. 2
( 7)


determine whether a black-led coalition or a white-led coalition will con-
trol the local political arena. With limited learning from the ¬rst term
of a black mayor and with blacks that much closer to actually taking
over, whites who originally feared black leadership should be even more

16 Black city council representation is closely correlated with the percentage of blacks in a
city. In the racially balanced cities examined in this book, blacks held, on average, a third
of the council seats at the time of the transition to a black mayoralty. This means that with
some support from white liberals (or in some cases Latinos and Asian Americans), the
black-led coalition probably controlled close to half of the city™s major political of¬ces.
Changing White Attitudes

pronounced in their opposition and white voters should continue to vote
against the black incumbent in large numbers. In these racially balanced
cities, white opposition should not decline and elections should continue
to be racially charged and competitive affairs. In sum, then, the infor-
mation model forecasts that the most pronounced and positive change
in white attitudes and votes should occur in minority white cities, with
more moderate positive change in majority white cities and little positive
change “ or even negative change “ in racially balanced cities.

existing research
Unfortunately, at this point we know very little about white reactions to
minority representation across the United States. The evidence to date is
sparse, anecdotal, and often inconsistent. One can ¬nd cases to support
each of the three different theoretical claims described here.
The earliest reports on black mayors and perhaps the bulk of reports on
all black leaders point to the positive effects of black leadership on white
residents and ¬t closely with an information model of white behavior.
In one of the ¬rst assessments of white reactions to black incumbents,
Peter Eisinger found, for example, that white elites in Atlanta and Detroit
“responded initially to the prospect of transition with fear, but living
under black government brought gradual and widespread acceptance”
(1980: 75). Another study considered white voting in two mayoral elec-
tions each in eight cities and found that white support for black candidates
jumped markedly after blacks became incumbents (Watson 1984). In per-
haps the starkest case, in Los Angeles, Tom Bradley, began his tenure with
62 percent of whites opposing his candidacy amidst widespread concerns
that the black mayor would hurt the white community. He ended his
career twenty years later, after winning reelection ¬ve times “ each time
with the majority of LA™s white voters offering him their support (Sonen-
shein 1993). A number of other accounts of a single black incumbent or
of voting patterns in a series of local elections in one city have also con-
cluded that incumbency increases white crossover voting for black candi-
dates (Vanderleeuw 1991; Stein and Kohfeld 1991; Bullock and Campbell
1984; Pettigrew 1976; J. Franklin 1989; Persons 1993). Surveying these
¬ndings, Colburn recently claimed that “the more often blacks served in
prominent political positions and as mayors, the more acceptable they
were to whites” (Colburn and Adler 2003: 40). This positive turnaround
in the white vote across these cases seems to offer support for an infor-
mation model of white political behavior.
Black Leadership: The Possibilities 31

There are, however, important exceptions to this pattern of growing
white support for black incumbents. Reports from the late 1980s and early
1990s tended to be much less sanguine about white acceptance of black
leadership. Accounts of politics in several cities seemed to af¬rm a white
backlash model more than an information model (Rivlin 1992; Abney
and Hutcheson 1981; Grimshaw 1992; Pinderhughes 1994). In Chicago,
for example, Harold Washington™s election as the ¬rst black mayor of the
city seemed to lead to more rather than less racial con¬‚ict (Rivlin 1992).
The Wall Street Journal noted that Chicago had been transformed during
Washington™s tenure from “the city that works” to “Beirut on the lake.”
Public opinion data in Atlanta similarly suggest that the election of a
black mayor led to increased rather than decreased distrust among whites
(Abney and Hutcheson 1981). When, amid severe racial tensions, black
incumbents lost a series of reelection bids in the nation™s largest cities,
many thought “the end of the rainbow” had occurred (Sleeper 1993).
Observers saw these incumbent defeats as a sign that whites ultimately
would not accept black leadership (Rivlin 1992; Holli and Green 1989;
Browning, Marshall, and Tabb 1997).
Still other studies, particularly those that focus on congressional elec-
tions, appear to demonstrate little or no reaction to black incumbents
(Bullock and Dunn 1999; Gilliam 1996; Bobo and Gilliam 1990; Gay
1999; Parent and Shrum 1986, Voss and Lublin 2001). Studies that have
followed a number of black congressional candidates as they move from
competing as challengers to running as incumbents have found that their
white support has remained relatively stable (Bullock and Dunn 1999;
Gay 1999). The same type of nonreaction has been repeated in several
cities, for example, New Orleans, where the presence of a black mayor
appears to have had little impact on either race relations or white vot-
ing patterns (Gilliam 1996; Bobo and Gilliam 1990; Parent and Shrum
1986). This white nonresponse offers at least some support for a racial
prejudice view of white Americans. Thus, while it looks like white Amer-
icans are usually responding positively to black leadership, the evidence
is somewhat inconsistent.
The main problem is that most of this evidence is anecdotal. Studies
have tended to focus on a particular individual or city and to general-
ize from these single cases. We have few quantitative estimates of black
incumbent success rates across a range of cases. As that is the case, it
is dif¬cult to make claims about the “average” white reaction to expe-
rience with black leadership. Another real problem is that most existing
studies have failed to distinguish between black challengers and black
Changing White Attitudes

