Sectionalism in North Carolina counties

Material Information

Sectionalism in North Carolina counties
Wolf, Matthew D
Publication Date:
Physical Description:
xii, 93 leaves : ; 28 cm


Subjects / Keywords:
Since 1951 ( fast )
Political parties -- North Carolina ( lcsh )
Politics, Practical -- North Carolina ( lcsh )
Political parties ( fast )
Politics and government ( fast )
Politics, Practical ( fast )
Politics and government -- North Carolina -- 1951- ( lcsh )
North Carolina ( fast )
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )


Includes bibliographical references (leaves 90-93).
General Note:
Department of Political Science
Statement of Responsibility:
by Matthew D. Wolf.

Record Information

Source Institution:
University of Colorado Denver
Holding Location:
Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
710808014 ( OCLC )
LD1193.L64 2010m W64 ( lcc )

Full Text
Matthew D. Wolf
B.B.A, The University of Iowa 1982
A thesis submitted to the
University of Colorado Denver
in partial fulfillment of
the requirements for the degree of
Master of Arts
Political Science

This thesis for the degree of Master of Arts
by Matthew D. Wolf
has been approved by
Tony Robinson
Mike Cummings

Wolf, Matthew D. (M.A., Political Science)
Sectionalism in North Carolina Counties
Thesis directed by Associate Professor Anthony Robinson
Utilizing long-term measures of county-level voting behavior, the change in
Democratic strength (CDS) was estimated for North Carolina and each of its one
hundred counties. While the state overall, experienced only a quarter percent increase
in DS, individual counties ranged from a decrease of 11.78 percent to an increase of
6.24 percent. Independent variables were tested for their ability to predict CDS using
multiple regression. In addition to the entire population of counties, several samples
of counties were tested in the attempt to tease out any characteristic differences
between those with the most radical CDS {Polar Counties) and Mainstream Counties
(CDS closer to the mean).
The results of these tests suggest that Bishops (2008) sorting of the population by
ideology is likely occurring in North Carolina. The South is the sole major US
region, as defined by the US Bureau of the Census, with substantial net domestic
migration during the past two decades, and North Carolina has been one of its leaders.
As inter-state and intra-state migrants choose where to live, they are sorting in
varying degrees by partisan identification, resulting in wide variances in county-level
changes in Democratic strength.
The evidence also supports Glaesers observation that segregation now reflects less
the confinement of Blacks to certain counties than it does the flight of Whites to
increasingly-White enclaves. Also supported here is the contention of Walters (2001)
that the Black vote has been highly influential in the South and will likely continue to
increase in importance. Changes in Black and Hispanic shares of county population,
voter turnout, and creative class share of jobs consistently predicted CDS in most
North Carolina counties.
This abstract accurately represents the content of the candidates thesis. I recommend
its publication.
"Anthony Robinson

This thesis is dedicated to my parents, David Lawrence Wolf and Carol Ann Zeman
Wolf, both of whom instilled in me an appetite for learning. It is also dedicated to my
wife, Joya Darby Wolf, who has been an inspiration to me and a faithful supporter of
this project from the beginning.

Tony Robinson, my thesis advisor, provided the inspiration to pursue inquiry into this
sector of political science by his leadership in, and enthusiasm for, electoral politics.
He has been a patient and instructive mentor and deserves my sincere gratitude. I
also wish to thank the other members of my committee, Mike Cummings and Michael
Berry, for their valuable input.

LIST OF FIGURES ......................................................... xi
LIST OF TABLES .......................................................... xii
1. INTRODUCTION ..................................................... 1
2. CHANGES IN SUB-STATE VOTING BEHAVIOR ............................ 11
The Method of Measurement .................................. 11
The Results ................................................ 12
Analysis of Results ........................................ 14
3. MULTIPLE-REGRESSION MODELS ...................................... 17
The Model .................................................. 17
Testing the Tails of the Distribution Curve ......... 17
Testing the Center of the Distribution Curve ........ 18
The Independent Variables .................................. 19
Migration Variables .................................... 20
Demographic and Socio-Economic Variables ............... 22
The Results of Multiple-Regression Tests ................... 25
All Counties ........................................... 25
Testing the Tails of the Distribution Curve ......... 26
Testing the Center of the Distribution Curve ........... 33
Summary of Test Results ................................ 35

4. POLAR AND MAINSTREAM COUNTIES ................................... 42
Mainstream Counties ......................................... 42
Gross Migration ......................................... 42
Racial Composition ...................................... 45
Creative-Class Jobs ..................................... 47
Polar Counties .............................................. 50
Education ............................................... 50
Voter Turnout ........................................... 51
White Flight ............................................ 53
Polar Red Counties .......................................... 60
Cherokee County ......................................... 60
Columbus County ......................................... 61
Polar Blue Counties ......................................... 63
Guilford County ......................................... 63
Edgecombe County ........................................ 65
Mainstream Counties ......................................... 67
Craven County ........................................... 67
Iredell County .......................................... 68
Summary ..................................................... 71

Migration ................................................... 73
The Theory of Racially Unbalanced Migration ........... 73
The Changing Face of the Carpetbagger ................... 75
Polarization ................................................ 77
Black Voters: A Democratic Force in the South ............... 79
Characteristics of the Black Electorate ................. 80
Implications ................................................ 82
BIBLIOGRAPHY ................................................................ 90

Figure Description Page
1 Distribution of Counties by Rate of Change in Democratic Strength........... 8
1 Distribution of Counties by Rate of Change in Democratic Strength........... 13

Table Description Page
1 North Carolinas Change in Democratic Strength.......................... 12
2 Extreme Changes in Democratic Strength in North Carolina................ 14
3 Populations and Sub-populations in Tests of Distribution Tails.......... 18
4 Sub-populations in Test of Center Distribution.......................... 19
5 Multiple-regression of All Counties (N=100) Predicting CDS.............. 25
6 Key to Short Notation Descriptions of Independent Variables............. 27
7 Analysis of the Tails of the CDS Distribution with Multiple-Regression 28
8 Analysis of the Center of the CDS Distribution with Multiple-regression. 34
9 Summary of Multiple-regression Model Results............................ 35
10 Selected County Data.................................................... 58

There is such a hunger for positive, non-partisan solutions.
John Hickenlooper, Mayor of Denver, Colorado1
Colorado Governor-elect and Denver Mayor Hickenloopers comment alludes
to the present state of American politics; one in which the country appears polarized
into blue and red states, counties and cities. Cahn and Carbone (2010) describe the
two contrasting paradigms of Red Families and Blue Families2 that portend
potential changes in the demographic profiles of these groups of Americans and
examine various laws regarding abortion, gay marriage, and other hot-button issues
that some States taking up the flag of either extreme have enacted.
Some would have us believe that American partisan polarization is more a
characteristic of elites than the population taken as a whole (Hacker and Pierson
2005; Lasch 1995). Echoing Fiorina (2005) and regarding North Carolina in
1 This was spoken at a fundraiser for his campaign for Governor of Colorado, March 17, 2010.
2 This is literally the title of their book.

particular, Prysby (2008) notes activists appear to be more polarized than do
voters, with the elected officials falling somewhere between these two groups in their
polarization. (p. 75), though Fiorina (2005) notes that many of todays elected
officials originally got into politics through the party activist. The Annenberg
Democracy Project (2007) enumerates various facets of the polarization of Americas
elites, including branches of government, legislators, political party brass, and
corporate executives. One could explore the extent of this kind of elite polarization
by following the workings of legislatures, campaigns, and activists in the daily news,
but what about the people themselves, the electorate? Are they polarized, and how
would one measure it?
Kaufmann, Petrocik, and Shaw (2008) argue, based on their analysis of fifty
years of ANES3 data, that while American elites are indeed polarized, the vast
majority of the electorate actually agree on many more issues than they disagree
upon, and are not strongly divided on most issues, except at the extremes. Fiorina
(2005) seems to makes this same argument with less data and a better framework, but
allows that:
In the past few years there have been increasing indications (see Chapter 1)
that high-level political actors are beginning to believe in the distorted picture
of American politics that they have helped to paint. This development
threatens to make the distorted picture a self-fulfilling prophesy as a polarized
political abandons any effort to reach out toward the great middle of the
country, (p. x)
3 American National Election Studies and (presumably) its predecessor, the Survey Research Center

Bishop (2008) uses presidential voting patterns, demographic changes, and
socio-economic data, from every US county4 to advance what Kaufmann, Petrocik,
and Shaw (2008) describe as the position of their opponents:
By this account, the American public had become increasingly divided
into warring camps that could be identified as much by geography as
by politics. The popular Red state Blue state metaphor suggested
that they also had organized themselves on competing terrains. The
unspoken (but often implied) subtext to this story was that the political
divide between Republicans and Democrats was so palpable and
personal that partisans chose to physically segregate themselves from
all but their ideological brethren (p. 49).
Part of the evidence for such claims of increasing polarization at the grass-
roots level is the growing number of counties that give one or the other political party
overwhelming support. By Bishops (2008) measure, thirty eight percent of US
counties experienced landslide election results5 in the 1976 presidential election, and
this number grew steadily to sixty percent by the 2004 presidential election. Based
on such patterns polarization theorists find that Americans are actively sorting
themselves geographically according to their values and beliefs, ideological
preferences, and/or partisan identity.
4 This includes equivalent administrative units.
5 A landslide victory represents a margin of victory greater than twenty percent.

A county-level comparison of the Presidential vote change from 2004 to 2008
seems to indicate that this trend continues. A distinct arc of red symbolizing
counties that voted more Republican in 2008 than in 2004 runs east from Oklahoma
and east Texas through the Ozark Mountains, across the Tennessee Valley and
northern Alabama, and then turns northeast along the Appalachian Mountains (a
somewhat broken red line also runs along the gulf coast). While the vast majority of
swing voters were voting more Democratic presumably in opposition to the
unpopular Bush administration the voters of these geographically concentrated
counties were bucking the national trend, and voting more Republican. This pattern
alone seems to provide evidence of spatial polarization of voters.
What might account for such patterns of spatial polarization? Gimpel and
Schucknecht (2003), analyzing twelve U.S. states in search of explanations for
political sectionalism, point to migration as a likely source of changing electoral
behavior, but do not provide the detail necessary to postulate much farther. Indeed, in
his review of their book, Knotts (2006) recommends that Future work should ...
look more closely at partisan changes within particular states. (p. 597)
As though following this advice, Robinson and Noriega (2010) analyzed the
eight states comprising the Rocky Mountain West, at the county level, and show that
in-migrants originating from places with higher Democratic strength than the county

of destination account for growing Democratic strength across the region. Much of
the influx appears to be coming from the Pacific coast.6 They also show correlations
between increases in Democratic strength and increases in creative-class jobs, the
share of population that is single and without children, and migration rates. This
evidence supports the theory that migrants likely will maintain their partisan identity
after moving, but is it isolated to the Rocky Mountain West?
Other studies also suggest that voter migration patterns might account for
changing ideology at the county level. Campbell et al (1966) show how resilient
partisan identification is over the long term. Analyzing specific cases of
congressional seats won in the previous presidential election, and lost by the party in
the following mid-term election, he demonstrates that a significant source of this
reverse is caused by former cross-over voters returning to the party with which they
usually identify. Campbell et al (1966) note, ... it is empirically clear that in the
lengthening period of our observation, vote shifts have not been accompanied by
conversion but rather have been followed by actual return to the party of original
choice. (p. 15)7 Due to the rare nature of significant levels of voter conversion from
one major party to the other, and the ability of voters to retain their party
6 This conclusion is indicated in US Census Bureau, Population Estimates Program, 2004, and
supported in greater detail in Robinson and Noriega 2010.
7 Ironically, later in the same book, he predicts the realignment of the South, which is the one big
recent exception to the rule that voters rarely convert.

identification after relocating geographically, even when they move to neighborhoods
where their party identification places them in a small minority, several authors,
including Converse (1966), Gimpel et al (2003), and Bishop (2008), point to high
rates of voter migration as the likely cause of spatial changes in voting behavior.
This study was designed to isolate sectionalism and polarization of the
electorate at the sub-state level, in a single state in order to assess the severity of
these factors and to attempt to explain in greater detail some of the variables, whether
having to do with migration, demographic, or other factors, that predict and/or
characterize these phenomena in greater detail and/or with a narrower geographic
focus than has been provided by other studies. Thus, when looking for evidence of
sorting and polarization outside of the West, the South seems a logical target, as it is
the one major region of the United States, as defined by the US Census Bureau that
has experienced significant net domestic in-migration in recent times8. Between 1990
and 2004 the US Census Bureau estimates that the South averaged annual net
domestic migration of 372,000 compared to about 21,000 for the West and losses of
295,000 in the Northeast and 98,000 in the Midwest.9 One might expect that, if
*The US Census Bureau divides the nation into four major regions; Northeast, Midwest, South and
West. Although the West has seen significant net migration in recent times, it is largely driven by
international migration, which has replaced the loss of residents who have moved from the Pacific
Coast to the Rocky Mountain region. Domestic migration between these two sub-regions is considered
intra-regional migration for the purposes of this study. Note also that the study is focused only on
domestic migration to the exclusion of international migration, which includes many residents who are
not naturalized and/or have no prior ideological or party affiliation.
9 US Census Bureau, Population Estimates Program, 2004.

