comparisons were ac-
complished using the total population of a county as a measure of urban
structure. The resulting rho value are given in Table 15. Again rho
values for both the total sample (Table 15A) and the total sample

minus the core counties (Table 158) are given. The level of rho

137

.uonuoe he ooaamsou

"Homoom

.Ho>aH mo. are an acaoaanaanm.

 

 

 

 

 

comm. «awn. ammo.t «mom. cmma
ammo. oNo. «hmv. «mom. coma

«mmv. «mum. «Nam. «mva. cmma

Aouou mscaz mammemc coauaasmom amuoe
m

ammm. emmm. ebmc.l «mom. coma
«mmm. amc. noon. «mam. coma

«mum. «ohm. camm. emaa. cmma

Aoamemm aeuoav scauoasmom amuoa
tha ooma coma vmma mvma tha moma voma coma omma Nmma mvma

 

amauoumcom amaucooamoum
coauauomsou mo ao>oa

 

d
MBZDOU hm ZOHBdADmOm A4809 UZH>AQ>ZH
mZOmHmfimEOU mom mBzmHUHhmmOU memO MZ¢m m.Z¢Zm¢mmm

ma Manda

138

values for these 24 comparisons (12 elections and 2 samples) was such
that in only one instance, that of the 1964 presidential election in
both samples, was it necessary to accept the null hypothesis of no
significant relationship. And though it is virtually impossible to
compare relative strengths in rho values, it is notable in this case
that the values of rho increased markedly when the total population
ranks were employed.

The relation between urban structure and voting structure in this
case is consistent across levels of competition. At the senatorial
level in both samples the null hypothesis could be rejected for all
five elections. In fact, if the rho values are any indication, the‘
relationship would give every indication of being more intense at the
senatorial level since the values of rho are higher at that level of
competition.

It would appear from this second round of rank order comparisons
that the assertion of an urban centered electoral change in the study
area is confirmed. However, care must be taken to note the nature of
this relationship may be changing. For example, consider the rho
values of the comparisOns made at the presidential level of competition
(Table 15). The rho values progress from lows of .113 and .148 in 1948
to highs of .366 and .427 in 1960. But by 1964, there is apparently no
significant relationship. And by 1972, the relationship apparently
becomes negative as the rho values reach -.087 and -.088 (Table 15).
Interestingly enough, this progression from.weak to strong to no ap-
parent relationship does not hold true for both levels of competition.
The negative relationship of the 1972 presidential election is not

duplicated at the senatorial level where positive values of .298 and

139

.336 suggest a significant positive relationship still exists between
urban structure and voting structure. This suggests two distinct
possibilities. First, that the presidential election of 1972 was an
anomaly and the levels of Republican support were abnormally high due
to the nature of the candidate choice offered. Given this possibility,
it is quite plausible that the same comparisons performed in 1976 would
produce rho values as significantly positive as those of 1968 or 1948
through 1960. In other words, the presidential elections of 1964 (no
significant relationship exists) and 1972 are aberrant and the rela-
tionship between urban structure and voting structure still exists. The
significantly high positive values of the 1972 senatorial level elec-
tions offer support for this possibility.

On the other hand, a second and equally plausible possibility is
that the nature of the relationship between Republican electoral support
and urban structure has moved from a weak but significantly positive
level in earlier elections to a stronger positive and significant
relationship in middle years. And finally in later elections the rela-
tionship is ceasing to exist. The high positive values at the sena-
torial level could be accounted for in terms of the lag between the two
levels of competition found to exist in previous analysis.

On the basis of the information we have thus far, it is difficult
to determine which of these possibilities is more likely. Has the Re-
publican party increased its competitive stance at the presidential
level to the point where it has become electorally viable at all levels
of urban structure within the study area? If so, can similar gains be
expected at the senatorial level? Or is the suggestion of such elec-

toral gain simply a facet of errant elections in 1964 and 1972 at the

140

presidential level? Perhaps a more detailed examination of the urban
structure and vote level relationship might provide answers to such
questions. To gain such detail, we now turn to a more specific analysis

of the relationship.

Specific Patterns of Relationship

In order to probe more closely the nature of the relationship
between urban structure and level of electoral support for the Repub-
lican party, it was decided to carry out an analysis based on the size
of the areal units. The technique involved is basically the same as that
employed by Adamany (1964) and Epstein (1950) in their analyses of
Wisconsin electoral responses. Essentially the technique involves
finding the average level of electoral support for a party at each given
level of population size. Seven population size categories were employed
in this inquiry and the number of counties in each category varied
depending on the census year in question. The category sizes and the
number of counties in size category are provided in Table 16. Two lists
of figures are provided for each cell since both the total sample and
the total sample minus the core counties are examined.

Examination of Table 16 reveals two salient points. First, the
last category of population size, those counties with populations
over 500,000 never includes over three total counties in any given
census year. Hence, care must be employed in interpretation of an
average value computed from such a small number of observations. The
mean is more subject to influence by deviant cases. And second, those
core counties eliminated from the analysis all fall into the 50,000

or less population categories. In fact, most appear to fall in the

141

TABLE 16

NUMBER OF COUNTIES IN EACH POPULATION SIZE CATEGORY

 

Census Year
1950 1960 1970
Total Sample Minus Minus
Sample Core Total Core Total Core

 

< 10,000 101 97 105 101 101 97
10,001- 25,000 180 162 178 161 168 153
25,001- 50,000 110 108 98 95 91 86
50,001—100,000 34 34 39 39 56 56

100,001-250,000 19 19 20 20 22 22
250,001-500,000 3 3 6 6 7 7

>500,000
TOTAL 448 424 448 424 448 424

 

SOURCE: Compiled by author.

142

10,000 to 25,000 category. Thus, even if Republicanism in the period
1948-1972 has been closely associated with larger urban populations, it
has not been so in those areas of traditional Republican support.

The core areas of southern Republican support are and have been largely
rural counties of less than 25,000 population. Interestingly enough,
this corresponds to the national norm of rural or small town Republican

bases of party support.

Size-of—Place Analysis

The results of the detailed breakdown of population size and level
of support are provided in Table 17. To facilitate analysis the data
are summarized in graphical as well as tabular form and these results
are provided in Figure 14.

Examination of these results reveals several distinct patterns of
relationship between size of county and level of vote. In general,
however, for those counties with population less than 50,000, the rela-
tionship is as hypothesized. Counties with larger populations offer
higher levels of support. This holds true for both levels of electoral
competition and for all elections except the presidential elections of
1964 and 1972. In these years the total variation between the mean
levels of support for counties in <10,000 population size category and
those in the 50,000 to 100,000 is less than 4 percent. Apparently in
these two elections electoral support for Republican candidates was
fairly consistent across all population sizes.