incumbents. Since all black candidates are simply lumped together, there
is no way to tell whether black incumbents do better than similar black
challengers. As a result, we really do not know whether race declines in
importance after blacks are elected to of¬ce.

improving on existing research
There are two main goals to the empirical analysis that follows. The ¬rst
is simply to offer a broader, more representative account of how white
Americans respond to black representation. Given the anecdotal nature
of previous studies, it is imperative that we get an overall picture of how
whites respond to their experiences with black leadership. The second
empirical goal is to test the information model. Is information the key
to understanding the impact of black representation, or is some other
mechanism such as social dominance or racial prejudice at play when
African Americans enter of¬ce?
Fortunately, these two empirical goals can be achieved by focusing
on the same set of empirical tests. First, to provide a broad picture of
how black leadership affects white political behavior, I need to assess
change not on one measure as past studies have tended to do but instead
on an array of key indicators. Therefore, I focus on change across two
sets of important measures: white voting patterns and white racial atti-
tudes. These two measures assess critical but different features of the
white response to black leadership. The vote, because it determines elec-
tion outcomes and the distribution of a wide array of public resources, is
the most direct and telling measure of white willingness to support black
leadership. Racial attitudes, because they provide us with a glimpse into
what white Americans are thinking about blacks and black leadership, are
the most direct and telling measure of the underlying motivations of the
white community. Change on one measure alone could be viewed by some
observers as ambiguous, but change on both measures is clear. If black
representation leads to a positive change both in white voting behavior
and in white racial attitudes, then we know that black leadership has led
to a positive and meaningful transformation of the white community.
To address the second goal of testing the information model, I need
to test a prediction that it uniquely and necessarily makes. Fortunately,
the predictions made by the information model contrast with the predic-
tions made by any other mainstream theories. If the information model
is true and white residents do learn from black representation, then two
changes in white behavior must logically follow. Whites will become more
Black Leadership: The Possibilities 33

willing to vote for black leadership and white attitudes toward blacks
will improve. No other theory predicts this pattern of white reactions. If
prejudice is the main force driving white behavior, then one would pre-
dict little or no change on either the white vote or white attitudes. A
social dominance perspective predicts that whites should react negatively
to any challenge to the racial status quo. And if we were to make pre-
dictions based on previous instances of black leadership (e.g. in the late
nineteenth century), we would expect whites to mobilize to try to oust
black leadership.17 Thus, to see if the information model explains white
reactions to black leadership, I can look at the same two measures: white
voting patterns and white racial attitudes. If black representation leads to
positive change on both measures, then we have compelling evidence in
favor of the information model.
To accomplish both of these tasks and to systematically assess changes
in white political behavior under black representation, I examine three
different data sets related to black mayoral leadership. In each case, I do
not examine a handful of black representatives in one or two cities “ as
past studies have generally done “ but instead I examine an entire universe
of cases of white reactions to black representation. The primary tool to
assess change in the white vote is a series of ¬fty-two black challenger and
black incumbent elections. Since I want to look at how experience with
a black incumbent changes the white vote, I compare and contrast the
white vote in two elections in each city “ the election in which a city elects
its ¬rst black mayor and the election immediately following, in which the
same black candidate runs for reelection. Also, since I want cases where
white voters were forced to choose between black and white leadership, I
exclude cities where the black candidate ran against a black opponent in
either election. Finally, since I want to be as comprehensive as possible I
include all cases in cities with over 100,000 people that ¬t this criterion.
The two key questions that I try to address with this ¬rst data set are: (1)
Are white voters more or less likely to support black challengers than they
are to support the same black candidates when they run as incumbents?
(2) Does black incumbency at the mayoral level lead to changes in the
nature of the white vote? In particular, as time goes by under black lead-
ership, is the white vote less governed by racial aspects of the election
and more focused on nonracial factors such as endorsements, candidate