Americans were ideologically sorting, electoral results would show up in the region
to which they are migrating (or in regions they were exiting, such as the Northeast),
because many in-migrants (and out-migrants) have made a choice about where to live
(Glaeser 2008). In the case of Bishop (2008) this is argued to be an ideological
Of the net domestic migrants to the South, during the fourteen-year period
ending with 2004, some seventy three percent settled in the South Atlantic division,
which includes Delaware, Maryland, the District of Columbia, Virginia, West
Virginia, North Carolina, South Carolina, Georgia, and Florida.10 Of these states,
North Carolina, which ranked fifth in the nation for net domestic in-migration
between 2000 and 2004,11 was chosen for this sub-state study. In addition to its
influx of voters, North Carolina has three large and growing urban centers, a
mountain region that has a long history of Republican voters, and one hundred
counties, which divides the state into enough geographic entities to potentially exhibit
distinct characteristics.
Using a measure of long-term voting behavior similar to that used by
Robinson and Noriega (2010) (described in detail below), it was determined that
10 Ibid.
11 Ibid.

Democratic strength for the state of North Carolina increased by one quarter of one
percent between the past two decades (the average Democratic strength in elections of
2002 through 2008, compared to the average Democratic strength in elections
between 1992 and 2000, inclusive)12. When assessed by county-level changes,
however, changes in Democratic strength were found to range from a loss of 11.78
percent to a gain of 6.24 percent. The distribution of counties by the rate of change
calculated formed a normal curve with a mean of negative 1.99 percent at the center,
and a slightly longer tail on the negative side (the greatest loss of Democratic
strength) as shown in Figure 2.
Number jo
Counties 5
CT> .....................................
od ' *
Rate of Change in Dem Strength (Percent) 1990s to 2000s
North Carolina Counties
Rate of Change in Dem Strength
Figure 1 Distribution of Counties by Rate of Change in Democratic Strength.
12 Voting data from presidential, gubernatorial, and US senate races was used.

These data suggest the existence of a geographic cleavage of voters based on
their party identification. In order to find out more about what is happening in North
Carolina counties, and how well this polarization hypothesis fits the data, several
multiple-regression models were created, using change in Democratic strength as the
dependent variable and several demographic, socio-economic, and migration
measures as independent variables. All one hundred counties were first tested, and
then the population of counties was reduced several times, so that each subsequent
group of counties tested represented a smaller subset of the total. Groups of counties
with increasingly extreme measures of change in Democratic strength (CDS) in both
directions were tested in one model, and counties with increasingly less extreme
measures of CDS were tested in another.
The results of these tests showed that CDS varied significantly with changes
in the share of Blacks and Hispanics in the county population, and with changes in the
creative-class share of jobs and changes in voter turnout. These relationships were
significant in most of the county groups tested. As the sample size was decreased
(and became increasingly representative of extreme CDS) some other relationships
were identified, and these relationships became stronger and/or more significant as
the CDS became more extreme. The counties where CDS is extreme exhibit different
behavior from counties with less extreme CDS, and these behavioral differences may
indicate two different kinds of ideological geographic polarization, one that is

somewhat mild and may be largely accidental, and another that may be more of an
active ideological polarization related to race and education.
Variables that predict CDS in polar North Carolina counties are the
attainment of college degree or higher, change in voter turnout, and net White
migration rates (between 1995 and 2000). This latter variable (net White migration
rate) is negatively correlated with the dependent variable (change in Democratic
strength) and suggests that White flight (primarily from urban to rural counties) may
be a significant force in the changing, sub-state voting behavior noted above, both by
increasing the conservative White population of rural counties and by decreasing that
same population in urban counties.
The results of these tests are consistent with Bishops (2008) sorting, but
seem to show clear signs of polarization only in the few counties at the extremes of
change in Democratic Strength. This trend is especially true of the four counties that
are represented at the negative-CDS tail of the distribution, which experienced loss of
Democratic strength (DS) at rates from 9.81 to 11.78 percent. The results of these
tests make it clear that some sorting is occurring in the majority of North Carolina
counties and that this sorting seems most pronounced in a minority of counties across
the state, yet as will be shown below there is little reason to believe that extreme
measures of political polarity have infected the majority of North Carolinas counties.

The Method of Measurement
In the chapter entitled The Concept of a Normal Vote, Campbell et al
(1966) instructed the reader to ...consider any particular vote cast by any particular
group the nation as a whole or some subpopulation as consisting of a long-term
and a short-term component (p. 14). They then show that, while many voters will
cross over and vote for another party due to the circumstances present in any
particular election, most voters -- even those who loosely identify with a party will
return consistently to the party with which they generally identify. Therefore, when
comparing actual votes in various elections, the analyst must be able to identify the
short-term effects in order to solve for any long-term changes in the underlying
partisan identification of the electorate. In order to analyze a long-term trend, such as
geographic partisan polarization, one must either use long-term partisan identification
data, such as that contained in American National Election Studies (ANES) survey
data, successor to the Survey Research Center survey data utilized by Campbell et al
(1966), or use long-term measures of voting behavior as a proxy for partisan
identification, with the expectation that over a series of elections the short-term
components will tend toward mean reversion, and will generally cancel each other

1212This latter method was used to estimate the Change in Democratic
Strength in North Carolina counties between the past two decades. Specifically,
Democratic Strength (DS) was measured as the total votes for Democratic candidates
for US President, State Governor, and US Senate,13 as a ratio of all votes for both
Democratic and Republican candidates (the two-party vote). The Change in DS
(CDS) is calculated by subtracting average DS during the period 1992 to 2000,
inclusive, from average DS during the period 2002 to 2008.
The Results
Using this method, to analyze voting data from the North Carolina State
Elections Board, DS and CDS gross values were calculated for the state of North
Carolina as shown in Table 1.
Table 1 North Carolinas Change in Democratic Strength
Source: North Carolina State Board of Elections 1992 2000 2002 2008
North Carolina 50.10% 50.36% 0.25%
The state-level changes clearly suggest that there is little meaningful electoral
change in North Carolina as a whole, in terms of changing partisan strength. When
13 Note that, in a departure from the method of Robinson and Noriega (2010), voting data for US
House races was excluded, due to the difficulty of sorting the vote of gerrymandered districts into
county level voting data. This approach also eliminates the confounding effects of any Dixiecrat
politicians conservatives who have not left the Democratic Party, but gamer a significant share of
Republican votes which are more likely to exist in races for the US House, as well as state and local
elections, than in the U S. Senate, Presidential, and Gubernatorial races.

calculated by county for all one hundred North Carolina counties, however, the data
resulted in a normal distribution of counties around a mean of negative 1.99% CDS,
but with a range from negative 11.78% to positive 6.24%, as presented in Figure 1.
North Carolina Counties
Rate of Change in Dem Strength
fc'b'V'vdPdRJVJVIbJb* ?> S> S*
f\. (N- '>
Rate of Change in Dem Strength (Percent) 1990s to 2000s
Figure 1 Distribution of Counties by Rate of Change in Democratic Strength.
Urban-rural differences in the nature of CDS are immediately apparent when
the county-level CDS distribution is compared to the overall statewide CDS; since the
overall state CDS is positive 0.25%, yet the average county experienced negative
1.99% CDS, there must be significantly higher population in the fewer counties that
showed positive CDS.

Analysis of Results
In order to get a better look at the polar extremes of CDS, a subset of counties
was selected for analysis based on CDS, and various categories of demographic,
socio-economic, and migration data. The counties selected consisted of those with
calculated CDS values at or above two percent CDS growth and those at or below
four percent CDS decline. These value limits are equidistant from the mean of two
percent CDS decline, and generate a sub-population of counties that includes nearly
equal numbers of counties in which CDS increased versus counties where it
Table 2 Extreme Changes in Democratic Strength in North Carolina counties.

Highest Decline DS State Average Highest Increase DS
Number of Counties 11 100 12
CDS < -6% -1.99% >2%
Avg. County Population 31,958 75,132 191,195
Avg. Change in Creative Job Share 2.12% 3.28% 3.75%
Avg. Change in Black Share Pop. -2.22% -0.79% 1.63%
Avg. Change in Hispanic Share 1.47% 2.84% 3.30%
Avg. Change in Voter Turnout 7.83% 9.98% 12.44%
Avg. Net White Migration 10.32% 6.46% 3.12%
Avg. Net Non-White Migration 3.93% 9.78% 12.40%
Avg. Net Black Migration 1.45% 4.73% 6.81%
This exercise rendered the results shown in Table 2, and a cursory
examination of the data contained therein turns up several correlations. First, the
counties that experienced the most extreme increases in DS are, on average, six times
more populous than those where the most extreme decrease in DS occurred.
Evidence of an urban versus rural partisan cleavage is indicated, and is very

pronounced in counties with extreme rates of CDS. Second, the average change in
Black and Hispanic share of the population is greater than the state average in
counties where DS increased and lower than the state average in counties where DS
decreased. This observation seems to support the sorting theory of Bishop (2008), at
least in these counties, which exhibit the most extreme CDS in the state.
Third, voter turnout appears significantly higher where DS increased and
lower where DS decreased than the average county. Voter turnout is generally
recognized as a short-term component of any particular vote (Campbell et al 1966),
with the most obvious example being the change in turnout between any mid-term
election and presidential election that follows or precedes it. There are, however,
other factors that cause the turnout in any particular election to vary, including the
public appeal of the candidates, and pressing issues of the day, often the most
effective of which include wars and economic recessions. The voter turnout measure
utilized here to compare with CDS is a long-term one, and indicates that changes in
voter turnout are correlated with CDS in North Carolina counties.
Fourth and finally, the change in creative-class share of jobs appears to be
growing faster than average in counties where CDS increased and slower than
average in counties where CDS decreased. Robinson and Noriega (2010) found a
similar relationship in the West, where DS was increasing in counties that also

experienced increases in creative-class jobs. This aspect will be discussed in more
detail below.
In summary, the voters of North Carolina may be sorting ideologically and
geographically statewide. The measurement of CDS used here indicates that
unbalanced racial migration likely plays some part in the change in voting behavior in
the twenty-three counties in which CDS was most extreme, both in terms of increase
and decrease in CDS. In this subset of counties, CDS is increasing in metropolitan
counties, where creative-class jobs are on the increase, as is long-term voter turnout,
and is declining in more rural counties where Creative-class jobs and voter turnout
are declining, or increasing at rates that are substantially lower than the state average.
While this information is reasonably informative, especially regarding the
most politically polarizing counties, a more comprehensive analysis of the data may
provide a better picture of what is causing the Changes in Democratic Strength
observed in North Carolinas counties.

The Model
The regression function in Microsoft Excels add-in Analysis ToolPak was
used to test the ability of several independent variables to predict CDS.14 In addition
to testing the entire population of North Carolina counties (N=100), several samples
of this population were tested.
Testing the Tails of the Distribution Curve
In order to observe any changes in the predictive capability of the independent
variables as one considers groups of counties that contain increasingly more extreme
measures (in both directions from the mean) of CDS, the first two sets of counties
removed from the total population of counties were (in a first model) those with CDS
measures between negative 1.00% and negative 1.99%, and (in a second model) those
between negative 2.00% and negative 2.99%. These two groups of counties straddle
the mean of negative 1.99%, and removing these mean counties from the regression
14 Though it has some shortcomings, this tool is easily acquired and used. In the future, ArcGIS, SPSS,
and other applications/tools may be used to test for multicolinearity, spatial autocorrelation,
heteroscedasticity, and other factors that may provide more insight into the results obtained, but such
additional tests are outside the scope of this study.

model effectively isolates a series of counties constituting roughly equal tails of the
CDS distribution curve (shown in Figure 2). In this manner successive pairs of
county groups were removed from the previous sample, so that the number of
counties decreased and the average CDS measure in the sample became increasingly
more extreme (again, in both directions from the mean).15 This process resulted in
the population and samples shown in Table 3. Note that model E includes the same
23 counties that were summarized in Table 2. Each of these subgroups of counties
was tested with a separate regression model.
Table 3 Populations and Sub-populations in Tests of Distribution Tails
CDS Included
CDS Excluded
-11.78 to 6.24
n 73 -0.99 to 6.24 -3.00 to-11.78 -1.00 to -2.99
c 56 0.00 to 6.24 -4.00 to-11.78 0.00 to -3.99
D 36 1.00 to 6.24 -5.00 to-11.78 0.99 to -4.99
E 23 2.00 to 6.24 -6.00 to -11.78 1.99 to -5.99
Testing the Center of the Distribution Curve
In search of predictive relationships between independent variables and CDS
in counties where CDS was less extreme, the process described above was reversed,
so that the tails of the curve were successively excluded from the sample populations
tested. In addition, two more subsets were tested: one in which all CDS values were
15 This could obviously have been done by making the cuts at each standard deviation, or half standard
deviation; however, this seems as arbitrary as the method used here, which resulted in a nearly uniform
reduction of the size of the sample of counties tested, several county groups (models), and interesting

positive and one in which all CDS were negative. These county groups are shown in
Table 4.
Table 4 Sub-populations in Test of Center Distribution and
Positive vs. Negative CDS
-0.99 to 6.24 -3.00 to-11.78
2.00 to 6.24 -6.00 to-11.78
1.00 to 6.24 -5.00 to-11.78
0.00 to 6.24
-4.00 to -11.78
-1.00 to -2.99
0.00 to -3.99
1.99 to -5.99
0.99 to -4.99
The Independent Variables
The independent variables tested for predictive value represent two broad
categories relating to two different, but not necessarily mutually exclusive, schools of
thought regarding spatial partisan change. One such school of thought is that partisan
change is the result of migration (Campbell et al, 1965; Gimpel et al, 2003; and
Robinson and Noriega, 2010), the other school of thought is that partisan change is
driven by changes in the underlying socio-economic characteristics of the voting
population. Because migration can itself cause changes in the underlying socio-
economic characteristics of a county, is seems likely that some blend of these factors
may be the cause of the CDS observed here.