Aside from these two elections the general positive relationship
between size and level of support is quite distinctive. The pattern

of increasing electoral support with increasing size proves remarkably

TABLE 17

THE RELATIONSHIP BETWEEN POPULATION SIZE AND LEVEL OF REPUBLICAN VOTE

A

 

Total Sample

 

Senatorial Elections

 

Presidential Elections

 

Category of Total

1952 1956 1960 1964 1968 ‘ 1972 1948 1954 1960 1966 1972

1948

Population Size

 

27.3 30.8 53.6 22.0 74.5 8.0 6.2 7.3 26.9 39.4
15.7 13.6 19.3

28.7

16.1

< 10,000

42.8

35.2

37.2 34.4 39.0 57.3 25.1 73.0

21.9

10,000- 25,000

38.0 35.9 41.1 55.6 28.8 74.4 14.5 14.5 24.5 43.2 48.4

19.1

45.6 41.4 46.4 57.1 32.4 75.1 18.1 21.0 31.8 44.3 50.9

22.8

50,001-100,000
100,001-250,000

50.1 44.9 49.9 57.2 34.0 73.0 14.0 20.4 27.7 45.0 48.6

24.0

43.2 49.7 50.4 52.2 39.9 70.5 15.3 6.0 25.3 51.0 55.6
26.2

25.0

250,001-200,000

>500,000

143

44.9 40.9

17.1

45.6 46.9 52.7 58.3 28.7 65.6 12.8

19.2

ml

Total Sample Minus Core Counties

 

5.5 25.5 38.4

16.2

4.0
9.6
13.9

27.3 25.7 29.4 53.5 20.3 74.7 5.8
10.9

14.3

< 10,000

40.6

32.3

33.5 30.4 35.4 56.8 21.6 72.5

16.9

10,000- 25,000

41.5 46.9

22.8

34.8 40.2 55.6 26.7 74.0 13.8
32.4 75.1

37.4

18.3

25,001- 50,000
50,001-100.000
100,001-250,000

21.0 31.8 44.3 50.9

18.1

45.6 41.4 46.4 57.1

22.8

50.1 44.9 49.9 57.2 34.0 73.0 14.0 20.4 27.7 45.0 48.6
70.5

24.0

6.0 25.3 51.0 55.6

26.2

15.3

43.2 49.7 50.4 52.2 39.9

25.8

250,001-500,000

>500,000

44.9 40.9

17.1

45.6 46.9 52.7 58.3 28.7 65.6 12.8

19.2

 

Compiled by author.

SOURCE:

Percent Of Total Vote

144

FIGURE I4

POPULATION SIZE AND LEVEL OF REPUBLICAN VOTE
(Total Sanple Minus Core Counties)

Presidential Elections Senatorial Elections

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Bo » ‘ ao . I
. I
75 {~ d" a ’\ \ 7S ’ I I,
x 3
70 L \‘fi 70 P t !
\ ' .
\ i
\‘ 1972 i l
65 r 65 - I I
I I ‘
60 h i I I
1964 . I i
55 I i
1960 I A
I l/ '\
50 - / '; \
‘I’ \
1956 : .. ’ \0
1952 45 .-.V \N 1966
I ‘ \
I I 1972
40 i
I
35 I
I
; 30 I
I 1968 .
/ . ‘ \g I954
I 25 1' ‘
I I / i I\. ..
4/ | I I \ '
K, I 20
2° I I948 .- y. i -\
i j' I . e _- I 1960
‘5 / 4 I ‘5 m
E I '. , .' ‘943
I .
‘0 ' I l I to . : .
i I I . .. ~I-
5 B I 1 1 i 5 .° Ir
i I . I 5
I I I
I i I I .
O L l o 1
t 2 3 4 s 6 7 I 2 3 4 s 6 7
< Ioooo '0 000— 25 oat, 90.001- I0000I- 280.00I- >500,000 canon I0 000 25.00I— so oou- Ioo oo~ zso oov >snn Am
25 000 50 000 I00 000 230.000 500.0“ 23.000 90 000 '00 000 250 000 500 00':

Level Of Population Size

145

consistent until the larger categories are considered. At the presi-
dential level of competition, after the 100,000 population size cate~
gory is surpassed, the relationship becomes far less regular and far
less predictable. In 1948, 1956 and 1968 electoral support continues
to rise with pOpulation size until the 500,000 category where support
drops off somewhat. In 1952 electoral support declines after the
100,000 to 250,000 level is surpassed. Only in 1960 does the general
positive relationship hold constant for all categories of population
size.

As with presidential elections, a regular pattern of increasing
support with increasing population size can be traced in all five
senatorial elections. However, at this level of electoral competition
this pattern of increase is broken at the 50,000 to 100,000 mark rather
than the 100,000 to 250,000 breaking point discovered at the presiden-
tial level. After the 50,000 to 100,000 mark is surpassed, the rela-
tionship becomes rather irregular. But, in general, in earlier
senatorial elections (1948-1960) electoral support begins to decline
with larger population sizes. For later elections (1966-1972), electoral
support continues on irregular rise until it drops off sharply for
counties in the 500,000 category. These last two examples are not un-
like the examples of presidential elections of earlier and middle periods
(1948-1960).

If the same relationships are placed in a more longitudinal
dimension, that is to say they are examined over the period from 1948
to the present, an interesting phenomenon is observed. In early elec-
tions (1948-1954) the level of support in lower categories (10,000-

S0,000) is far less than that at larger population size categories

146

(500,000 or more). In fact, except for 1948 and 1968, at the presiden-
tial level of competition the larger size categories are returning vote
levels that lie at least in the competitive electoral range (>45%). This
phenomenon is repeated at the senatorial level, only the mean levels of
support at all size categories are less. This tendency towards com-
petitiveness in larger populated counties and markedly lower levels of
support in smaller populated counties begins to disappear, however,
with increasing more recent elections. In fact at the presidential
level after 1960,1 the differences between the mean levels of electoral
support for the lowest size categories and the highest size categories
are very small. At the senatorial level the differences are larger
but if 1966 and 1972 are any indication, the differences are decreasing.
The results of the size-of-place analysis offer strong evidence
of the validity of the hypothesized positive relationship between popur
lation size2 and electoral support for the Republican party. At both
levels of electoral competition, counties with larger populations
offer higher levels of electoral support than those with smaller popu-
lations. In the longitudinal sense the positive relationship has
apparently become less intense with successive elections. At both
levels of competition, the positive relationship between population size
and mean level of support as measured for those elections in the pre-
1960 period were much stronger than the positive relationships noted

in the elections in the post-1960 period. In fact, by the presidential

 

1Again the 1968 presidential election proves an exception.

2Larger population size is equated with a higher level of urban
population.

147

election of 1972, the population-size categories above 50,000 maintain

a negative relationship, since the mean level of support actually
decreases with increasing size: These results give evidence of the pro-
gression of Republican electoral support. Apparently higher levels of
electoral support for the GOP first came from counties with larger
populations. With successive elections electoral support has gradually
moved down the urban hierarchy into counties with smaller populations.
By 1972, little total difference in mean levels of support separates

the higher categories of population size from the lower.

Although the relationship between size and vote has generally
been positive, the electoral support at larger size categories has
demonstrated marked irregularity. And if a break in the generally
positive relationship of population size and level of support is to be
found, it lies at the largest population categories. The largest
counties have a tendency (increasing in later elections) to offer pro-
portionately less support to the Republican candidates at both levels
of competition. If the most recent elections are a valid indication,
this involves far more than simple irregularity. For in later election
years, the differences between the level of electoral support in the two
top population size categories appear to be increasing. In the 1968
presidential election 11.2 percent points separate the means of two
largest population size categories; in 1972 there is a difference of
nearly 10 percent; and hithe 1972 senatorial elections an average of over
14.0 percent less support was given to GOP senatorial candidates in
counties over 500,000 population than in those counties in the 250,000
to 500,000 size category. While the difference in mean levels of sup-

port between the largest and smallest category sizes is gradually

148

decreasing the support in larger population categories is apparently
falling off somewhat.

Thus, on the one hand the spatial patterns of electoral response
have apparently moved from geographic concentration in one portion of
the study area to more uniformly distributed support that apparently
has a distinct urban bias. That is to say, larger population clusters
apparently facilitated the growth of Republican support. But as this
support has grown, the relationship to the urban structure has changed.
In fact, recent indications suggest a pattern of Republican support in
rural or smaller population size categories and a drop in Republican
support in larger population clusters, a pattern which approximates
the national norm. If this pattern can be verified in subsequent
elections, then scientists may begin to consider giving the South

something other than a unique position in electoral theory.