17 Even a pure realistic group con¬‚ict model, which might predict a positive change in white
voting behavior “ given that black leadership proved ineffective “ would still not predict
a positive change in white attitudes.
Changing White Attitudes

quality, and the economy? In addition, to see how voters behaved across a
larger array of cases and over a longer time period, I examine the outcome
of every reelection bid of every black mayor in cities with over 50,000 res-
idents in the twentieth century. The goal here is to show that, across the
range of cases, white voters regularly support black incumbents.
The main tool to assess change in white racial attitudes is a series of
nationally representative public opinion surveys conducted by the Ameri-
can National Election Study (ANES). With an array of questions on racial
attitudes, a suf¬ciently large sample of white respondents, enough geo-
graphic dispersion, and a long time period, the ANES is the only survey
that can be used to assess the effects of black representation on white
attitudes.Two questions are critical to gauging the impact of black lead-
ership. First, controlling for selection effects, do whites who live under
black mayors express less racial fear and less negative views toward black
leadership and the black community than do whites who live in other
cities? Second, in cities that have elected a black mayor, is there an appre-
ciable change in white attitudes over time? In particular, do white atti-
tudes toward blacks and black leadership improve as experience with
black of¬ceholding increases? These main data sets, coupled with other
secondary data, should provide a fairly complete picture of white reac-
tions to black mayoral representation and a fairly discerning test of the
information model.
There are, of course, some limitations to the analysis. Although I am
able to measure the white vote and white views at key points before,
during, and after the transition from white to black leadership, I offer
less in the way of direct evidence of the speci¬c information that whites
residents have about black leaders. Exactly what is it that white residents
do or do not know about local black leadership at different points in
time? Strictly speaking, this kind of evidence is not required to test the
information model but this kind of data might help to give us a more
detailed understanding of the learning process. Unfortunately, survey data
asking whites what they did or did not know about different aspects of
black leadership at different points in time is quite limited and exists in
only a few cities. Nevertheless, in the book I do utilize the data that are
available. As I detailed earlier in this chapter, survey data not only strongly
suggest that most white residents know if they have a black mayor but
also that they are quite willing to offer evaluations of local conditions
and mayoral performance. As Chapters 5 and 6 will demonstrate, these
evaluations have a substantial impact on mayoral approval (see also Stein,
Ulbig, and Post 2005; Howell and Perry 2004; and Howell and McLean
Black Leadership: The Possibilities 35

2001). In addition to data from larger surveys, I also present data from
a range of interviews with a number of white residents in a variety of
cities. These interviews do not emerge out of a random sample design
but they nevertheless indicate what it is that many whites fear before
black leaders are elected and more importantly how those fears decline
as the years under black leadership go by. Finally, I also attempt to tie the
political behavior of white residents more directly to the actions of black
leadership by matching changes in local conditions under black mayoral
leadership to changes in the views of white residents over time in two
in-depth case studies of Los Angeles and Chicago.
The other main limitation of the analysis is that I could not follow a
representative set of individuals as they experienced the transition from
white to black leadership. As bene¬cial as such panel data would be to
assess change over time, they are simply not available. Instead, I use two
alternate types of analyses to assess change over time. To measure changes
in the white vote, I focus on aggregate rather than individual data. Using
this aggregate data I can easily compare the white vote at different points
over the course of black leadership and note any changes that occur
over time. To measure changes in white attitudes using pooled individual
data, I develop a set of more complex methodological tests that attempt
to isolate the effects of black leadership. Speci¬cally, I develop a two-
stage least-squares model that controls for selection of individuals into a
particular city and I incorporate in that model a range of factors other than
black leadership that might in¬‚uence white attitudes. In addition, to assess
changes over time in white views, I not only compare white attitudes in
black-led cities to white attitudes in white-led cities, I also compare white
attitudes early in a black mayor™s tenure to white attitudes in the latter
years of a city™s experience with black mayoral leadership “ a comparison
that allows me to get reasonably close to measuring change over time.
Finally, in the analysis of both the white vote and white racial attitudes, I
explicitly control for changes in voter turnout and out-migration to help
ensure that the changes we see under black leadership are real. No data
in social science, short of experiments, can prove a temporal or causal
connection with absolute certainty, but the tools that I employ should
provide a fairly accurate assessment of the changes that occur over time
under black representation.
Another potential concern is selection bias. If the cities that elect black
mayors are cities where white residents are especially tolerant or cities
where whites are on the cusp of accepting black leadership, it is possible
that the positive change we see under black representation is the result
Changing White Attitudes