Migration Variables
As noted previously, many authors point to migration as one of the most likely
suspects underlying partisan change in geographic locales (sectionalism), whether
neighborhood, census block, or region. The South is the one major region of the
United States that has experienced sustained, positive net domestic migration during
the past two decades. This migration stream includes many Black and Hispanic
voters, and these two racial subsets of migrants are generally strong and reliable
identifiers with the Democratic party North Carolina Blacks, in particular voted
eighty-eight percent and ninety-five percent Democratic, in the 2004 and 2008
presidential elections, respectively. Thus, it appears likely that racially unbalanced
migration could influence the voting behavior of North Carolina counties. Because of
this strong partisan identification, most of the migration variables control for race (in
this case White or non-White).16
Several migration variables were calculated for each North Carolina county in
search of variables which might predict CDS. These variables include the following
County Gross Migration / 2000 County Population
Net White Migration / 2000 County Population
Net Non-White Migration / 2000 Non-White County Population
Net Non-White Intrastate Migration / County Net Migration
16 Note that, throughout this report, White refers to non-Hispanic White and non-White includes

Net Non-White Interstate Migration / County Net Migration
Non-White / County Intrastate Migration
Non-White / County Interstate Migration
These migration variables were all calculated using US Census Bureau, 2000
Census figures, including Gross and Net Migration Tabulations and County-to-
County Migration Flow Data (1995 to 2000), which is derived from the long-form
questions regarding where the respondent lived in 1995. The migration variables are
therefore not exact proxies for migration that might have effected electoral change,
due to the temporal mismatch with the time period for which CDS was measured in
this study. This census data covers migration patterns roughly in the middle of the
term during which CDS was measured (1992 through 2008). Due to the possibility
that migration patterns might have changed during this period, they leave some
potential for error in the test results, but it is reasonable to assume that migration
patterns measured by this census data is a good proxy for migration patterns that
likely prevailed in the entire period measured by this study. In any case, the
migration patterns measured over five years can be expected to influence voter
behavior over a longer period of time, as most of the new migrants remain in their
new home counties and vote in successive elections.

Demographic and Socio-Economic Variables
As noted in the discussion of migration variables, Black and Hispanic voters
identify strongly with the Democratic Party. Thus, variables representing the change
in county-level racial demographics between the 1990 and 2000 Census were
calculated, resulting in the following variables:
Change in Black Share of County Population
Change in Hispanic Share of County Population
Change in White Share of County Population
Change in Non-White Share of County Population
Educational Attainment
Cahn and Carbone (2010) contrast Red versus Blue family paradigms, in which
the two types of families exhibit distinctly different characteristics in terms of
household income, educational attainment, divorce rate, and age at marriage and first
pregnancy. Household income and changes in household income between 1990 and
2000 were tested and did not add to the predictive capacity of any regression models
and were therefore excluded. Cahn and Carbone (2010) indicate that these variables
are interrelated. For example, marriage and/or having children at a younger age 17
17 Red represents conservative families, who generally identify with and vote Republican, while Blue
represents liberal families who generally identify with and vote Democrat.

proves to be a significant barrier to completing higher levels of education, which
ultimately limits household income and subsequently can limit the educational
attainment of successive generations. For this reason, and due to the location of many
colleges and large universities in North Carolina, the study focuses on educational
attainment by testing the following variables against CDS:
Percent of County Population with Less Than a High School Degree
Percent of County Population with a High School, But Not College Degree
Percent of County Population with a College Degree or higher
These values are calculated as of the 2000 Census.
Change in Voter turnout
Turnout is a classic component in many scholars theories of changes in
partisan voting behavior.18 As noted previously, the voter turnout values used here
are calculated using the actual votes cast in presidential elections during the same
span of time as that used to calculate CDS.19 In each county the average votes cast in
each presidential election were divided by the eligible population of the county (those
18 and older) estimated for that year by the US Bureau of the Census to arrive at a
rate of voter turnout for the election. Then an average for the elections in 1992,1996
18 See for example, Campbell et al (1966), Gimpel and Schuknecht (2003), and Bishop (2008)
19 The exception here is that only presidential election turnout was used, whereas the CDS calculations
use mid-term election results also. Eliminating mid-term election year turnout patterns was done to
avoid the short-term swings in turnout that normally happen in these off-year elections.

and 2000 was calculated, as was one for the elections of 2004 and 2008. Change in
turnout equals the difference between these two average decade values.
Change in Creative-class Jobs
Creative-class job growth was found by Robinson and Noriega (2010) to be
predictive of CDS in the Rocky Mountain West, and appears to be a good variable to
test for predictive value in North Carolina. As most Creative-class Jobs require a
college degree, this variable may be co-linear with educational attainment. The
concept of the Creative-class and Creative-class jobs (Ray and Anderson, 2000 and
Florida, 2002) will be discussed in more detail in the analysis of the results of these
This variable was created using US Department of Agriculture data available
online; creative-class share of each countys jobs in 1990 was subtracted from the
same ratio reported for 2000, to arrive at change in creative-class share jobs by
20 It appears that, although the creative class is a more recent concept, once one identifies which job
codes should be included in this class, data can be mined on creative class jobs as far back as
consistent information about jobs exists.

The Results of Multiple-Regression Tests
All Counties
First examined was the entire set of one hundred North Carolina counties
(N=100) to see if any of the variables described in the previous section predict CDS
statewide. The results of this first multiple-regression estimation are presented in
Table 5 below:
Table 5 Multiple-regression of All Counties (N=100) Predicting CDS
Independent Variables Coefficients S. E.
County Gross Migration / 2000 County Population 0.0475 0.03405
Net White Migration / 2000 County Population -0.0495 0.08200
Net Non-White Migration / 2000 Non-White County Population -0.0569 * 0.03231
Net Non-White Intrastate Migration / County Net Migration 0.0050 0.00381
Net Non-White Interstate Migration / County Net Migration 0.0051 0.00549
Non-White / County Intrastate Migration -0.0023 * 0.00118
Non-White / County Interstate Migration -0.0233 * 0.00901
Demographic and Socio-Economic:
Change in Black Share of County Population 0.8582 * 0.37778
Change in Hispanic Share of County Population 0.4096 * 0.24573
Change in White Share of County Population 1.9268 1.16848
Change in Non-White Share of County Population 2.0467 1.23132
Educational Attainment:
% of County Population With < High School Degree 0.0095 0.01503
% of County Population With High School Degree, Not College 0.0020 0.00984
% of County Population With College Degree or More 0.0088 0.00873
Change in Voter Turnout 0.1242 * 0.07398
Change in Creative-class Jobs 0.7019 *** 0.15259
Adjusted R Squared 0.5384
Standard Error 0.0236
F-Statistic 8.2159
*p < .10; < .01; ***p < .001

Adjusted R Squared is presented due to the small size of the samples (and this
is the largest sample of counties tested in this study) and its value of 0.53836
indicates that the model explains, or more accurately, predicts a significant share of
the CDS in North Carolina counties. Change in Creative-class jobs, Voter Turnout,
and county share of population represented by Blacks and Hispanics all predict CDS,
and are positively correlated (an increase in the Black share of population in a county
would be expected to predict an increase in Democratic Strength), with particular
significance found in the predictive capacity of the change in Creative-class share of
jobs in all counties.
The migration variables also show some significance in predicting CDS, but
have low coefficients and contradicting signs. The negative signs contradict the
expected positive correlation between non-White migration and CDS, so that non-
White migration was expected to be an underlying cause of the changes in Black and
Hispanic shares of county populations, which as noted in Table 5, are a strong
predictor of CDS. Perhaps analysis of smaller groups of counties, those harboring
more extreme values of CDS, will shed some light on these results.
Testing the Tails of the Distribution Curve
In order to present the results of tests of smaller samples of North Carolina
counties, so that they are more readily comparable to each other, the descriptions of

the independent variables were shortened, so that they and all the data may be
presented in fewer tables. A key relating the longer descriptions contained in Table 5
to the shortened versions presented in the tables that follow is provided in Table 6.
Table 6 Key to Short Notation Descriptions of Independent Variables
Long Description Short Notation
County Gross Migration / 2000 County Population M Gross
Net White Migration / 2000 County Population M Net White
Net Non-White Migration / 2000 Non-White County Population MNetNW 1
Net Non-White Intrastate Migration / County Net Migration M Net NW 2
Net Non-White Interstate Migration / County Net Migration M Net NW 3
Non-White / County Intrastate Migration M NW 1
Non-White / County Interstate Migration MNW2
Demographic and Socio-Economic:
Change in Black Share of County Population CCS Black
Change in Hispanic Share of County Population CCS Hisp.
Change in White Share of County Population CCS White
Change in Non-White Share of County Population CCSNW
Educational Attainment:
% of County Population With < High School Degree % of County Population With High School Degree, Not College HS +
% of County Population With College Degree or More College +
Change in Voter Turnout Chg TO
Change in Creative-class Jobs ChgCCSJ
The sample of counties tested was reduced from 100 to 73 (model B), 56
(model C), 35 (model D) and 23 (model E). The county count was reduced by
progressively eliminating the counties represented by bars in the center of Figure 1,
beginning with the two on either side of the mean (equal to -1.99% CDS). The first
two sets of counties eliminated had changes in Democratic strength of -1.0 to -1.99
and -2.0 to -2.99, a total of twenty seven counties, resulting in a population of
counties numbering seventy three. The next counties eliminated are represented by

the bars outside of these two on the graph, and so forth; working away from the mean.
This method resulted in testing the characteristics of samples of counties that exhibit
progressively more extreme values of Changes in Democratic Strength. Results of
the multiple-regression estimations for each of the county subsets is shown in Table 7
(note that the model identity matches that of Table 3 above).
Table 7 Analysis of the Tails of the CDS Distribution with Multiple-regression
Model A B C D E
Sample Size N=100 N = 73 N = : 56 Z II K N = 23
Variables Coeff. P Coeff. P Coeff. P Coeff. p Coeff. P
M Gross 0.0475 0.0368 -0.0004 -0.0993 -0.3932 *
M Net White -0.0495 -0.0326 -0.2311 * -0.3090 * -0.8697 *
MNetNW 1 -0.0569 * -0.0732 -0.0715 -0.0561 0.6444
M Net NW 2 0.0050 0.0059 0.0115 * 0.0061 -0.1072
MNetNW 3 0.0051 0.0066 0.0108 0.0015 -0.0548
M NW 1 -0.0023 * -0.0021 * -0.0026 * -0.0016 0.0029
MNW2 -0.0233 * -0.0192 * -0.0226 -0.0157 -0.0558
CCS Black 0.8582 * 1.1166 * 1.6869 * 2.2451 * -1.7456
CCS Hisp. 0.4096 * 0.5152 * 0.7487 * 1.9540 ** 0.5671
CCS White 1.9268 1.6023 2.1929 3.0736 9.8478
CCS NW 2.0467 1.5225 1.26941 1.2477 11.4273
< HS 0.0095 -0.0058 -0.0269 0.0466 0.2046
HS + 0.0020 -0.0157 -0.0327 0.0089 0.1034
College + 0.0088 0.0358 0.0460 * 0.2426 * 0.4678 *
ChgTO 0.1242 * 0.1296 * 0.1633 * 0.3554 * 1.0206 *
Chg CCSJ 0.7019 *** 0.7817 ** 1.2466 *** 1.5371 ** 0.5297
Adj. R Sq. 0.5384 0.6075 0.6870 0.6993 0.8363
S. E. 0.0236 0.0255 0.0257 0.0298 0.0257
F-Statistic 8.2159 7.9655 8.5458 6.0862 8.0247
*p < .10; **p < .01; ***p < .001
The standard error for each variable was omitted from Table 7 in order to
present the regression estimates of each model proximate to each other for
comparison. Before considering each group of variables in turn, it is important to

note that adjusted R squared values indicate that the predictive power of this model
gets progressively stronger as the number of counties in the model is decreased.21 In
addition, the standard error, which gradually increased as N was decreased, actually
decreased between models D and E. Similarly, the F-statistic, which had decreased
significantly by model D, due in part to the decrease in sample size, increased
significantly between models D and E. In light of the steady increase in adjusted R
squared, especially the rather large increase noted between models D and E, one
might get the impression that certain predictive variables in the model become
stronger in the counties which report the most radical CDS (model E). Recall that
this includes radical CDS in both directions; both increase and decrease. In other
words, if net White migration is shown to be significantly negatively correlated with
CDS, this means that not only are White voters moving to counties where DS is
declining, but also that White voters are moving out of counties where DS is
The results of this test of the tails can be separated into two different types,
one in which the predictive variables seem to be characteristic of the state as a whole,
showing up in all or most of the county groups, and another type in which there
appears to be little or no significance for the population as a whole, but significance
21 R Squared for groups A, B, C, D, and E were 0.6130, 0.6896, 0.7575, 0.8367, and 0.9554,
22 As is the case with distribution of CDS, the mean of a particular independent variable does not
necessarily occur at the zero value. Thus the model, in calculating an independent variables deviation
from the mean of all county values does not necessarily result in gain versus loss straddling the mean;
it may result in negative or positive values that lie on opposite sides of the mean. Bertie County, North
Carolina, for example, registered a decline in DS of 1.23 percent, but deviates from the mean (the
average North Carolina county had al.99 percent loss of DS) in the direction of positive DS growth,
because its DS declined at a rate that was slower than that of the average North Carolina county.

as one progresses to more extreme values of CDS (or models representing counties
with more extreme CDS). For example, change in creative-class jobs and Hispanic
share of population are significant in models A through D, but lose their significance
in model E, while net White migration and the attainment of a college degree or more
show no significance in models A and B, but become significant in models C, D, and
E. The former will be referred to as the Mainstream Variables, because they
generally predict CDS in most North Carolina counties. The latter as will be referred
to as the Polar Variables of North Carolinas partisan cleavage, because they
predict CDS in the counties where the most extreme CDS occurs.
Mainstream Characteristics
The variables that show strong predictive value throughout the series of
county groups include change in Black and Hispanic share of county population,
change in Creative-class share of jobs, and change in voter turnout. Coefficients of
the first three of these increase steadily as N is decreased, then suddenly decline in
model E (change in Black Share coefficient becomes negative in model E). In
contrast, the change in voter-turnout coefficient increases steadily from model A
through model D, but then it increases substantially in model E, giving it properties of
both mainstream and polar variables. The fact that coefficients for all these variables
increase somewhat as one approaches the poles of CDS may make them appear
similar to polar characteristics; however, most of them are not significant in the most
extreme set of counties (model E).