CHAPTER VI

A PERSPECTIVE OF THE RESULTS

The previous chapters have presented some of the more salient
findings of the analysis of the research problem. Although some
evaluation of these results was offered in conjunction with their
presentation, every attempt was made to tailor such interpretation to
the limitations of the data and the methods of measurement and analysis
employed. In this fashion was the reader left to form a personal evalua-
tion and interpretation of the findings. But just as the reader is
entitled to such opinion, so must the author be allowed license for a
measure of personal evaluation. This final chapter is directed to that
end. I

An effort is made to provide an overall perspective of the results
in relationship to both contemporary political reality and general
electoral theory. Specifically, the results are interpreted in terms
of the regional political framework, their meaning in light of con-

temporary theory, and the prospects for future research efforts.

The Results in a Regional Context
The most dramatic component of the Goldwater outcome
in the Deep South was--change (Cosman, 1966b, 60).
This inquiry has sought to empirically investigate the spatial
mechanisms inherent in changing patterns of electoral response.

149

150

Electoral change has been operationalized in the form of continued
growth of electoral support for the Republican party in the American
South. While the literature provides adequate justification of this
operational definition of change, it provides insufficient evidence of
the spatial mechanisms inherent in such change. This research is
designed to fill that need. The problem involves analysis of the
spatial mechanisms inherent in the changing structure of Republican
electoral support in a selected study area of the American South from
1948 to 1972. Basically what we are saying is that here are the elec-
toral response patterns as they existed in 1948 and this is how they
look at selected intervals over the 24 years since then. Is there
spatial regularity in the response? How has the pattern of response
changed? Is there spatial regularity in the change? And finally,
what is the mechanism of the change? In the latter instance, it be-
comes necessary to draw upon the literature to narrow the limits of
the inquiry. The result is an investigation of the urban structure

as a probable medium of the growth of Republican electoral support

in the study area.

For the most part the results conform to the expectations out-
lined in Chapter II. The distribution of electoral support for the
Republican party has changed. This change is evident in the electoral
parameters. The Republican party has moved into a position where it
can now offer a substantive challenge, at least at the two levels of
electoral eompetition examined in this inquiry, to Democratic hegemony
within the study area. Although these advances have come at ngh_

the presidential and senatorial levels of competition, the level of

151

support is consistently higher1 at the presidential level. An expected
volatility in the levels of support at the presidential level of com-
petition was realized; an expected stability at the senatorial level

was not. While the senatorial level of competition proved less volatile
than the presidential, the stability of the standard deviation and mean
level of support of the five distributions examined was insufficient to
allow acceptance of the hypothesis of greater electoral stability at

the senatorial level of competition.

It is interesting that much of the instability of the senatorial
response patterns appears to be a result of the relatively large in-
crease of electoral support at that level from the 1960 to 1966 elec-
tions. Since this increase is replicated at the presidential level in
the period from 1960 to 1964, the early 1960's take on the appearance of
a turning point in the fate of Republican electoral fortunes in the study
area. While a measure of relative success can be identified in earlier
elections, the 1964 presidential and 1966 senatorial elections mark a
significant break in traditional response patterns. At one level a
Republican presidential candidate wins a majority in the study area.

At the other there are senatorial candidates in every possible sena-
torial race. Thus, while previous discussion has alluded to the national
ineffectiveness of the 1964 southern strategy of the GOP, a re-evalua-
tion of its influence on the South is apparently in order. Evidently

its long range effect on the South proved quite substantial.

 

1One exception involves the 1968 presidential election. The
mean level of support was 26.85 percent; Republican senatorial can-
didates at the comparable date of 1966 gained 36.89 percent of the
total vote.

152

Certainly the coincidence of Republican successes in the South
after 1964,with calculated strategy to win southern votes in that year
is insufficient cause to suggest that the turning point in Republican
fortunes was a consequence of GOP presidential candidate Barry Gold-
water and his "southern strategy." The change could well have come
as a consequence of a long term southern strategy on the part of the
national GOP or more importantly it could have come as a consequence
of effective organization, at least at these two levels of competition,
by southern Republicans. But in either case, Goldwater's candidacy
provided a f0cus for change. For Goldwater was the first Republican
presidential candidate to make victory in the South a major element of
his strategy for winning the presidency. As such, he led the way for
other candidates to follow. As for the influence of effective local
and state organizations, Bernard Cosman (1966a; 1966b) has indicated
that the 1964 candidacy of Goldwater stimulated a grassroots Repub-
licanism in the Deep South--

One very dramatic element that the Goldwater candidacy

left behind in the Deep South was a number of state and

local Republican parties demonstrably stronger than at

any time in the past . . . (Cosman, 1966b, 128).
This is not to imply Goldwater was the cause of Republican success in
later years; rather he can be seen as an agent of Republican electoral
growth. The South was ripe for change and the Goldwater candidacy
presented southern Republicans with an opportunity for effective party
organization (Cosman, 1966a). Goldwater became the catalyst that pre-
cipitated organization activity. And this flurry of organization was

the foundation for post 1964 success. Thus, while Goldwater and states

153

rights might have provided the stimulus for increased GOP support in
the 1964 presidential election, in the wake of Goldwater--in the period
after l964--it is the skeleton of an enduring party mechanism that
precipitated continued success. For only the most generous of Gold-
water proponents would attribute the successes in 1966 senatorial elec-
tions to any coattail effect. How does one coattail a candidate who
has suffered such an overwhelming defeat only two years previous?
Successes such as the 1966 gubernatorial candidacies of Georgia's

Bo Callaway1 and Florida's Claude Kirk, and the 1966 senatorial candi-
dacy of Tennessee's Howard Baker, coming as they did in the off-year
elections two years after Goldwater's attempt, bear only a remote con-
nection to his candidacy.

In the years since Barry Goldwater, Republican candidates have
proven somewhat more successful in attempts at the presidency. And
this success has not proven detrimental to the fate of Republican
candidates for southern political offices. For example, southern GOP
candidates have benefited from the incumbency of a Republican president.
Richard Nixon has practiced his own form of "muted" southern strategy.
The Republican administration's position on crime, busing, civil rights
and other issues salient to southern interests represent an improvement
at least in southern frame of reference, to the earlier Democratic

administrations. This constrained form of southern strategy falls a

 

1Callaway won the Georgia gubernatorial election on the basis
of popular vote. But under Georgia electoral procedure the small
margin of his victory made it necessary for the election to be de-
cided in the Georgia state legislature. Callaway lost the decision
to the Democratic controlled legislature and Lester Maddox.

154

calculated measure short of what Alexander Heard (1952) called "going
to the South," something a national based party could not do and sur-
vive as 1964 has demonstrated. The results of this reserved southern
strategy has certainly not done harm to the Republican organizational
efforts in the South.

But in the final analysis the success of Republicanism in the
study area represents more than simple influence of an unsuccessful
presidential candidate or of an incumbent president. Consider the
fate of Republican candidates since 1966. Tennessee now has two Re-
publican senators--Baker (elected in 1966 and 1972) and Brock (elected
in l970)--and a Republican governor-~Dunn (elected in 1970). North Caro-
lina elected a Republican governor, James Holshouser, Jr., and senator,

Jesse Helms, in 1972.1

And Florida elected Republican Edward Gurney

as its senator in 1968. While these successes may have benefited from
national party policy, they could hardly have come about without
organization-~basic level organization. These successes represent

a measure of grassroots Republican organization designed to endure at
all levels of competitive electoral politics. While the organization
effort may not have reached all_levels of competitive politics in equal
measure, this appears to be only a question of time.