Polar Variables
The variables that show predictive strength only in the counties with more
extreme measures of CDS are the proportion of the county population that has a
college degree or higher, change in voter turnout, and two migration rate variables -
gross migration and net White migration. Both of the migration variables are
negatively correlated with CDS, suggesting that White flight from metropolitan
counties to rural counties is effectively decreasing White share of population in
metropolitan counties at the same time it is increasing White share of the population
of certain non-metropolitan counties, and this same pattern helps predict the changing
voting behavior of a subset of North Carolina counties. In fact this migration pattern
may be driving these counties to the most extreme partisan change recorded in the
state (by the measure of CDS utilized here).
As noted above, the counties included in model E include two distinct samples
of North Carolina counties: one group represents the counties with the most extreme
decreases in CDS and the other group represents counties with the most extreme
increases in CDS. The latter group has on average six times the population of the
former group, which suggests White flight from urban to rural counties. Glaeser
(2008) and Fuguitt (2008) show that brain drain-the continuing migration of more
educated people from rural to urban, or metropolitan locales-has continued into the
time frame under study here, and this is consistent with these findings. Regarding the

significance of change in voter turnout in model E, one might speculate that it is
caused by either, or a combination, of several factors. Those who take White flight
may have voted with their feet, so to speak, and may be less likely to make it to the
ballot box. Those in urban, or metropolitan centers, may be influenced by
contextual factors described by Bishop (2008) and Gimpel (2003) which may be more
effective in areas that have higher population density, and are result in some voter
conversion and increasing voter turnout in these locales. Finally, the high and
growing share of urban and metropolitan areas represented by Black and Hispanic -
voters who identify strongly with the Democratic party are turning out in increasing
numbers and having a significant effect on overall turnout rates, as is observed by
Walters (2001).
Before exploring these characteristics of sub-state North Carolina and their
implications for the present and future in depth, the next section will investigate the
center of the distribution of counties based on CDS values, in other words, the
counties with progressively less extreme changes in DS will now be investigated, in
search of variables that seem to drive electoral change in the mainstream of the state,
as opposed to the extremes. 23
23 In other words, they just want to get away to borrow the term used in Southwest Airlines

Testing the Center of the Distribution Curve
Because the stratification of North Carolina counties in the Test of the
Tails produced interesting results with the potential to predict much of the observed
values in calculated CDS in extreme counties, it was decided to reverse the process of
exclusion of counties from the population, such that, as the county sample size is
decreased, the model tested increasingly represents counties with less radical CDS
values. Multiple-regression results are presented in Table 8. Note that the county
models are the same as those presented in Tables 3 and 4.
The predictive power of these models, as indicated by the low adjusted R
squared and F-statistic values, decreases quickly as one lops off successive strata at
the extremities of the CDS distribution. Yet the model still has a story to tell,
especially when the results are contrasted with the Test the of Tails. While the
Creative-class share of jobs remained significant until everything on the chart falls
apart (in models H and I, which have negative adjusted R squared, and F-statistic
below 1.0, and show little significant predictive power in any individual variable) the
change in relationship between CDS and voter turnout (and several other independent
variables) disappeared once the first strata of the tails were removed.

Table 8 Analysis of the Center of the CDS Distribution with Multiple-regression
Model A F G
Sample Size N=100 N = 77 N = 64
Variables Coeff. P Coeff. p Coeff. p
M Gross 0.0475 0.0801 * 0.0925 **
M Net White -0.0495 -0.0618 -0.1182
MNetNW 1 -0.0569 * -0.0391 -0.0144
M Net NW 2 0.0050 0.0048 -0.0002
M Net NW 3 0.0051 0.0066 0.0003
MNW 1 -0.0023 * -0.0023 -0.0020 *
MNW2 -0.0233 * -0.0070 -0.0054
CCS Black 0.8582 * 0.5134 0.3175
CCS Hisp. 0.4096 * 0.1397 0.0936
CCS White 1.9268 1.7263 0.7828
CCS NW 2.0467 1.7198 0.8149
< HS 0.0095 -0.0057 0.0062
HS + 0.0020 -0.0031 0.0059
College + 0.0088 -0.0030 -0.0052
Chg TO 0.1242 * -0.0682 0.0109
ChgCCSJ 0.7019 *** 0.4724 *** 0.3521 **
Adj. R Sq. 0.5384 0.20051 0.19932
S. E. 0.0236 0.01803 0.01439
F-Statistic 8.2159 2.19131 1.98018
*p< 10; **p < .01; ***p < .001
Several of the race variables have high coefficients in this model, but never a
low enough p-value to be considered significant, and the only educational variable to
achieve significance is High School Degree Plus, which does so with a negative
coefficient of 0.0297 in model H, a value which is suspect due to the low adjusted R
squared and F-statistic values estimated for this model of counties. The migration
variables, on the other hand, indicate some predictive power. Gross County
Migration, in particular, shows a positive correlation with CDS in models F and G.

Summary of Test Results
The volume of numbers presented in the above tables is large enough to
require an intermediate distillation, which appears in Table 9. The test of positive and
negative CDS counties is omitted here, as it at best reiterates relationships already
indicated in the other models, and at worst is an unreliable model. Table 9 displays
the coefficient, (or the range of coefficients, in the case that there are more than one
coefficient with p-value less than 0.10, denoted *) along with the p-value significance
level (one to three *s), and the specific model or models within which the coefficient
value or range occurs.
Table 9 Summary of Multiple-regression Model Results
All Tails Center
A B through E F through I
Model Statistics
Adjusted R Sq. 0.5384 0.6075 to 0.8363 0.2005 to-0.1171
S.E. 0.02364 0.0257 to 0.0298 0.0144 to 0.0059
F-statistic 8.2159 7.9655 to 8.0247 2.1913 to 0.8297
Migration Variables:
Gross - 0.3932 (E only) 0.0801 to 0.0925 **(F-G)
Net White -0.2311 to -0.8697 (C-E)
Net NW 1 - 0.0569 *
NW 1 - 0.0233
CCS Black 0.8582 * 1.1166* to 2.2451 *(B-D)
CCS Hisp. 0.4096 * 0.5152* to 1.9540 ** (B-D)
HS + - 0.297 (H only)
College + 0.0460 to 0.4678 (C-E)
Chg TO 0.1296* to 1.0206 (B-E)
Chg CCSJ 0.7019 *** 0.7817** to 1.5371 ** (B-D) 0.4724*** to 0.3521 ** (F-G)
*p < .10; **p < .01; ***p <.001

Table 9 allows the isolation of a particular variable found to have some ability to
predict CDS in one or more of the models, and observe its behavior under the
different circumstances. This observation will be done for each variable in turn in the
following sections.
Migration Variables
One of the particularly notable relationships shown in Table 9 is the gross
migration variable, which changes signs between model E and models F and G.
When considered with Net White migration, for which the coefficient increases
steadily from model C through model E (where it reaches 0.8697*), gross migration
in county model E seems to indicate White flight from generally more populous
metropolitan areas to less populated, more rural ones, as was apparent in Table 2.
Remember that gross migration only shows p-value significance (-0.3932*) in
model E, in the Test of Tails models. Then in the Test of Center models it changes
signs and has significance (0.0801*) as soon as model E is removed from the
population of counties tested (model F, in other words, is simply model A minus
model E). This suggests that the relationship between gross migration and CDS
changes quite substantially between models E and F (G also).
Consider that the change of sign (when compared to model E) and strong
significance of gross migration found in county models F and G, (the first two sub-
groups created by removing counties with the most extreme CDS) means that in
counties where the most inter-county migration is occurring is also where DS in

increasing (or decreasing at a slower rate) and where there is less inter-county
migration than average, DS is declining at higher rates.
Socio-Economic Variables
Lets now consider change in county share of population of Blacks and
Hispanics, as well as creative-class share of jobs, at the same time, as they show very
similar behavior in the different models. One notable aspect of this behavior is the
fact that, while all three of these variables show strong significance with increasing
coefficients in the test of tails as one moves from model A to model D, they all fail to
register p-value significance in model E, the most radical of county samples in terms
of CDS value. This pattern echoes the gross migration variable pattern in these
models, which showed a change of sign when the model E-F border was crossed.
The other two socio-economic variables, college degree plus and change in
turnout, behave in a different manner. They begin to show p-value significance in
model B and have increasing coefficients as the county sample size decreases in the
Test of Tails. Unlike the race and creative-class variables, these two variables are
significant; and even reach their peak coefficient value, in model E. Because of this
fact, they seem to fit into the category previously defined: polar variables, or the
variables associated with electoral changes of counties at the polar extremes of the
CDS distribution.

Mainstream and Polar Counties
There truly seems to be a chasm24 between county model E and models A
through D (F and G also). For this reason, rather than discuss mainstream and polar
variables, it seems more accurate to describe the characteristics of mainstream and
polar counties to better describe the behavior identified in this study. One of the
most significant results of all this data manipulation and analysis is the identification
of two distinct groups of counties in North Carolina in which different characteristics
predict the Changes in their Democratic Strength by county, to wit:
Mainstream Counties.
Mainstream counties are those that exclude model E counties (the most radical
in terms of CDS) and in which variables that predict CDS are found significant in
models F or G (both of which exclude model E counties) and/or in any of models A
through D. In spite of the fact that models A through D all include model E counties
as part of the population of counties tested, the significance of most of the
mainstream variables is lost when model E counties are isolated, and other, polar
variables become significant.
Notwithstanding the fact that there is some overlap, such that some polar
variables are found significant in mainstream counties, here the independent variables
24 This term is borrowed from Moore (1991) and, unlike his chasm, this refers to two chasms, one at
each extreme of the CDS distribution.

that explain CDS are different from those that generally predict CDS in polar counties
(model E counties). The overlap noted is attributed in significant degree to the fact
that the cut off between models is arbitrary. One could reset the cut-off between
polar and mainstream counties, but would have to recognize that each independent
variable might have its own unique cutoff, where it becomes insignificant as one
includes more or fewer counties in the models identified as polar or mainstream. The
important point is that the evidence clearly shows, in spite of its imperfections, that
different variables predict CDS as one moves from the counties with less extreme
CDS values to those with more extreme CDS values.
Thus, Democratic Strength is increasing in mainstream counties where:
Black and Hispanic share of the population is increasing,
Creative-class share of jobs is increasing, and
Gross migration is higher.
Conversely, Democratic Strength is decreasing in mainstream counties where:
Black and Hispanic share of the population is decreasing,
Creative-class share of jobs is decreasing, and
Gross migration is lower. 25
25 As noted earlier, increasing may mean decreasing at a slower pace than the mean, and decreasing
may mean increasing as a slower pace than the mean. These statements are simplified for clarity.

Polar Counties
Polar Counties are those that have the most extreme measures of CDS. These
counties number twenty-three and are identified as model E throughout this study.
Polar Counties set themselves apart by the predictive significance of independent
variables studied in the various models described above. The independent variables
found to have significance and high coefficients either in model E only, or in a series
in the Test of Tails, with increasing coefficients that peak in Model E are predictors
of CDS in Polar counties
Democratic Strength is increasing in polar counties where:
The attainment of a college degree or higher is more common,
The increase in voter turnout is higher,
Gross migration is lower, and
Net White migration is lower
Conversely, Democratic Strength is decreasing in polar counties where:
The attainment of a college degree or higher is less common,
The increase in voter turnout is lower,
Gross migration is higher, and
Net White migration is higher

The models, therefore, indicate that there are distinct differences between a
relatively small subset of North Carolina counties (the polar counties in model E) and
the rest of the states counties (the mainstream counties). Not only do the polar
counties set themselves apart from the mainstream counties by the differences in rates
of CDS measured, but they differentiate themselves by the variables that predict
changes in CDS. This fact suggests that, while some polarization and ideological
sorting is occurring in most (or all) North Carolina counties, a subset of counties is
experiencing ideological sorting of a more profound type, or to a greater extent, and
the independent variables that predict these more extreme measures of CDS may be
the key to better understanding what is driving this more radical form of polarization.
The next chapter will attempt to analyze this situation in greater detail.