The results of this inquiry suggest that such organization has
developed at upper levels of electoral competition; for at the two
levels of competition examined.Republican inroads have been substantial
since 1964. And although a third party candidate in 1968 mars the

successful record of Republican presidential candidates, the spatial

parameters derived from this inquiry indicate the nature of the setback

 

1Republican candidates appear to do best against non—incumbents.

155

was numerical only. The level of support drapped, but the drop was
shortlived.

Description of the spatial distribution of electoral responses
in the period after 1964 supports the suggestion that the Goldwater
phenomenon was not unique. The legacy of 1964 was change. For in the
period after the 1964 presidential election, support does not retrogress
to traditional concentration in mountain cores. The spatial parameters
indicate that support continues to move toward uniformity in patterns
of electoral support. In fact, from the point of view of Republican
party fortunes, these spatial parameters are one of the most promising
features of the analysis. The hypothesized change from spatial concen-
tration of Republican support to spatial uniformity of support is
confirmed. Migration of mean centers of electoral support from tradi-
tional core areas into Democratic strongholds and a steady increase
in the measure of density toward values indicative of uniformity in
electoral response are both symptomatic of movement towards increasing
support in areas formerly deficient in Republican sentiment. This in-
crease is in agreement with the hypothesis that more uniform patterns
of electoral support occur first at the presidential level of competi-
tion and then, at later intervals, at the senatorial level. Even if,
as in 1968, the electoral support that develops in an area does not
always prove to be competitive, the potential for competition is there.

Thus, into the void left by southern disaffection with the na-
tional Democratic party policy has come the national Republican party.
But the intrusion has not been confined to higher level political
competition. Victories at congressional, gubernatorial and state

legislative levels attest to this. Whether the intrusion represents

156

a realignment of southern partisanship is not a question open to this
inquiry. Although there is a strong indication that massive and
rather permanent shifts in voting responses have occurred with the
study area, the data do not permit any statements relating to atti-
tudinal realignment. However, the findings do lend themselves to a
suggestion of increasing partisan competitiveness over a broader
spatial arena. And nowhere in this analysis did this become more
evident than in the examination of the relationship between electoral
change and the urban structure.

A principal thesis of this inquiry was that increases in the
competitiveness of the GOP in the study area was facilitated by in-
creased population size or urban structure. The hypothesis as derived
from the literature was that electoral change occurred first in larger
urban areas. The examination of spatial regularity in the voting
response provided some support for this thesis. The description of
the voting structure appeared to indicate some coincidence of larger
concentrations of urban population and higher levels of Republican
support. More detailed examination via the medium of tests for rank
order association revealed a positive association between the level of
Republican support and the level of urban population. This positive
association was strengthened by consideration of total population (in
lieu of percentage of urban population) as a surrogate measure of
urban population. But the examination of this association revealed a
curious phenomenon. The association between population size and level
of Republican vote is apparently weakening with successive elections.
At the presidential level in 1964 there is no significant association.

In 1972 the association becomes negative. And these two elections

157

represent the two most significant victories of the GOP in the study
area! Perhaps these elections represent an indication of what is to
come of competitive electoral politics in the South. This does not
imply that in the future Republican candidates will dominate competitive
electoral politics; rather it implies that as the Republican party
becomes more competitive the relationship between urban structure and
GOP electoral support will assume the appearance of the national pat-
tern. Rural areas and smaller cities will support Republican candi-
dates and larger cities and urban areas will support Democratic candi-
dates. In the South this can only come as the fruits of grassroots
Republican party organization are realized. For only through such
organization will the party become competitive throughout the study
area. And this uniformly pervasive competitiveness is apparently what
has begun to develop.

In the earlier elections analyzed in this inquiry uniformity of
support did not exist. If support was to be found it was in the tradi-
tional cores of Republican support or in some counties with larger
populations. As the Republican party began to become increasingly more
competitive, this increase was facilitated by larger populations.

Change took place more rapidly in counties with larger populations.
Support was relatively higher there. But this relationship was developed
without benefit of the organization that apparently occurred after

1964. With the onset of this type of organization, the rural areas

were as susceptible to Republican electoral efforts as were the urban
areas of earlier periods. And with this the character of the urban-vote
relationship changed markedly. The relationship began to resemble

the national norm. Rural counties and smaller cities were as likely

158

as larger cities to offer higher levels of Republican support. In

fact if the movement of large numbers of blacks into larger urban en-
vironments is considered perhaps rural areas were more likely to offer
higher levels of electoral support than larger cities. In this inquiry
detailed analysis of this relationship by category sizes offers a
measure of support for this type of interpretation.

For the earlier elections the size—of_place analysis revealed
a strong positive relationship between population size of a county and
the level of Republican support. But in subsequent elections the size
of place analysis appears to offer conformation of the weakening influ-
ence of urban size. At the presidential level little difference can be
noted between the level of support at the lowest category and that at
the higher categories. The only noticeable trend is that upper cate-
gories of population size appear to be less stable. But at the highest
category, counties of 500,000 or more population, there is a noticeable
tendency toward lower levels of support. Perhaps this signals the ex-
istence of a situation not unlike that of larger northern industrial
cities, where the Democratic party has enjoyed a measure of pre-eminence.
If so, this would indicate the emergence of a New Democratic party in
the South-~one centered on larger urban concentrations.

Thus, while it is possible to accept the hypothesis of a positive
relationship between electoral support for the GOP and level of popula-
tion size in every instance but two (1964 and 1972 presidential elections)
the data suggest the relationship is weakening. But the fact remains
that initial growth of support was facilitated by larger population
size. And only when the party organization became pervasive enough to

reach all sectors of the study area did the relationship decline.

159

The Results in a Theoretical Context

The results of this inquiry support the thesis that electoral
change is facilitated by larger populationinuiurban size. In the study
area, the growth of electoral support for a political party occurred
initially at higher echelons of the urban structure. These findings
support the conclusions of earlier studies such as Epstein (1956) and
Adamany (1962) and in general appear to offer further empirical veri-
fication of the nature of the relation between electoral change and
urban structure. But the coincidence of an erosion in the strength
of the association between change and urban structure raises obvious
questions as to the influence of party structure in such a relationship.
If, as is indicated, the pace of electoral change was stimulated by
fundamental changes in internal party structure then our understanding
of the process of change would be greatly enhanced by closer scrutiny
of the party structure itself. Did a grassroots level organizational
movement provide the stimulus for the increases in support for Repub-
lican candidates in the middle 1960's. If so, from where did the
leaders and activists in this movement come? Have they come from the
urban or rural sector? Have they come from traditional areas of sup-
port or from emergent areas? And the candidates--from where are they
drawn? Probing these questions in the southern context would extend
our understanding of the role that party organization plays in elec-
total change.

However, in working to augment our grasp of the causes of change,
we must not ignore the spatial mechanic: of this change. This inquiry
provides an indication that change is associated with one segment of a

spatial system, i.e. , the upper level of the urban sector. But the

160

actual mechanism of this change remains a matter of conjecture. It

is possible,for example,that electoral change continues to operate
within the upper echelons of the urban hierarchy until a saturation
level is attained. Beyond this point little additional change is
feasible. Hence, some spill over into rural sectors must take place.
Using the example of the growth of Republicanism in the study area, it
can be established that support grew initially at the upper echelons
of the urban structure until continued development of higher levels

of support was no longer feasible. From there increased support could
only come from areas where additional converts might be reached. Thus,
electoral growth spilled over into the remainder of the hierarchy--
into more rural sectors. It is quite feasible that some external
stimulus, party reorganization for example, served as a catalyst for
more rapid movement into rural areas. These results appear to indicate
that it did. But the mechanics of the process of change itself have
yet to be verified.