Before attempting to glue this data all together into a cohesive picture of
recent electoral (and migratory) behavior in North Carolina, additional information -
including the findings of other scholars, regarding the behaviors observed in this
study should be considered.
Mainstream Counties
Gross Migration
As noted above, the change of sign-to indicate a positive correlation of gross
migration with CDS-found in the mainstream models suggests that in counties where
the most inter-county26 migration is occurring is also where DS is increasing (or
decreasing at a slower rate than the mean). Moreover, where there is less inter-county
migration than average, DS is declining at higher rates than the mean.
Note here that inter-county migration refers to migrants that change counties, which
includes both intra-state and inter-state migrants, as opposed to intra-county migration. This is how
the data are tracked by the US Census Bureau; based on the assumption that an inter-county move -
whether /wfer-state or intra-state is more significant than an intra-county move.

This pattern might reflect one of the aspects of the declining segregation of
Blacks and Whites since 1970, referred to by Glaeser (2008) as follows:
The decline has generally taken the form of small numbers of blacks living in
areas that were previously all white and has been steepest in the growing areas
of the Sunbelt and suburbs. One interpretation of that fact is that the desired
level of segregation is lower today than in the past, so we see more integration
in those areas that are newer and less driven by long-standing historical
... the 2000 Census offered a confirmation of whether this trend continued.
Glaeser and Vigdor (2001) found a continuing decline in segregation,
especially in the newer cities ... High socioeconomic status African-
Americans are today much less segregated than low socioeconomic status
African-Americans ... (p. 184)
Fuguitt et al. (2008) similarly find that the recent migration stream of Blacks
into the South from other major U.S. regions is made up of significantly more highly
educated persons than the average Southern non-migrant Black. It seems possible
that models F and G where non-White migration is significantly and positively
related to CDS represent counties that contain both newer and older suburbs. Where
Blacks settle in these newer suburbs, their strong identification with the Democratic
Party may be generating small, but significant increases in DS, while the older
suburbs may be long-standing White enclaves maintaining the status quo, or even
registering declines in DS. Note that this migration pattern of incoming, better-
educated Blacks to newer suburbs in the South may differ substantially from
migration patterns identified in other regions due to the possible persistence of
longstanding historical patterns (Glaeser, 2008) of segregation that may exist in the
South and not in other US regions.

Migrants may also be sorting into these enclaves based on the perceived
cultural differences described by Bishop (2008), which tend to cause identifiers with
the Republican Party to be more likely to settle in certain places and identifiers with
the Democratic Party to settle in others. Due to the strong predictive nature of
changes in creative-class job share, one might also expect that newer suburbs contain
more inter-state immigrants of all races who have relocated to take advantage of
growing supplies of creative-class jobs, while the older enclaves may be more likely
to be experiencing declines in the traditional mainstay industries of furniture and
textiles. (Eamon, 2008, p.21)
In contrast to the positive correlation between gross migration and CDS found
in mainstream counties, the correlation becomes negative in the polar counties (model
E). One subset of model E counties includes many urban or metropolitan centers,
while the other subset of model E counties tends to represent counties that are more
rural, or at least lesser populated. Taken as a whole, model E counties {polar
counties) show a significant negative correlation between gross migration and CDS
measures, indicating that greater gross migration rates predict decreases of CDS in
these counties. White flight from urban to rural counties may be driving this result by
increasing gross migration rates seen in the lesser populated destination counties in
greater proportion than the loss of the same number of migrants affects the gross
migration rates in the more populated counties departed. The significant negative
correlation with White migration in model E further supports this conclusion.

Racial Composition
The state of North Carolina was 21.6 percent Black and 4.7 percent Hispanic
as of the 2000 census. These groups vote very strongly Democratic Blacks alone
voted ninety-five percent and eighty-eight percent Democratic in North Carolina in
the 2008 and 2004 presidential elections, respectively. Therefore, it should not be
surprising that the changing racial composition of various counties is a force in North
Carolina politics. What is surprising is the fact that Non-White migration statistics
did not score higher in the multiple-regression tests as predictors of CDS in the same
counties where the changing racial composition does predict CDS.
While the birth-rate differential between White, Black, and Hispanic women
is a possible explanation of why Black and Hispanic populations of some counties are
growing, but do not correspond with non-White migration, there is another
possibility: that white flight is causing the Hispanic and Black share of county
populations to increase. Establishing why people live in cities, Glaeser (2008) relates
important truths about residents of all places, not just cities:
The economic approach to cities starts with the assumption that locations are
chosen and that those choices are not entirely irrational. Adults are not
randomly sprinkled across space. They select one place over another, and in
particular they choose cities. This statement does not mean to deny that many 27
27 Differences in the birth rates by race are explored in more detail in the next chapter.

people just choose to stay put to keep their ties to friends and family. Still,
even that is a choice and even in Europe, where mobility is much less than in
the US, there are millions of people each year that move from place to place,
choosing new locations, (p.2)
Regarding this choice, Bishop (2008) found that:
Between 1995 and 2000, 79 percent of the people who left Republican
counties settled in counties that would vote Republican in 2004 and they
were most likely to move to counties that would be Republican landslide
counties.28...By contrast, people who left counties that would vote
Democratic in 2004 migrated to both Republican and Democratic counties
without showing much of a preference for either although they were
unlikely to move to counties that would become Republican landslide
counties, (p.44)
This result suggests that sorting in North Carolina may be more driven by
those who identify with the Republican Party moving into existing Republican
counties and changes in Black and Hispanic share of population may be due in
significant measure to the loss of these White migrants. Never the less, the strong
predictive power of these two race variables, and the fact that significant numbers of
Black and Hispanic migrants are moving into the state each year, create the
expectation that Non-White migration must play some role. After controlling for
race, however, the models failed to estimate predictive much significance for any of
the Non-White migration variables.
28 A landslide county is defined as one ... where one party won by twenty percentage points or
more. (p. 9)

Creative-Class Jobs
The highly significant and consistent predictive power of this variable across
nearly all models makes it an important characteristic of the political change
observed. Yet, most Creative-class jobs require a college degree or more, suggesting
potential for co-linearity of this variable with educational attainment.29 30 In addition to
this, the genesis of the concept of the Creative-class job description (see Florida, 2002
and Ray and Anderson, 2000) appears to simply describe the type of jobs that are
more common in Democratic enclaves, or places where DS is growing; this chicken-
and-egg dilemma is argued by many of Floridas (2002) critics. Such locales, in
North Carolina, appear generally to be metropolitan areas that have major
universities, high-tech industry, and a younger, well-educated work force. Glaeser
and Berry (2005) show (assuming that educational attainment is a reasonable proxy
for creative-class jobs) that, while Creative-class jobs seem to have a strong
association with metropolitan areas, they are not in all metropolitan areas. Bishop
(2008) summarizes their findings:
By 2000 ... there were sixty-two metropolitan areas where less than 17
percent of adults had college degrees and thirty-two where more than 34
percent had finished college. The differences were even more dramatic
among the young. More than 45 percent of twenty-five to thirty year olds in
29 Unfortunately, Excel doesnt test for co-linearity problems.
30 Creative Class also includes jobs that are more common in beach or mountain resort, non-
metropolitan communities, and this is likely another reason why population density was not able to
predict CDS in the models. Apparently creative class jobs are also not all in metropolitan areas either.

Raleigh-Durham, North Carolina, had a college degree in 2000; that figure
was only 16 percent in Las Vegas, (p. 130)
It is easy for most to see why Aspen, Colorado might grow faster than, say
Cherokee, Iowa, but why would people leave Culpepper, Virginia for Raleigh-
Durham? Florida (2002) found that what Glaeser (2008) refers to as agglomeration-
the advantages that close physical proximity, afforded by cities, has on the spread of
ideas and innovation and the efficiency of operations-works best when the highly
educated create a critical mass that includes institutions of higher education and high-
tech, arts, and other information-based industry. Bishop (2008) indicates that such
cities grow faster than their counterparts, and create wealth at twice the pace. Las
Vegas, Nevada is an obvious exception, as it has grown substantially over the past
few decades, yet is not creative according to Florida (2002). Las Vegas has grown
to supply the growing demand for recreation by an increasingly affluent (on average)
America, but many of the jobs created in this process are in the hospitality,
entertainment, and gaming industries, and are not identified as creative. Just south
of the North Carolina state line, Myrtle Beach, South Carolina the Vegas of the
southeast probably looks similar in terms of creative-class jobs. The Research
Triangle in North Carolina, however, has evolved by leveraging top-notch
educational and research institutions and high tech industry. This area accounts for
significant numbers of counties that register exceptional growth in creative-class
share of jobs, and as the data analyzed here indicates many of these same counties

are those that have experienced the highest increases in CDS during the period
studied, along with socio-economic measures.

Polar Counties
Voters who live in Democratic landslide counties have been achieving higher levels
of educational attainment in the past few decades than those who live in Republican
landslide counties. For example, Bishop (2008) found that:
In 1970, the county groups were well balanced in the proportion of the
population that had a college degree. After that, the percentage of college-
educated people increased in every group but the well-educated were
especially attracted to Democratic counties... People with college degrees
increased the most in the Democratic landslide counties, where 29 percent of
the adult population had at least a college degree in 2000. In the Republican
landslide counties, 20 percent of those over twenty-five years of age had a
bachelors degree or higher in 2000. (p.50)
Bishops (2008) landslide counties are more or less the equivalent of models
D and E, in which predictive power for College Degree attainment was found, and
therefore, the results here appear to be in rough agreement with his observation.
Thus, in North Carolina, educational attainment is particularly predictive in only
thirty-six, or just over a third of counties.31 32 Remember that these are the third that
show the most extreme CDS. From the earlier discussion of the data regarding the
population of counties included in model E, it is obvious that there is a metropolitan -
non-metropolitan cleavage involved in polar North Carolina counties, and this fits
31 The difference is that DS can be declining at a pace that makes the county polar in this study, yet if
the county begins at a high value of DS, it may take declining DS sustained over long period of time
before the county becomes a Republican county, let alone a landslide Republican county as defined by
Bishop (2008).
32 Note that Bishops landslide counties make up approximately half of the population tested.

rather well with the above discussion of creative-class jobs in this state as well. It
would be interesting, however, to understand why these two variables (College
Degree and Creative-class Share of Jobs) behave so differently in models B,C,D, and
E, one variable being mostly a polar variable (college degree attainment) and the
other (change in share of creative-class jobs) being a mainstream variable showing
effects on almost every sub-group of counties tested in these models.
Voter Turnout
Non-White voters have historically turned out to vote at lower rates than
White voters (see Gimpel and Schuknecht, 2003, p.375). In the 1998 midterm
election, North Carolina bucked a national trend toward declining turnout, when
compared to the 1994 midterm election (Walters, 2001, p.219). While many states
set records for low voter turnout, North Carolina saw a significant increase. Gaither
and Newburger (2000) note that, in contrast to declining turnout in the nation as a
whole, Black turnout at the polls rose by some 3.1 percent.
Walters (2001) similarly notes that:
Since the 1998 elections, it has been clear that most political observers have
missed an exceedingly important fact: the black vote has been a critical factor
in limiting and to some extent reversing the much vaunted Republican
realignment in the South. Some analysts of the 98 election suggested that
Democrats experienced a resurgence in the once solid-South fielding large
majorities in reliably Republican parts of the region. And although some of

this performance was attributed to the anemic political performance of the
Christian Right, it was also clear that black voters played a substantial role.
He predicts that ...the combination of higher fertility rates among blacks in the
South compared to whites and the continuing migration of blacks back to the South
means that the influence of blacks is certain to grow... (p.227)
It seems likely that the increase in voter turnout, which began in the 1990s in
North Carolina, extended into the 2000s. The continuing influx of Black voters to the
state, who, as Fuguitt (2008) shows, are more highly educated, on average, than the
Black population they are joining (and the more educated are more likely to vote)
further supports the contention that increasing Black turnout continues to play an
important role in the long-term increases in state-wide turnout measured here. The
strong identification Black voters have with the Democratic Party may be the primary
reason that change in voter turnout predicts CDS, and the fact that this is more
pronounced in the polar counties may reflect some contextual effects on voter turnout
in these more populous counties. The closer concentration of residents in urban and
metropolitan areas as compared to rural and exurban areas may cause voters to
have more contact with their neighbors. Gimpel (2003, p.22), however, notes that;
... arguments have not been developed to explain why living in areas of
differing size or density would make voters oppose one another. Those who
have moved beyond compositional explanations to explain urban-rural
cleavages in British political behavior have suggested that the gap rests upon
the institutional arrangements that organize social life, including churches,
workplaces, trade unions, and political parties. Working-class voters in rural