Although the results of this inquiry strongly hint that electoral
change does indeed operate well in an urban context, the aggregate
nature of the data analysis precludes pinpointing the actual process of
movement of support. The data units are too large. However, even
though voting data are not particularly well suited to classical diffu-
sion theory, this type of spatial movement might well be examined by
isolating a southern metropolis for detailed analysis of the movement.of
support. By resorting to detailed analysis in a single urban area, the
confusion resulting from the overlapping influences of a number of urban
places could be avoided. Through longitudinal analysis of the growth

of electoral support, the pattern of development could be isolated and

161

the influence of the urban concentration upon the surrounding country-
side could be examined. If the more recent aggregate statistics were
to be supplemented with survey data, the spatial structure of voting
preference in a southern city might easily be compared with other
structures in other cities. Thus, could our empirical evidence of the
nature of the Spatial structure of voting preference in an urban en-
vironment be augmented by providing a record of the changes in this
structure over time? Such detail is essential if we are to delve more
deeply into the social and economic components that underlie both the
voting preference and the changes in that preference.

There is, however, the possibility that the examination of the
relationship between voting preference and additional social economic
characteristics need not await the availability of survey data. By
the use of non-parametric statistics, such variables as black popula-
tion, income levels, education levels and professional or employment
status could be incorporated into analyses similar to that conducted
in Chapter 5. One very promising arena would be an examination of the
effect of rates of urban growth on the level of electoral support. Do
faster growing areas give higher levels of support to the southern
GOP? Again such analyses culd be conducted at either the regional
level or at the individual metropolitan level.

Reducing the level at which we examine the spatial structure of
voting preference need not effect the approach used to describe the
spatial distribution of voting preference. In some ways the techniques
employed in this inquiry to describe the voting structure become even
more effective at lower levels. At any rate such description is essen-

tial for purposes of familiarization with the voting structure of a

162

study area, if for no other reason. But they have a practical side as
well. For, as a method of describing the Spatial and numerical distri-
bution of voting response, these techniques of spatial description could
well be a way of accomplishing long range party planning. They are
certainly less costly than continuous surveys. The idea is that alloca-
tion of party funds devoted to improvement of party structure could be
accomplished on the basis of constituency performance. And attempts

to improve the effectiveness of grassroots party organization could be
concentrated in mean range constituencies, thereby increasing the ef-
fectiveness of the organization that must work to win over such con-
stituencies to Republican candidates. Thus, long range planning and
"between election" planning could be accomplished with considerably

less expense than by continuous updating of survey data.

Numerical and spatial descriptive techniques may or may not prove
of practical use to party organization. That question depends on the
feasibility of long range planning based on constituency performances.
But the feasibility of employing such techniques in voting analysis
is more certain. Their effectiveness in providing a more accurate and
objective means of describing voting structure has been demonstrated in
this inquiry. Certainly political geographers have need to explore more
fully the use of such spatial descriptive techniques in electoral
analysis.

In summary, although these results offer continued verification
of existing notions of the relationship between change and urban struc-
ture, they give cause for reassessment of our thinking on the nature of

southern voting response. They offer additional insight into what are

163

apparently new vistas in the consideration of southern electoral be-
havior. The convergence suggested by Converse (1963) a decade ago has
continued. But it has, contrary to his predictions, apparently gone
further than urban environments. It has gone as Strong (1963) has
suggested to something resembling two party behavior. Certainly an
evaluation of the total impact of electoral change on the political
behavior of the electorate in the study area must be tempered somewhat
with the possibility of impermanence. Verification of the permanence
of new patterns of electoral support must await 1974 and 1976. And
verification of partisan attitudes must await adequate longitudinal
survey analysis. But it is plain that if the conclusions drawn from
the data analyzed in this inquiry are sound then we must begin to re-
evaluate our thinking on southern political behavior. Indeed by the
end of the decade of the 1970's, there may be little left of a

reliable form of "southern" political behavior.

BIBLIOGRAPHY

BIBLIOGRAPHY

Abler, Ronald, John S. Adams and Peter Gould. Spatial Organization.
Englewood Cliffs, New Jersey: Prentice-Hall, Inc., 1971. In)

 

Adamany, David. "The Size-of—Place Analysis Reconsidered." Western 1;
Political Quarterly, XVII, No. 17 (1964), 477-487. : 1

Ayers, H. Brandt and Thomas H. Naylor, eds. You Can't Eat Magnolias.
New York: McGraw-Hill Book Co., 1972.

 

 

H
Barnett, J. Ross. "Scale Components in the Diffusion of the Danish k;
Communist Party, 1920-64." Geographical Analysis, V (1973),
35-44.

. "Scale, Process and the Diffusion of a Political Move-
ment." Proceedings of the Association of American Geographers
IV (1972), 9-13.

 

Billet, H. "L'Expression Politique en Gresivaudan et son Interpretation
Géographique." Revue de Géographie Alpine, XLVI, Part 1 (1958),
97-128 as quoted and cited by J. R. V. Prescott, "The Functions
and Methods of Electoral Geography." Annals of the Association of
American Geographers, XL (1959), 296-304.

 

Birdsall, Stephen S. "Preliminary Analysis of the 1968 Wallace Vote
in the Southeast." Southeastern Gepgrgpher, Ix (Nov., 1969),
55-66.

Board, Christopher, Richard Chorley, Peter Haggett and David R. Stoddart,
eds. Progress in Geography, Vol. I. London: Edward Arnold, Ltd.,
1969.

Brunn, Stanley D. "The wallace Presidential Vote in Southeast Cities."
Paper presented at the 1970 Annual Meetings of East Lakes Divi-
sion of the Association of American Geographers.

Brunn, Stanley D. and Gerald L. Ingalls. "The Emergence of Republicanism
in the Urban South." Southeastern Geographer, XII, No. 2 (1972),
133-144.

Brunn, Stanley D., Wayne L. Hoffman and Gerald H. Romsa. "The Defeat

of a Youngstown School Levy: A Study in Urban Political Geog-
raphy." Southeastern Geographer, Ix (Nov., 1969a), 67-79.

164

165

Brunn, Stanley D., Wayne L. Hoffman and Gerald H. Romsa. "Some Spatial
Considerations of the Flint Open Housing Referendum." Proceedings
of the Association of American Geographers, I (1969b), 26-31.

 

Bureau of the Census. County and City Data Book, 1956 (A Statistical
Abstract Supplement). U.S. Government Printing Office, Washington,
D. C., 1957.

Bureau of the Census. County and City Data Book, 1962 (A Statistical
Abstract Supplement). U.S. Government Printing Office, washington,
D. C., 1962.

Bureau of the Census. General Population Characteristics, Alabama.
Census of Population, Washington,D. C.: U.S. Government Printing
Office, 1970.

Bureau of the Census. General ngulation Characteristics, Florida.
Census of Population, Washington, D. C.: U.S. Government Printing
Office, 1970.

Bureau of the Census. General Population Characteristics, Georgia.
Census of Population, Washington, D. C.: U.S. Government Printing
Office, 1970.

Bureau of the Census. General Population Characteristics, Mississipp;,
Census of Population, Washington, D. C.: U.S. Government Printing
Office, 1970.

Bureau of the Census. General Population Characteristics, North
Carolina. Census of Population, Washington, D. C.: U.S.
Government Printing Office, 1970.

Bureau of the Census. General Population CharacteristicsL_South
Carolina. Census of Population, Washington, D. C.: U.S.
Government Printing Office, 1970.

Bureau of the Census. General ngulation Characteristics, Tennessee.
Census of Population, Washington, D. C.: U.S. Government
Printing Office, 1970.

Campbell, Angus, Phillip E. Converse, Warren E. Miller and Donald E.
Stokes. The American Voter. New York: John Wiley and Sons,
Inc., 1966.

Campbell, Angus, Phillip E. Converse, Warren E. Miller and Donald E.
Stokes. Elections and the Political Order. New York: John
Wiley and Sons, Inc., 1967.

Congressional Quarterly Special Report. Presidential Candidates from
1788 to 1964. washington, D. C.: Congressional Quarterly Ser-
vice, 1964, p. 26.

 

 

166

Converse, Phillip E. "A Major Political Realignment in the South?"
Change in the Contemporary South. Ed. Allan P. Sindler. Durham,
N.C.: Duke University Press, 1963, pp. 195-222.

. "On the Possibility of a Major Political Realignment in
the South." Elections and the Political Order. Angus Campbell
et al. New York: John Wiley and Sons, Inc., 1967, pp. 212-242.

 

Converse, Philip E., Warren E. Miller, Jerrold G. Rusk and Arthur C.
Wolfe. "Continuity and Change in American Politics: Parties
and Issues in the 1968 Election." The American Political Science
Review, LXIII (1969), 1083-1105.

Cosman, Bernard. The Case of the Goldwater Delegates: Deep South
Republican Leadership. University, Alabama: Bureau of Public
Administration, University of Alabama, 1966a.

. Five States for Goldwater: Continuity and Change in
Southern Presidential Voting_Patterns. University, Alabama:
University of Alabama Press, 1966b.

 

. "Presidential Republicanism in the South, 1960."
Journal of Politics, XXIV (May, 1962), 303-322.

 

 

Cox, Kevin R. "Residential Relocation and Political Behavior: Con-
ceptual Models and Empirical Tests." Acta Sociologica, XIII
(1970), 40-53.

. "The Spatial Components of Urban Voting Response Sur-
faces." Economic Geography, XLVII (1971), 27-35.

 

. "The Spatial Structuring of Information Flow and Parti-
san Attitudes." Quantitative Ecological Analysis in the Social
Sciences. Eds. Mattei Dogan and Stein Rokkan. Cambridge:
M.I.T. Press, 1968a, pp. 157—186.

 

. "Suburbs and Voting Behavior in the London Metropolitan
Area." Annals of Association of American Geographers, LVIII
(1968b), 111-127.

 

"The Voting Decision in a Spatial Context." Progress

in Geography, Vol. 1. Eds. Christopher Board, Richard J. Chorley,
Peter Haggett and David R. Stoddart. London: Edward Arnold

Ltd., 1969, pp. 81-117.

 

 

Dogan, Mattei and Stein Rokkan, eds. Quantitative Ecological Analysis
in the Social Sciences. Cambridge: M.I.T. Press, 1968.

 

Editorialization in Support of Al Smith. The Watchman. Greensboro,
Alabama, Oct. 25, 1928 as quoted in Donald S. Strong. "Durable
Republicanism in the South." Change in the Contremporary South.
Ed. Allan P. Sundler. Durham, N.C. Duke University Press,

1963, p. 175.

 

167

Epstein, Leon D. "Size of Place and the Division of the Two-Party
Vote in Wisconsin." Western Political Quarterly, IX (1956),
138-150.

Gould, Peter R. Spatial Diffusion. washington, D. C.: Association of
American Geographers, Commission on College Geography, Resource
Paper No. 4, 1961.

Grantham, Dewey W., Jr. The Democratic South. Athens: University of
Georgia Press, 1963.

Grantham, Dewey W., Jr., ed. The South and the Sectional Image. New
York: Harper and Row, Publishers, 1967.

Grantham, Dewey W., Jr. "The South and the Politics of Sectionalism."
The South and the Sectional Imag_. Ed. Dewey Grantham, Jr.
New York: Harper and Row, Publishers, 1967, pp. 36-55.

Haring, Lloyd L. "An Analysis of the Spatial Aspects of Voting Be-
havior in Tennessee." Unpublished Ph.D. Dissertation, State
University of Iowa, 1959.

Havard, William C., ed. The Changing_Politics of the South. Baton
Rouge: The Louisiana State University Press, 1972.

Heard, Alexander. A Two Party South? Chapel Hill: University of
North Carolina Press, 1952.

Hoffman, Wayne L. "A Comparative Analysis of Two Urban Non-Partisan
Referendums: A Factor Analysis Solution." Unpublished Ph.D.
Dissertation, University of Florida, 1970.

Hultquist, John, John Holmes and Lawrence A. Brown. CENTRO: A Program
for Centrographic Measures. Discussion Paper No. 21, Department
of Geography, The Ohio State University, No date.

Kasperson, Roger E. "On Suburbia and VOting Behavior." Annals of
Association of American Geographers, LIX (1969a), 405-410.

"Ward Systems and Urban Politics." Southeastern
Geographer, IX (Nov., 1969b), 17-25.

 

Kasperson, Roger E. and Julian V. Minghi. The Structure of Political
Geggraphy. Chicago: Aldine Publishing Company, 1969.

Key, V. 0. Southern Politics in State and Nation. New York: Random
House-Vintage Books, 1949.

Krebheil, Edward. "Geographical Influences in British Elections."
Geographical Review, II (March, 1916), 419-432.

168

Lewis, P. F. "Impact of Negro Migration on the Electoral Geography
of Flint, Michigan, 1932-1962: A Cartographic Analysis."
Annals of Association of American Geographers, LV (1965), 1-25.

Lewis, Peter and G. E. Skipworth. Some Geograpgical and Statistical
Aspects of the Distribution of Votes in Recent General Elections.
University of Hull, England, 1966.

Mack, Raymond W., ed. The Changing South. United States: Aldine
Publishing Co., 1970.

Masters, Nicholas and Deil Wright. "Trends and Variations in the Two
Party VOte. The Case of Michigan." American Political Science
Review, LII (1958), 1078-1090.

 

Matthews, Donald R. and James W. Prothro. Negroes and the New Southern
Politics. New York: Harcourt, Brace & WOrld, 1966.

 

. "Southern Images of Political Parties: An Analysis of
Negro and White Attitudes." Journal of Politics, XXVI (Feb.,
1964), 82-111.

 

Morris, John. Nonparametric Statistics. East Lansing: Computer
Institute for Social Science Research, Technical Report 40,
1967a.

. Rank Correlation Coefficients. East Lansing: Computer
Institute for Social Science Research, Technical Report 47,
1967b.

 

Morris, William, ed. The American Heritage Dictionary of the English
Language. Boston: American Heritage Publishing Co., Inc. and
Houghton Mifflin Co., 1971.

Murphey, Reg and Hal Gulliver. The Southern Strategy. New York:
Charles Scribner's Sons, 1971.

Orum, A. M. and E. W. McCrane. "Class, Tradition, and Partisan Align-
ments in a Southern Urban Electorate." Journal of Politics,
XXXII (Feb., 1970), 156-176.

Phillips, Kevin P. The Emerging Republican Majority. New Rochelle,
N.Y.: Arlington House, 1970.

Prescott, J. R. V. "Electoral Studies in Political Geography."
The Structure of Political Geography. Ed. Roger E. Kasperson and
Julian V. Minghi. Chicago: Aldine Publishing Co., 1969,
pp. 376-383.

. "The Functions and Methods of Electoral Geography."
Annals of Association of American Geographers, XLIX (1959),
269-304.