Britain, for instance, are less likely to be trade union members, which helps to
account for their greater support for Conservative party candidates (Johnston
Social interaction, whether institutionalized or encouraged by the informal gathering
at the third places described by Oldenburg (1989), such as bars and coffee houses,
certainly has contextual effects on attitudes toward political participation and
ideologies Bishop (2008) provides a fairly comprehensive summary of group
behavior, more specifically, group polarization. Yet until convincing evidence that
urbanization and/or higher density populations create more social interaction, the
most that can be said here is that the urban or metropolitan setting generally makes
canvassing work easier for politicians and get-out-the-vote activists than does the
greater distance between households in rural areas (not to mention barriers to access
in gated exurbs).
White Flight
Using the concept of spatial equilibrium The key theoretical element in
urban economics... (Glaeser, 2000,pi8)- Cutler, Glaeser, and Vigdor (1999) show
that Black and White segregation changed between 1940 and 1990, from centralized
to decentralized racism... (Glaeser, 2008, p. 185) He explains:
... segregation in 1940 primarily reflected barriers to black mobility, since
African-Americans appear to be paying more for housing in more segregated

cities. By 1990, segregation appears to primarily reflect the white taste for
living with other whites, since whites pay more in more segregated areas...
While these results are more suggestive than definitive, they do match the
institutional history of American segregation, where barriers to black mobility
that were once ubiquitous became illegal. The declining number of lily-white
census tracts and their replacement with slightly more33 segregated census
tracts that include small numbers of generally more educated African-
Americans are also compatible with the view that the barriers to black housing
choices have declined.
Thus, no longer able to pen Blacks in geographically (centralized
segregation), some Whites have been moving to whiter locales and paying higher
prices for the exclusivity (decentralized segregation). This pattern appears to
continue in North Carolina, causing several counties to exhibit the highest rates of
decrease in DS in the state. This finding is also in agreement with Bishops (2008)
findings that the polarizing trend he discovered ...was stronger in Republican than in
Democratic counties. (p.43) He further notes;
... that when people left counties that would vote Republican in 2004, they
were two and a half times more likely to move to other counties that would
vote Republican than to those that would vote Democratic. By contrast,
people who left counties that would vote Democratic in 2004 migrated to both
Republican and Democratic counties without showing much of a preference
for either although they were unlikely to move to counties that would
become Republican landslide counties, (p.44)
In other words, as a source of polarization in the form of geographic partisan
cleavage, White flight and decentralized segregation appear to be stronger influences
33 This appears to be a typographical error; the author must have meant to say less segregated, which
better fits the context of this statement, and is in agreement with the point he makes elsewhere in this

than any opposing force, such as purposeful racial desegregation, or conscious
partisan segregation by those sorting into counties with increasing DS. For this
reason, it may be more appropriate to view the chasm identified earlier as existing
only at the conservative tail of the distribution, the apparent destination of White
flight. Due to the fact that this variable was significant in models C and D in addition
to E, the location of the chasm would place all of model D and part of C on the polar
side of the chasm. Conservatively, however, assuming that all of the counties in
model C were considered polar, one could never the less confine most White flight to
just over one quarter of North Carolina counties (27) representing less than one eighth
of the states population (12.44%) in terms of the county of destination. The former
estimate is in rough agreement with Bishop (2008), who divided counties into four
quadrants based on the occurrence of landslide elections. Without ready access to
Bishops (2008) data, one cannot easily compare the share of population represented
by these counties to his results with nation-wide data. Yet, it seems reasonable to
believe that the counties in Bishops (2008) Republican landslide quadrant of
counties have smaller populations than those in the Democratic landslide quadrant,
based on the overall urban versus rural nature of Democrat versus Republican
enclaves, respectively, that have persisted nationwide for some time now. It also
seems likely that large decreases in DS occur in Republican-landslide counties.
The inability to apply the results of this study to the nation and other states
renders this result more subjective than would otherwise be the case. While
undoubtedly some may feel that decentralized segregation in an eighth of the North

Carolina population is far too extensive and is representative of dangerous polarity,
many others-especially those bombarded with accounts of severe and growing
polarity-might find relief that such behavior appears to be limited to this subset of the

In order to see how the variables found significant above might apply to
specific North Carolina counties, six were selected for further analysis. Two each
were chosen from the mainstream (model F), polar blue (model E portion with the
greatest increase in CDS), and polar red (model E portion with greatest decline in
CDS) county samples, as defined by the values calculated for CDS herein. In
addition, in the attempt to attain some diversity, no two counties from the same
sample are from the same of the three general regions of North Carolina, which
include the coastal, piedmont, and mountain regions. These distinct regions are
defined by their elevations above sea level, but exhibit other defining characteristics
as well.
The piedmont lies approximately two hundred to one thousand feet above sea
level and contains over two thirds of the states population, including the three CSAs
that form the Research Triangle. These major metropolitan areas have largely
replaced the former agricultural uses of land with commercial, industrial, residential,
and retail uses. The mountain (above one thousand feet elevation) and coastal (below
200 feet elevation) regions have some similarities in that they are less densely
populated and have more remote areas than the piedmont. They also have significant

resort industries. In contrast to each other, the coastal region is much more involved
in agriculture and has substantially higher Black populations than the mountain
Table 9- Selected County Data
County Region CDS Pop. HH Inc. Black Migration CC Jobs
Cherokee Mountain -7.30% 24,298 $27,992 1.59% 2.60% 14.23%
Columbus Coastal -6.68% 54,749 $29,805 30.93% -0.62% 12.22%
Iredell Piedmont -2.58% 122,660 $41,920 13.67% 12.60% 19.88%
Craven Coastal -0.89% 91,436 $35,966 25.12% 1.79% 19.13%
Edgecombe Coastal 3.70% 55,606 $30,983 57.46% -2.28% 11.73%
Guilford Piedmont 3.69% 421,048 $42,618 29.27% -0.12% 27.33%
State Avg. -1.99% 80,493 $34,874 21.53% 2.31% 17.25%
All values are as of the 2000 Census, except for CDS (described above) and Migration, which is Net White Migration as
described above.
The counties selected for this analysis are shown in Table 10 along with a
limited amount of information regarding each county. Unfortunately, not all of the
statistics used in the analysis that follows is able to fit in the table.
The selected counties will be analyzed for their fit to the appropriate model or
models based on various statistics, some of which are the independent variable values
for that county; others are statistics that are related to the variables. For example,
whereas the variable used in the multiple-regression models was change in share
creative-class jobs, the share of creative-class jobs in the county (as of 2000) is also
used to provide additional insight. The polar county model used in this chapter will
consider any variables found significant in models D and /or E. These variables

include gross migration (negatively correlated to CDS in this model), net white
migration, change in Black share of population, attainment of a college degree or
more, change in voter turnout, and change in creative-class share of jobs. The
mainstream county model used in this chapter will consider any variables found
significant in models F and/or G. these variables include gross migration (positively
correlated with CDS in this model) and change in creative-class share of jobs. In
addition to these variables (because there are so few) the mainstream counties will
consider general model variables (found significant in model A, or all one hundred
counties), which include change in voter turnout and change in Black share of
As the multiple-regression models do not necessarily indicate that
independent variables found to be significant predictors of CDS in a particular model
(or sample of counties) are equally significant in each county included in the sample,
the expectation is not that these selected counties will all fit the models. Rather, the
following discussion will attempt to explain any variance from the models based on
the particular characteristics or circumstances of that county.

Polar Red Counties
Cherokee County
Cherokee County lies at the western tip of North Carolina in the Appalachian
Mountains. Its primary town, Murphy, is 1,604 feet above sea level. The two-party
vote in Cherokee County split 70 percent Republican and 30 percent Democrat in the
2008 presidential election compared to 55 percent Republican and 45 percent
Democrat in the 1996 presidential election. Thus the extreme CDS of negative 7.3
percent experienced by this county has changed it into a landslide county during
this twelve year period.
Cherokee County is only 1.50% Black and has experienced estimated net
White migration of 2.6 percent and net Black migration of negative 27.18 percent
between 1995 and 2000. This represents a gain of only 564 White migrants and a
loss of only 106 Black migrants, and though it seems to fit the polar model which
tells us that white flight and change of Black share of the population (negative 0.197
percent) are significant predictors of CDS (negative 7.30 percent in this case), when
compared to state averages of 4.41 percent for net White migration, Black net
migration of positive 3.33 percent and change in Black share of population of
negative 0.79 percent, it seems that Black flight is the more pertinent migration
predictor of CDS in this county.

In other ways Cherokee County fits the polar model better. Its increase in
turnout of 8.83 percent, when compared with the state average increase of 9.98
percent, is in line with the model. Increase of creative-class job share of 0.63 percent
is well below the state average increase of 3.28 percent, also in agreement with the
polar county model, as is the attainment of a college degree, which stood at 12.98
percent of the population aged twenty five years or more as compared to the state
county average of 15.14 percent. Gross migration, however, at 36.67 percent was less
than the mean of 37.72 percent, which would not be predicted by the polar model.
Also of note are the facts that density of 49 people per square mile is
substantially lower than the mean of 158, average household income of $27,992 is
much lower than the mean of $34,874, and creative-class share of jobs is 14.23
percent versus the state mean of 17.25 percent. The remoteness of this hamlet, along
with its dearth of population and limited access to higher education (only the two-
year Tri-County Community College resides in Cherokee County) seem to account
for its underperformance in socio-economic terms.
Columbus County
This coastal county resides in the southeastern part of the state. Its county seat,
Whiteville, is 98 feet above sea level, and the countys population of 54,749 is below

the state average of 80,493. In the 2008 presidential election Columbus County split
its two-party vote 54 percent Republican and 46 percent Democratic compared to the
1996 presidential election, in which it voted 40 percent Republican and 60 percent
Democrat. The extremely negative CDS of 6.68 percent has therefore changed
Columbus County from a Democratic county to a Republican county over this twelve
year period.
Like Cherokee County, its density of 57 per square mile, average household
income of $26,805, and creative-class share of jobs of 12.22 percent are well below
state averages. Somewhat isolated and comparatively sparsely populated, Columbus
county is the home of only Southeastern Community College, a two year institution.
Thus, in spite of the fact that the county ranked 7th among US counties in tobacco
production, it appears more socio-economically challenged than Cherokee County.
Also resembling Cherokee County, Columbus Countys migration statistics do
not fit with the polar model. Net Black migration of negative 1.09 percent is in
agreement with the model, yet increase in Black share of population at the rate of
0.32 percent (compared to state average of a decrease of 0.79 percent) does not
amount to the kind of numbers that would predict a decrease in CDS of 6.68 percent.
And net White migration of negative 0.63 percent does not indicate that this county is
the terminus of white flight in any substantial numbers. In agreement with the polar
model are change in creative-class share of jobs (1.84 percent) and the attainment of a

college degree (11.03 percent of the population aged twenty five years and more).
And while change in voter turnout of 10.33 percent compared to the state average of
9.98 percent fails to support the models expectation, it seems to support Walters
(2001) observation of increasing Black turnout at the polls based on the fact that the
county is 30.93 percent Black as compared to the state average of 21.53 percent.
At the end of the day, one is hard pressed to account for the change in this
countys voting record based on the polar county model and other data considered
here. Is Columbus County one of the last bastions of the South to realign? Or have
Black tobacco farmers voted increasingly Republican during the past two decades?
Polar Blue Counties
Guilford County
Located in the piedmont, Guilford County has a 2000 population of 421,048
and is the home of Greensboro and High Point (elevation 900 feet above sea level).
Greensboro alone has an estimated 2009 population of 260,083, is in a metropolitan
statistical area (MSA) estimated at 714,765 people in 2009 (the Greensboro High
Point MSA), and a combined statistical area (CSA) of 1,581,122 people in 2009 (the

Greensboro Winston Salem High Point CSA). Guilford County is located in one
of the three major metropolitan areas (CSAs) that make up the Research Triangle,
which was established in the 1950s to leverage world-class institutions of higher
education and research by inducing for-profit entities and researchers to corroborate
with each other in order to commercialize new technologies in the computer and life
The two-party vote in Guilford County split 41 percent Republican and 59
percent Democrat in the 2008 presidential election compared to 49 percent
Republican and 51 percent Democrat in the 1996 presidential election. Thus the
extreme CDS of positive 3.69 percent experienced by this county has nearly changed
it from a dead heat into a landslide county during this twelve year period.
Household income of $42,618, attainment of a college degree or more of
23.73 percent, creative-class share of jobs of 27.33 percent and its increase of 4.77
percent, and change in voter turnout of 12.38 percent are well above state averages
and in line with the polar model, as strong predictors of CDS. Also in agreement with
the polar model are net White migration of negative 0.12 percent, net Black migration
of positive 6.00 percent, and increase in Black share of the population of 2.89 percent.
All told, this county is the poster girl for the polar blue county model described in
Chapter 4.

Edgecombe County
In the northeast part of the state, Edgecombe County is in the coastal low
country at roughly 100 feet above sea level. In the 2008 presidential election
Edgecombe County split is two-party vote 33 percent Republican and 67 percent
Democratic compared to the 1996 presidential election, in which it voted 36 percent
Republican and 64 percent Democrat. The extremely positive CDS of 3.70 percent
has taken an already landslide county and increased the Democratic margin of victory
during this twelve year period.
Edgecombe County is 57.46 percent Black and registers a change in voter
turnout of 14.87 percent; well above the state average of 9.98 percent. The only other
variable in line with the polar model that would predict the strong positive CDS is the
increase in Black share of the population, which is an increase of 1.48 percent
compared to a state mean of loss of 0.79 percent. The utility of the polar model stops
there. Creative-class share of jobs of 11.73 percent and its change of 1.57 percent and
attainment of a college degree or more of 8.57 percent run somewhat strongly counter
to the modeled expectations. And although net White migration of negative 2.28
percent may be in agreement, net Black migration of negative 3.36 percent does not,
especially in light of the fact that the Black population is bigger than the White
population in this county.

Like Columbus County, Edgecombe County is indeed a curiosity. With much
better access to higher education34 than Cherokee and Columbus Counties, the
attainment of a college degree in Edgecombe County is only 8.57 percent as
compared to the 12.98 percent and 11.03 percent noted in Cherokee and Columbus
Counties, respectively. Yet the average household income in Edgecombe County of
$30,983 is higher than both of those counties. The explanation for this may be the
fact that a small part of the city of Rocky Mount is located in Edgecombe County.
With a population of 57,010, Rocky Mount is also part of an MSA with population of
143,026 as of the 2000 census, and the Rocky Mount Wilson CSA, estimated at
over 200,000 residents. With fewer college degrees and creative-class jobs, the
residents of Edgecombe County utilize the more efficient access to jobs provided by
this more populous, adjacent city. Glaesers (2008) agglomeration (existing in these
metropolitan areas) may provide opportunities for these lesser educated workers to
earn a higher average household income than their more isolated, rural brethren.
Another possible explanation is that Edgecombe County has experienced the socio-
economic exchange of Blacks identified by Fuguitt (2001) such that, as Black
students are graduated by the local Colleges and Universities they move to more
urban or metropolitan environs and are, to some extent replaced by lesser educated
Blacks moving out of those areas (some driven by gentrification, and others possibly
34 This includes North Carolina Wesleyan University, Shaw University, and access to East Carolina
University and North Carolina State University through the Gateway Technological Center.

attracted back to family and/or friends in the Black majority population of
Edgecombe County).
Interestingly, adjacent Nash County, which hosts the more substantial part of
Rocky Mount and its MSA and CSA (described immediately above) has statistics that
are more in line with the polar county model, including 20.22 percent creative-class
job share growing at 4.29 percent, 14.06 percent change in voter turnout, change in
Black share of the population of 2.45 percent, attainment of a college degree at 15.51
percent, and population of 87,420. Its average household income is $37,147.
Mainstream Counties
Craven County
New Bern, the county seat, and the one-time capitol of the North Carolina
colonial government, then for a short time the state capitol, is the second oldest town
in North Carolina. At just thirty feet above sea level, New Bern is located at the
confluence of the Trent and Neuse Rivers near the Atlantic Ocean. Craven County
had some 90,436 residents as of the 2000 census, and has CDS of negative 0.89
percent, which is less than the state mean of negative 1.99 percent. The two-party
vote in Craven County split 57 percent Republican and 43 percent Democrat in the

2008 presidential election compared to 56 percent Republican and 44 percent
Democrat in the 1996 presidential election. Thus the small negative CDS
experienced by this county has given the Republican Party a slight advantage during
this twelve year period.
Gross migration of a whopping 60.60 percent as compared to the state average
of 37.72 percent fits the mainstream model, predicting less than average decrease of
Democratic strength. Change in creative-class job share of 2.01 percent as compared
to state average of 3.28 percent, however, does not agree with the mainstream model.
Also of interest are the facts that college degree attainment of 16.87 percent, creative-
class job share of 19.13 percent, and average household income are all above state
averages. Considering the high gross migration rate and the fact that it is located near
the coast within 90 miles of Wilmington, it seems plausible that this county has
significant resort industry, including arts, sporting, and other leisure activities. Many
such industries under-employ college graduates in non-creative-class jobs for short
tenured stints, in Colorado we call them ski bums.
Iredell County
Located in the piedmont, Iredell is one of the fastest growing counties in
North Carolina (thirty two percent population growth between 1990 and 2000). It is a

bedroom community, or exurb of both the Charlotte Gastonia Salisbury CSA (part
of which is in South Carolina) and the Greensboro Winston Salem High Point
CSA described in the Guilford County section. Iredell is closer to and included in the
Charlotte Gastonia Salisbury CSA, which had an estimated population of
2,574,787 as of 2008. In the 2008 presidential election, Iredell Countys two-party
vote split 63 percent Republican and 37 percent Democrat compared to the 1996
presidential election in which it split 62 percent Republican and 38 percent Democrat.
Thus, already a Republican landslide county as defined by Bishop (2008) it has
slowly lost Democratic strength at a rate of 2.58 percent. This is faster than the state
average decrease in Democratic strength of 1.99 percent, but not fast enough to
qualify the county as a polar county as defined here.
The change in creative-class jobs of 5.43 percent in Iredell County is higher
than the state average of 3.28 percent, which contradicts the mainstream model.
Creative-class share of jobs in the county stands at 19.88 percent is above the state
mean. While gross migration is not below the state mean of 37.72 percent, as would
be predicted by the model, at 37.83 percent, it is only slightly higher. Republican
voters holding creative-class jobs seem to have an affinity for this locale, and the
substantial growth of the county likely is a factor in causing the gross migration
statistic to lie outside the models expectation.

Change in voter turnout is slightly above the mean and out of line with the
general model; however the reduction of the Black share of the population at the rate
of 2.34 percent is in line with the general model. This is likely related to an
interesting facet of Iredell County; that fact that net White migration stood at 12.60
percent between 1995 and 2000 substantially higher than the state average of 4.41
percent. This strong showing in a polar county characteristic might lead one to
expect the county to show polar-level CDS, yet its CDS of negative 2.58 percent
places it squarely in the mainstream. It seems plausible that many of the White influx
consists of identifiers with the Democratic Party, as one might expect of an exurb in
the Research Triangle. With such high net White flight, even the loss of Black share
of population does not rule out the possibility that there exists some level of Black
influx into newer suburbs in metropolitan areas of the South, as described by Glaeser
(2008). Both of these factors may be responsible for offsetting what might otherwise
be a strong polarizing rate of decrease in Democratic strength.
Its substantial growth in population, creative-class jobs, and college degree
attainment account for its high average household income level. At $41,920, Iredell
Countys average household income is substantially higher than the state average of
$34,874, and the second only to Guilford County in our selection.

As expected, the selected counties did not all fit the appropriate models
perfectly. Yet most of them fit the models to a large extent. The counties that did not
fit seem more interesting. Edgecombe County, for example, with its better access to
education, but extremely low attainment of college degree may be an example of the
Fuguitt et al (2003) observation that higher educated Blacks are moving to
metropolitan areas, while the less educated are leaving metropolitan areas for the
more rural settings. More importantly, this county seems to emphasize the
importance of agglomeration to the socio-economic status of residents. With
significantly lower college degree attainment and availability of creative-class jobs,
the residents of Edgecombe County earn a higher average household income than
their counterparts in Cherokee and Columbus Counties.
Another testament to the power of agglomeration is Iredell County. What at
first appears to be the destination of excessive levels of White flight seems to be an
attractive locale for both White people leaving urban settings, as well as interstate
migrants choosing to live in a suburban setting with access to Research Triangle jobs.
The CDS experienced in Iredell County suggests that many of the influx of White
population identifies with the Democratic Party, though this is surely a minority of
the total influx based on CDS and the changes in voting patterns noted here. Finally,
positive Black net migration to Iredell County seems to support Glaesers (2008)

contention that Black interstate migrants from without the South are settling in small
numbers in such locations.

With approximately ten million inhabitants, North Carolina is now the tenth-
largest state in the union. It has much in common with its Southern neighbors,
including geographic location and a large African-American population. Yet North
Carolina has a long history of distinguishing itself from other Southern states. Key
(1949) observed that It enjoys a reputation for progressive outlook and action in
many phases of life, especially industrial development, education, and race relations
(p.205). While each state in this union has its own unique combination of
characteristics, some of the information provided here for North Carolina may be
descriptive, to some degree, of other states in the South, as well as the nation as a
whole. The following sections will explore some of the more interesting findings
arrived at through the analysis provided in earlier chapters.
The Theory of Racially Unbalanced Migration
What earlier may have looked like partisan change driven primarily by
racially unbalanced migration did not hold up very well under multiple-regression

tests. The only unbalanced racial migration observed here is that described as White
flight in Chapter 4. Thus, while one may conclude that unbalanced racial migration is
causing a degree of sectionalism in North Carolina, it affects only about half of the
states counties.35 As noted previously, Bishop (2008) establishes that such White
flight or decentralized segregation, to use Glaesers (2008) term is likely a
bigger factor in racial segregation than the concentration of Blacks in specific
neighborhoods. Recall that White flight affects both the county of destination (by
increasing the ratio of White residents) and the county of origin (by decreasing the
ratio of White residents and increasing the ratios of non-Whites). Thus, one can see
in this study evidence in support of Glaesers (2008) contention that, since civil rights
legislation and enforcement (which more or less began in the 1960s) systematically
removed the ability of Whites to restrict Black mobility which Glaeser (2008) refers
to as centralized segregation White flight has now become the primary
segregation vehicle. This study shows that White flight36 is likely to have been the
primary segregation vehicle in North Carolina during the past two decades.
35 More precisely, fifty six counties, which includes all counties tested in models C, D, and E where
net White migration was found to be significant.
36 This study found that net White migration out of more populous urban or metropolitan counties and
into less populated counties, some relatively remote (such as Cherokee and Columbus Counties) and
others more suburban (such as Iredell County) predicted high rates of increase of DS in those Urban or
Metropolitan counties and high rates of decrease in DS in the less populated counties that are
experiencing positive net White migration.

Notwithstanding the curious fact that, although White flight is a significant
predictor of CDS in the polar counties (model E), the change in the Black share of
county population is not, and the fifty six counties tested in models C through E seem
to correspond rather closely to Bishops (2008) Democratic and Republican landslide
counties. Recall that these fifty six counties are split, so that nearly half are those
with the greatest increase in CDS (twenty nine) and nearly half are those with the
greatest decrease in CDS (twenty seven). As these county groups are roughly
quadrants of North Carolinas total of one hundred counties, one might expect that
they match up closely with Bishops (2008) landslide counties in this state.
Following this line of reasoning, one might also wonder how great an impact White
flight has in Bishops (2008) nation-wide population of counties, or even just in those
in Southern states.
The Changing Face of the Carpetbagger
Campbell et al (1966) identify Republican, upper middle class migrants to the
South as a significant factor in the changing partisan landscape in the 1950s and
1960s. At that time the influx contained many migrants who identified with the
Republican Party, because at that time this was the prevalent party affiliation in the 37
37 As noted in Chapter 5, there certainly is some divergence between the subsets of counties that are
identified as polar here and Bishops (2008) landslide counties. Yet, based on his contention that
during the period of his study, few if any counties that became landslide counties reversed the trend,
one might expect the divergence to be small. A more thorough comparison of these two sets of
counties might yield insights as to the validity of this one-way-valve observation of Bishop (2008).

states of their origin. They were infiltrating the solid Democratic South, creating an
increasingly competitive two-party landscape, and many were taking advantage of the
opportunity to participate in the industrialization of the South. This trend meant that
many were moving to metropolitan areas and concurrently contributing to the
urbanization of the South. The only real change between that era and the current one
is the partisan affiliation of the South and its domestic in-migrants; todays South is a
Republican stronghold and its in-migrants are more likely to be Democrat, due to the
preponderance of this affiliation in the regions of their origin. Thus, it is now middle
and upper middle class Democrats, moving to the metropolitan areas of the South
where the higher paying jobs that require higher education exist, along with warmer
retirement locales, who are changing the political landscape.
The multiple-regression results obtained in models F and G seem to support
this view of interstate migration. Once the counties containing white flight (in
particular, the counties tested in models D and E) are removed from the population
tested, CDS shows a significant, positive correlation with gross migration. Although
this result may to some extent reflect the mobile nature of the more highly educated,
it seems likely that it also reflects migrants from other regions, coming to North
Carolina for creative-class jobs. And note that the change in creative-class job share
variable also shows significance in models A through D, each of which include the

polar counties in model E, yet the correlation is strong enough to show significance
until the population of counties is reduced to the polar counties alone (model E).
This phenomenon is not very likely confined to North Carolina, though it may
be more pronounced here due to the more advanced state of education and industry
- compared to some other states. One would expect that in-migration from other
regions will continue to provide Democratic identifiers in some measure, who will
continue to temper the strength of the Republican Party in many Southern states.
The purpose of this study was, in part, to assess the extent of polarization of
the people of North Carolina. The trend in sub-state partisanship cleavages was
measured to arrive at a long-term measure of change in Democratic Strength (CDS),
which is a reasonable proxy for county polarization. Many independent variables
were tested in various ways to learn more about the causes of observed CDS. A
summary of the results of these tests with respect to polarization must
acknowledge that there is some sorting by partisan identification occurring state-
wide. Yet the only obviously overt polarizing behavior identified here is White flight

which, as discussed above, is confined to just over one quarter of the counties in the
state,38 representing less than one eighth of North Carolinas population.
While decentralized racial segregation seems overtly racist, it may represent
less of a conscious, ideological event, than cultural flight.39 Rich Benjamin (2009),
a Black author, describes his conversation with a White woman, from an all-White
church in Idaho, at a dinner to which she had invited him, Alice says that many
Californian transplants to Idaho are not racist, but want to stick to their own kind.
The plump, soft-spoken grandmother gently puts her hand on my armrest and asks,
Without disrespect, dont you want to stick to your own kind? (p. 118)
This interpretation, of course, gives the benefit of the doubt to such cultural
flight, which Glaeser (2008) blatantly refers to as decentralized racism40 (p.185)
Yet some of this particular form of migration is surely based on racism and ideology
in North Carolina. The simple estimate of the numbers of counties impacted by
38 This contention makes the distinction between the fifty six counties affected by White flight, and
those counties that are the destination of White flight. Whereas the counties of origin are affected by
White flight, it is the destination counties that contain the citizens who have actively, or overtly, voted
for segregation, culturalflight, and/or polarization with their feet. Arguably, this is different from the
counties of origin, which though affected may be quite passive in the process, especially in light of
Bishops findings regarding the relocation preferences of migrants from Republican versus Democratic
landslide counties cited above.
39 Ron Rankin, a commissioner of Kootenai County, Idaho form 1996 to 2002, quoted by Rich
Benjamin, 2009, page 118.
40 Glaeser also uses the term decentralized segregation, which is generally used in this report to
avoid making the more severe assumption about the behavior.