 

169

Prothro, J. W., E. Q. Campbell and C. M. Grigg. "Two-Party Voting in
the South: Class vs. Party Identification." American Political
Science Review, LII (March, 1958), 131-139.

Reynolds, David R. "A Spatial Model for Analyzing Voting Behavior."
Acta Sociologica, XII (1969), 122-131.

Reynolds, D. R. and J. C. Archer. "An Inquiry into the Spatial Basis of
Electoral Geography." Discussion Paper Series. Paper No. 11.
Department of Geography, University of Iowa, 1969.

Roberts, M. C. and K. W. Rumage. "The Spatial Variations in Urban Left
Wing Vbting in England and Wales in 1951." Annals of Association
of American Geographers, LV (1965), 161-178.

 

Rossiter, Clinton. Parties and Politics in America. Ithaca, N.Y.:
Cornell University Press, 1960.

Sack, Rdbert David. "Geography, Geometry, and Explanation." Annals
of Association of American Geographers, LXII (March, 1972),
61-78.

Scammon, Richard M. America Votes. Vol. 1. Governmental Affairs
Institute. New York: The MacMillan Company, 1956.

. America Votes. Vol. 2. Governmental Affairs Institute.
New York: The MacMillan Company, 1958.

. America Votes. Vol. 3. Governmental Affairs Institute.
Pittsburgh: University of Pittsburgh Press, 1959.

. America Votes. Vol. 4. Governmental Affairs Institute.
Pittsburgh: University of Pittsburgh Press, 1962.

. America Votes. Vol. 5. Governmental Affairs Institute.
Pittsburgh: University of Pittsburgh Press, 1964.

. America Votes. Vol. 6. Governmental Affairs Institute.
Washington, D. C.: Congressional Quarterly, 1966.

. America Votes. Vol. 7. Governmental Affairs Institute.
Washington, D. C.: Congressional Quarterly, 1968.

. America Votes. Vol. 8. Governmental Affairs Institute.
Washington, D. C.: Congressional Quarterly, 1970.

. America Votes. Vol. 9. Governmental Affairs Institute.
Washington, D. C.: Congressional Quarterly, 1972.

Sindler, Allan P., ed. Charge in the Contempprary South. Durham, N. C.:
Duke University Press, 1963.

170

Strong, Donald S. "Durable Republicanism in the South." Change in the
Contemporary South. Ed. Allan P. Sindler. Durham, N. C.: Duke
University Press, 1963, pp. 174-194.

 

. The 1952 Presidential Election in the South. University,
Alabama: Bureau of Public Administration, 1955.

 

"The Presidential Election in the South, 1952." Journal
of Politics,XVII (Aug., 1955), 343-389.

 

 

. Urban Republicanism. University, Alabama: Bureau of
Public Administration, 1960.

 

 

Twelve Southerners. I'll Take My Stand: the South and Agrarian
Tradition. New York: Harper and Brothers, 1930.

United States Bureau of the Census. Statistical Abstract of the United
States. 72nd-93rd editions. Washington, D. C.

Wittick, Robert I. GEOSYS: An Information System for the Description
and Analysis of Spatial Data. East Lansing, Michigan: Computer
Institute for Social Science Research Technical Report 73-6,
1973.

. Some General Statistics Prggrams Used in Spatial Analysis.
East Lansing, Michigan: Computer Institute for Social Science
Research, Technical Report 71-1.

 

Woodward C. Vann. Origins of the New South. Baton Rouge: Louisiana
State University Press, 1951.

APPENDICES

APPENDIX 1

A LIST OF COUNTIES IN THE STUDY AREA

State of Alabama (67 Counties)

APPENDIX 1

A LIST OF COUNTIES IN THE STUDY AREA

 

Autauga
Baldwin
Barbour
Bibb
Blount
Bullock
Butler
Calhoun
Chambers
Cherokee
Chilton
Choctaw
Clarke
Clay
Cleburne
Coffee
Colbert
Conecuh
Coosa
Covington
Crenshaw
Cullman
Dale
Dallas
Dekalb
Elmore
Escambia
Etowah
Fayette
Franklin
Geneva
Greene
Hale
Henry
Houston
Jackson
Jefferson
Lamar

Lauderdale
Lawrence
Lee
Limestone
Lowndes
Macon
Madison
Marengo
Marion
Marshall
Mobile
Monroe
Montgomery
Morgan
Perry
Pickens
Pike
Randolph
Russell
St.C1air
Shelby
Sumter
Talladega
Tallapoosa
Tuscaloosa
Walker
Washington
Wilcox
Winston

171

State of Florida (34 Counties)

Alachua
Baker

Bay
Bradford
Calhoun
Clay
Columbia
Dixie
Duval
Escambia
Flagler
Franklin
Gadsden
Gilchrist
Gulf
Hamilton
Holmes
Jackson
Jefferson
Lafayette
Leon
Liberty
Madison
Nassau
Okaloosa
Putnam
St. Johns
Santa Rosa
Suwannee
Taylor
Union
Wakulla
Walton
Washington

172

State of Georgia (159 Counties)

Appling
Atkinson
Bacon
Baker
Baldwin
Banks
Barrow
Bartow
Ben Hill
Berrien
Bibb
Bleckley
Brantley
Brooks
Bryan
Bulloch
Burke
Butts
Calhoun
Camden
Candler
Carroll
Catoosa
Charlton
Chatham
Chattahoochee
Chattooga
Cherokee
Clarke
Clay
Clayton
Clinch
Cobb
Coffee
Colquitt
Columbia
Cook
Coweta
Crawford
Crisp
Dade
Dawson
Decatur
De Kalb
Dodge
Dooly
Doughtery
Douglas
Early
Echols
Effingham

Elbert
Emanuel
Evans
Fannin
Fayette
Floyd
Forsyth
Franklin
Fulton
Gilmer
Glascock
Glynn
Gordon
Grady
Greene
Gwinnet
Habersham
Hall
Hancock
Haralson
Harris
Hart
Heard
Henry
Houston
Irwin
Jackson
Jasper
Jeff Davis
Jefferson
Jenkins
Johnson
Jones
Lamar
Lanier
Laurens
Lee
Liberty
Lincoln
Long
Lowndes
Lumpkin
Macon
Madison
Marion
McDuffle
McIntosh
Meriwether
Miller
Mitchell
Monroe

Montgomery
Morgan
Murray
Muscogee
Newton
Oconee
Oglethorpe
Paulding
Peach
Pickens
Pierce
Pike
Polk
Pulaski
Putnam
Quitman
Rabun
Randolph
Richmond
Rockdale
Schley
Screven
Seminole
Spalding
Stephens
Stewart
Sumter
Talbot
Taliaferro
Tattnall
Taylor
Telfair
Terrell
Thomas
Tift
Toombs
Towns
Treutlen
Troup
Turner
Twiggs
Union
Upson
Walker
Walton
Ware
Warren
Washington
Wayne
Webster
Wheeler

White
Whitfield
Wilcox
Wilkes
Wilkinson
Worth

173

State of Mississippi (18 Counties)

Alcorn
Clarke
George
Greene
Harrison
Itawamba
Jackson
Kemper

Lauderdale

Lee
Lowndes
Monroe
Noxubee
Perry
Prentiss
Stone

Tishomingo

Wayne

State of North Carolina (46 Counties)

 

Alexander
Alleghany
Anson
Ashe
Avery
Buncombe
Burke
Cabarrus
Caldwell
Catawba
Cherokee
Clay
Cleveland
Davidson
Davie
Forsyth
Gaston
Graham
Guilford
Haywood
Henderson
Iredell
Jackson
Lincoln
Macon
Madison
McDowell