White flight contained here seems somewhat limited in its ability to assess the total
effect net White migration may have on the counties of North Carolina. On a more
positive note, while one might expect to find substantial White flight in states like
Idaho (with a history of appealing to white supremacist groups, such as Aryan
Nation), and other Southern states that have a deeper tradition of racism (such as
Mississippi), its occurrence is likely more limited in other states outside the South.
Black Voters: A Democratic Force in the South
One of the most interesting results found in this study is the many faceted role
African-Americans have had in the balance of Southern political power in recent
times. The strong and cohesive support of Black voters in North Carolina for the
Democratic Party causes the relationship between Black share of county population
and CDS to stand out in the multiple-regression estimates undertaken here. Between
this and other studies cited here, one gets a sense of the significant and growing
electoral power of North Carolina Blacks. Note that many of the characteristics
discussed here in connection with Blacks hold for Hispanics as well. The Hispanic
share of North Carolinas population remains fairly small, but is the fastest growing
ethnic segment in recent years.

Characteristics of the Black Electorate
Higher fertility rates are experienced by Black women than White,41 yet the
changes in Black share of county populations observed here are not likely due to this
factor alone. Nor are these changes likely caused solely by the exodus of White
migrants from particular counties, though, as shown above, this trend too plays a
significant part. It appears likely that an important migration factor is the in-
migration of Blacks to the South from other US regions. As observed by Fuguitt et al
(2008), Black in-migrants generally have higher educational attainment and economic
status than Blacks in the areas where they settle. Thus, as inter-state Black migration
continues, it both increases the ranks of those who identify with the Democratic Party
and gradually increases the overall educational and, therefore, socio-economic status
of Black populations in the South (which makes them increasingly more likely to turn
out at the polls). The continuation of inter-state migration of Blacks to the South
promises continued growth in the political power of the Black community there,
41 The US Census Bureau reports that, of women age 15 to 44 as of the June 2006 CPS and the 2006
ACS, non-Hispanic White women were estimated to have had 1.09 births, Black women 1.29 births,
and Hispanic women 1.44 births.

especially when coupled with the fact that Black women have higher fertility rates
than non-Hispanic White women, and they have their children at younger ages.
Walters (2001) observes that between the elections of 1994 and 1998, Black
voter turnout bucked a national trend toward lower turnout, especially in the South;
Since the 1998 elections, it has been clear that most political observers have missed
an exceedingly important fact: the black vote has been a critical factor in limiting and
to some extent reversing the much vaunted Republican realignment of the South
(p.219). The consistent significance of voter turnout in the same county populations
that indicated strong significance of changes in Black share of population as
predictors of CDS (see Table 10) suggest that the trend identified by Walters (2001)
has continued into the next decade. Unlike migration and fertility, the turnout factor
is more likely to be susceptible to a diminishing return. Yet, combined with growth
in the Black population, increasing Black turnout at the polls has the potential to
continue to supercharge Black electoral influence for some time to come.

The Black vote has been instrumental in balancing out the electoral shape of
the South and may be an important reason why North Carolina recently (in 2008)
placed its electoral votes in favor of the Democratic candidate for president for the
first time since it backed Jimmy Carter in 1976, prior to the completion of the
realignment42 In spite of the fact that the 2010 mid-term elections produced declines
in DS in many of North Carolinas counties, continued regional migration to the
South promises to continue countering Republican electoral strength there. This
leaves Arizona, Idaho, and some states in the Great Plains as the only bastions of
Republican strength unopposed by significant populations of Black voters, though a
latent Hispanic turn-out could ultimately serve the same purpose in these states.
According to the Cahn and Carbone (2010) blue-family paradigm, identifiers
with the Democratic Party tend to put off marriage and children in order to attain
higher levels of education and higher economic status through the higher-paying jobs
available to college graduates. Then, once they have established the human and
financial resources, blue families are created, which have lower divorce rates and
fewer children, who in turn focus on their own educational attainment. The
42 The realignment refers to the transition of the solid Democratic South into more of a Republican
stronghold, of which Campbell (1965) saw signs in the late 1950s and predicted circa 1965. For a more
comprehensive chronology and analysis see Black and Black (2002).

occupations of these people are more likely to be creative-class than those of their
counterpart, the red families. In contrast to blue families, the red families (who
identify with the Republican Party) tend to get married and have children much
younger, utilizing resources (especially time and money) that might otherwise go to
creating human and financial capital. This choice leaves the red family
breadwinners) increasingly challenged to obtain higher-paying, creative-class jobs
and limits the resources available for the education of their children.43
While the characteristics described by Cahn and Carbone (2010) are not found
uniformly throughout the different samples of North Carolina counties tested here, the
test results suggest that these two paradigms may be increasingly accurate in
describing county inhabitants as one progresses toward the more extreme measures of
CDS. If this conclusion is true, the future may hold a growing wealth disparity
between Republican and Democratic enclaves in North Carolina, such that the better
educated faction of Democratic identifiers increasingly take the better-paying,
creative-class, information economy jobs, possibly even narrowing the gap between
their upper middle class status and that of the highly affluent faction of Republican
voters. In contrast, the less affluent, middle- and lower-middle-class faction of
43 One might argue that, with public education available universally in the United States, parental
assistance is more important to a childs education than financial resources (especially at early ages).
However, with more children per red family (according to Cahn and Carbone, 2010) even parental
time is necessarily diluted for children of red families, not to mention that, with more mouths to feed
and clothes to buy, red family breadwinner(s) are likely to work longer hours on average to meet basic

Republican voters if indeed challenged to attain higher education as described by
Cahn and Carbone (2010) may continue to take the brunt of the disappearing supply
of jobs that pay a living wage or better in older industries that have been decimated
in the past decades by global competition44 In North Carolina these include the
furniture and textile industries. This working class might see their economic status
continue its gradual slide in the direction of the least affluent segment of North
Carolina society, that faction of Democratic voters who are living in poverty and look
to the Democratic Party to provide government assistance.
Having voted Democratic in vast numbers since the New Deal, the working
class has been in flux in recent times; after becoming disillusioned with the Bush
administration its members voted for Obama in large numbers in 2008, but then voted
Republican in the 2010 midterm, ostensibly due to continuing high unemployment
rates. In North Carolina the swing back to Republicans does not appear as
pronounced as in other states; U.S. Senator Burr retained his seat by more than a
twelve percent margin, and the only U.S. House seat gained for the Republican
delegation was in District two, where Republican Renee Elmers unseated Bob
Etheridge by a narrow margin.45
44 This would be especially true of remote counties with limited access to job opportunities and higher
education, such as Cherokee and Columbus Counties, as discussed in Chapter 5.
45 Although, where Republican US House incumbents won, they won by landslides (twenty
percent margin or greater). And where Democratic US House incumbents won, the margins were
uniformly less than a landslide (which seems to echo the one-sidedness of polarization observed in
Republican enclaves when compared to Democratic enclaves observed by Bishop (2008) in terms of

Referring to the working class in the aftermath of the 2010 midterm elections,
David Brooks (2010) notes that, American politics are volatile because nobody has
an answer for these people. They will remain volatile until somebody finds one. In
light of the polarization trends in North Carolina and the prospects for the different
family paradigms described by Cahn and Carbone (2010) which are likely to be
reinforced by the effects of group polarization one might reasonably expect that the
non-Hispanic, White working class in North Carolina will increase in political
importance down the road as they search for answers to their increasing socio-
economic plight.
Finally, the assessment of the extent of sectionalism in North Carolina
presented here, while not alarming, is certainly a cause for concern in light of ... the
phenomenon of group polarization that groups over time become more extreme in
the direction of the average opinion of individual group members. (Bishop, 2008,
p.67) In his chapter entitled The Psychology of the Tribe Bishop (2008) presents a
well-organized summary of social psychology studies from its origins to the present
understanding of group polarization. Thus, not only is some level of sorting likely to
relocation preferences, and further supported by White flight detected in the models created in this

continue, but communities where the more extremely liberal or conservative are
thereby concentrated are likely to become more extreme in their viewpoints.
Based on the results obtained here, Fiorinas (2005) self-fulfilling prophesy
seems to be occurring to some extent in North Carolina. If this is, as Lasch (1995)
and Hacker and Pierson (2005) argue, a case of elites leading the masses to divisive
confrontation, they seem to have tied ...knots no sailor ever knew,46 which appear
extremely difficult to unwind. Fiorina (2005) shows how the American political
system now favors minorities with extreme positions on issues that are not of general
concern to the masses, and seems rather loath to provide practical solutions. The
solution may be held in Laschs (1995) opening comment:
Once it was the revolt of the masses that was held to threaten social order
and the civilizing traditions of Western culture. In our time, however, the
chief threat seems to come from those at the top of the social hierarchy, not
the masses. This remarkable turn of events confounds our expectations about
the course of history and calls long-established assumptions into question.
Finding common ground both geographically and ideologically may ultimately
require a revolt of the masses, and it seems reasonable to expect that the working
class will, sooner or later out of economic necessity, lead the charge in North
Carolina, if not elsewhere. Franks (2004) describes a Kansas electorate that has
essentially been duped by the Republican machine that has for three decades now
46 From The Music Man, by Meredith Wilson.

held social-conservative issues such as prayer in schools, abortion, gay marriage, and
such, on the stick. In Franks (2004) estimation, as Kansas voters continued to reach
for these carrots, the Republican machine went about penalizing the lower socio-
economic class, which constituted an important part of the staunch core of the
Republican constituency. Meanwhile, the Republicans failed to deliver anything
substantial in the way of social conservative reform. At the state-level there has been
some legislation passed to restrict gay marriage and make abortion more difficult to
obtain, which, along with state legislation directed at other social-conservative issues,
are enumerated and analyzed by Cahn and Carbone (2010). And in spite of their
findings of some trends toward more state laws, the fact remains that tax cuts and the
dismantling of the financial and other regulation was done at the national level and
has cost the average tax payer dearly, while benefitting (in many instances created)
the wealthy.
Social-conservative extremists in Kansas might not ever have been the minority that
Fiorina (2005) describes. Consider the song printed inside the cover of Frank (2004):
Oh, Kansas Fools! Poor Kansas Fools!
The banker makes of you a tool.
-Populist song, 1892

It is difficult to believe that polarization in North Carolina places those who identify
with the Republican Party there on the path toward Kansas style fervor for social-
conservative issues, but possible. With so much economic pain to go around lately, it
would not be surprising, however, if middle class Republican began to vote more for
economic reasons than social-conservative ones.
The willingness of many working class voters to cross over and vote Democrat in
2008 may be an early indication of their ability to break out of the chains of
ideological polarization in search of solutions, for which there is great hunger, as
John Hickenlooper observed. In light of the Campbell et al. (1966) observation
regarding the natural return of voters at mid-term, and David Brooks recent
comments regarding the working-class, Mitch McConnells assessment of the 2010
midterm election results, characterized by Jim Manley (a spokesman for Harry Reid,
the Democratic Senator representing Nevada) as ...its our (Republican congress)
way or the highway. (Steinhauer, 2010) may exaggerate the message voters were
sending the Republican Party. This is especially true in light of the fact that the first
great cause toward which the Republican machine has focused its no-compromise
position is extending tax-cuts for the wealthy. As of press time they had refused the
Democrats offer to raise the cut-off to $1 million in taxable income. Confident that
they would not need to place a dollar-value cap on the extension of tax cuts, they

offered a bone to the Democrats: extended benefits for the unemployed of the
Democrats pay for it.
Hacker and Pierson (2005) show how painfully few Americans desired the Bush tax
cuts in the first place, when asked how they prioritized tax cuts. (Herszenhom and
Calmes, 2010) As Allen Toussaint wrote,
How Long Can This Go On?47
With any luck the Obama administration can deliver, by the end of its term, more of
the change voters sought and that Bishops (2008) one-way-valve can be reversed.
47 Written by Allen Toussaint, recorded by Lee Dorsey in 1965-66, but much more popular in recent
times is the Devo 1981 cover on their album New Traditionalists.

Benjamin, Rich. (2009). Searching for Whitopia: An Improbable Journey to the Heart
of White America. New York, NY: Hyperion.
Bishop, Bill with Cushing, Robert G. (2008). The Big Sort; Why the Clustering of
Like-Minded America Is Tearing Us Apart. Boston New York, Houghton
Mifflin Company.
Black, Earl & Black, Merle (2002). The Rise of Southern Republicans. Cambridge,
Massachusetts and London, England, The Belknap Press of Harvard
University Press.
Bositis, David A. (2008). Joint Center for Political and Economic Studies. Blacks
and the 2008 Elections: A Preliminary Analysis. Retrieved October 12, 2010
Brooks, David (2010, November 4). Midwest at Dusk. Retrieved November 11,
2010 from The New York Times website:
http://www.nytimes.eom/2010/l 1 /05/opinion/05brooks.html?scp=l &sq=mid
Brown, Thad A. (1988). Migration and Politics: The Impact of Population Mobility
on American Voting Behavior. Chapel Hill and London: The University of
North Carolina Press.
Cahn, Naomi & Carbone, June (2010). Red Families v. Blue Families: Legal
Polarization and the Creation of Culture. Oxford University Press.
Campbell, Angus, Converse, Philip E., Miller, Warren E., & Stokes, Donald E.
(1966). Elections and the Political Order. New York, NY: John Wiley and
Sons, Inc.
Eamon, Thomas F. (2008). The Seeds of Modem North Carolina Politics. In C.A.
Cooper and H.G. Knotts (Eds.) The New Politics of North Carolina, (pp. 15-
37). Chapel Hill, NC: The University of North Carolina Press.