Mecklenburg

Mitchell

Montgomery

Polk
Randolph
Richmond
Rockingham
Rowan
Rutherford
Stanley
Stokes
Surry
Swain
Transylvania
Union
Watauga
Wilkes
Yadkin
Yancey

State of South Carolina
(42 Counties)

Abbeville
Aiken
Allendale
Anderson
Bamberg
Barnwell
Beaufort
Berkeley
Calhoun
Charleston
Cherokee
Chester
Chesterfield
Clarendon
Colleton
Darlington
Dorchester
Edgefield
Fairfield
Florence
Greenville
Greenwood
Hampton
Jasper
Kershaw
Lancaster
Laurens
Lee
Lexington
McCormick
Marlboro
Newberry
Oconee
Orangeburg
Pickens
Richland
Saluda
Spartanburg
Sumter
Union
Williamsburg
York

174

State of Tennessee (82 Counties)

Anderson
Bedford
Benton
Bledsoe
Blount
Bradley
Campbell
Cannon
Carroll
Carter
Cheatham
Chester
Claiborne
Clay
Cocke
Coffee
Cumberland
Davidson
Decatur
DeKalb
Dickson
Fentress
Franklin
Giles
Grainger
Greene
Grundy
Hamblen
Hamilton
Hancock
Hardin
Hawkins
Henderson
Henry
Hickman
Houston
Humphreys
Jackson
Jefferson
Johnson
Knox
Lawrence
Lewis
Lincoln
Loudon
McMinn
McNairy
Macon
Marion
Marshall

Maury

Meigs
Monroe
Montgomery
Moore
Morgan
Overton
Perry
Pickett
Polk
Putnam
Rhea

Roane
Robertson
Rutherford
Scott
Sequatchie
Sevier
Smith
Stewart
Sullivan
Sumner
Trousdale
Unicoi
Union
VanBuren
warren
washington
Wayne
White
Williamson
Wilson

APPENDIX 2

CANDIDATES IN PRESIDENTIAL AND SENATORIAL ELECTIONS
1948-1972

APPENDIX 2

CANDIDATES IN PRESIDENTIAL AND SENATORIAL ELECTIONS
1948-1972

Presidential Elections

 

 

 

Date Republican Candidate Democratic Candidate
19481 Thomas E. Dewey Harry S. Truman

1952 Dwight D. Eisenhower Adlai E. Stevenson
1956 Dwight D. Eisenhower Adlai E. Stevenson
1960 Richard M. Nixon John F. Kennedy

1964 Barry M. Goldwater Lyndon B. Johnson
19682 Richard M. Nixon Hubert H. Humphrey
1972 Richard M. Nixon George McGovern

Senatorial Elections

 

 

Alabama
% Total % Total
Date Republican Candidate Vote Democratic Candidate Vote
1948 Paul G. Parsons 16.0 John J. Sparkman 84.0
1950 Lister Hill 76.5
1954 Foy J. Guin 17.5 John J. Sparkman 82.5
1956 Lister Hill 100.0
1960 Julian Elgin 29.8 John J. Sparkman 70.2
1962 James D. Martin 49.1 Lister Hill 50.9
1966 John Grenier 39.0 John J. Sparkman 60.1
1968 Perry Hooper 22.0 James B. Allen 70.0
1972 Winston M. Blount 33.0 John J. Sparkman 62.3

 

1 . . .
A third major candidate was J. Strom Thurmond of the States Rights
party.

2A third major candidate was George C. Wallace of the American
Independent party.

175

176

Senatorial Elections

 

Florida
% Total % Total
Date Republican Candidate Vote Democratic Candidate Vote
1950 John P. Booth 23.7 George A. Smathers 76.2
1952 Spessard L. Hollard 99.8
1956 George A. Smathers 100.0
1958 Leland Hyzer 28.8 Spessard L. Hollard 71.2
1962 Emerson H. Rupert 30.0 George A. Smathers 70.0
1964 Claude R. Kirk 36.0 Spessard L. Holland 63.9
1968 Edward J. Gurney 55.9 Leroy Collins 44.1
1970 William C. Cramer 46.1 Lawton Chiles 53.9
Georgia
1948 Richard B. Russell 99.9
1950 Walter F. George 100.0
1954 Richard 8. Russell 100.0
1956 Herman Talmadge 100.0
1960 Richard B. Russell 99.9
1962 Herman Talmadge 100.0
1966 Richard B. Russell 99.9
1968 Earl E. Patton 22.5 Herman Talmadge 77.5
1972 Fletcher Thompson 46.0 Sam Nunn 54.0
Mississippi
1948 James O. Eastland 100.0
1952 John Stennis 100.0
1954 James A. White 4.4 James O. Eastland 95.6
1958 John Stennis 100.0
1960 Joe A. Moore 8.2 James O. Eastland 91.8
1964 John Stennis 100.0
1966 Prentiss Walker 26.7 James O. Eastland 65.6
1970 John Stennis 88.4
1972 Gil Carmichael 38.7 James O. Eastland 58.1
North Carolina
1948 John A. Wilkinson 28.8 J. M. Broughton 70.7
19508 E. L. Gavin 32.6 Willis Smith 67.0
1950 Halsey B. Leavitt 31.3 Clyde R. Hoey 68.7
19543 Sam J. Ervin 100.0
1954 Paul C. West 34.1 William Kerr Scott 65.9
1956 Joel A. Johnson 33.4 Sam J. Ervin 66.6
19588 Richard C. Clarke 30.0 Everett B. Jordan 70.0
1960 Kyle Hayes 38.6 Everett 8. Jordan 61.4
1962 Claude L. Greene 39.6 Sam J. Ervin 60.4
1966 John S. Shallcross 44.4 Everett B. Jordan 55.6
1968 Robert V. Somers 39.4 Sam J. Ervin 60.6
1972 Jesse A. Helms 69.5 Nick Galifianakis 28.9

 

 

 

 

177

Senatorial Elections

South Carolina

 

% Total % Total
Date Republican Candidate Vote Democratic Candidate Vote
1948 Bates J. Gerald 3.6 Burnet R. Maybank 96.4
1950 Olin D. Johnston 99.9
1954 Edgar A. Brown 36.8
19568 Strom Thurmond 100.0
1956 Leon P. Crawford 17.8 Olin D. Johnston 82.0
1960 Strom Thurmond 100.0
1962 W. D. Workman 42.8 Olin D. Johnston 57.2
19668 Marshall Parker 48.7 Ernest F. Hollings 51.3
1966 Strom Thurmond 62.2 Bradley Morrah 37.8
1968 Marshall Parker 38.1 Ernest F. Hollings 61.9
1972 Strom Thurmond 63.3 Eugene N. Zeigler 36.7
Tennessee
1948 Carroll B. Reece 33.4 Estes Kefauver 65.3
1952 Hobart F. Atkins 20.9 Albert Gore 74.2
1954 Thomas D. Wall 30.0 Estes Kefauver 70.0
1958 Hobart F. Atkins 19.0 Albert Gore 79.0
1960 Bradley A. Frazier 28.2 Estes Kefauver 71.7
19648 Howard H. Baker, Jr. 47.4 Ross Bass 52.1
1964 Daniel H. Kuykendall 46.6 Albert Gore 53.6
1966 Howard H. Baker, Jr. 55.7 Frank G. Clement 44.3
1970 William E. Brock 51.3 Albert Gore 47.4
1972 Howard H. Baker, Jr. 61.5 Ray Blanton 37.8

S Elections designated by a S following the year were for short terms

 

to fill vacancies.

 

HICHIGAN STATE UNIV. LIBRARIES
WI ”WI Ill W “l "W II" V“ l1 IW W W Ml N WWI
31293010877516