of selection and not a sign of a real change in white political behavior.
There are a number of factors that indicate that selection bias is not a
signi¬cant problem in this case. Many of these are discussed in Chapters 2
and 3 but several are worth highlighting here. The most obvious reason
to discount this selection hypothesis is the strongly negative reactions of
the vast majority of white residents to the candidacies of black mayoral
challengers. Some 70 percent of all whites voted against the successful
black challenger in the cities that I examine. Many of these white voters
were quite willing to express their animosity toward black leadership and
their fears about what would happen if blacks gained control. In about
half of these cities, whites turned out at record or near record rates to try
to prevent a black victory. Moreover, as Chapter 2 will show, when whites
voted in these elections, racial concerns all but determined the white vote
in these cities. Clearly, in the cities that did elect black mayors, most white
residents were far from ready to support black leadership. Another reason
to doubt the selection hypothesis is the fact that an examination of surveys
by the ANES shows that the racial attitudes of white residents in cities
prior to the election of a black mayor were not appreciably different from
the racial attitudes of whites in cities that were not about to elect a black
mayor. Put another way, whites in cities that elect black mayors for the
¬rst time are similar to whites in cities that do not elect black mayors.
Third, the available evidence suggests that a particularly racially tolerant
white population is generally not the main reason for the election of a
black mayor (Karnig and Welch 1980). In fact, the presence of a black
mayor has much more to do with the size and resources of the local black
community than with the nature of the local white community (Karnig and
Welch 1980). Finally, when I do assess changes in white racial attitudes, I
undertake a series of tests to help ensure that changes in white views are
not due to selection bias. The ¬rst and most important tool is the use of
a two-stage least-squares model that directly controls for selection into
cities with black mayoral leadership. In addition, I look to see if the change
in white views under black mayors is evident even when I only include
whites from cities that have elected a black mayor at some point in their
history. By con¬ning the analysis to whites who live in cities that have
or will elect black mayors, any change in white attitudes that is evident
during the years when a black incumbent is in power cannot be due to
especially racially tolerant cities electing a black mayor since all of the
cities in the analysis elect black mayors. Also, as previously noted, when
assessing changes in white views, I compare white views early in the tenure
of a black mayor to white views later in the tenure of a black mayor. This
Black Leadership: The Possibilities 37

additional test indicates that even when the analysis is con¬ned to whites
who currently live under a black mayor, there is a marked improvement
in white views over the tenure of the black mayor “ a change that is not
likely to be due to especially racially tolerant cities electing black mayors.
Combined, these points strongly suggest that the changes we see under
black mayors are real.

The Transformation of the White Vote

In this chapter, I begin to assess how experience with black leadership
affects whites™ political behavior. The goal of the ¬rst part of the chapter
is to offer a test of the three competing explanations of white voter behav-
ior developed in Chapter 1. If the information model is accurate, white
Americans should be more supportive of black incumbents than of black
challengers. If the predictions of the white backlash model are correct,
whites™ opposition to black incumbents should increase with time. And if
the racial prejudice model is accurate, the white vote should be relatively
unaffected by black leaders™ incumbency. The results of this ¬rst test are
fairly clear: white voters are signi¬cantly more willing to support the same
black candidate when he or she runs as an incumbent. Regardless of who
they face or where they run, black incumbents usually win reelection.
This ¬rst test cannot tell us why white voters change their minds about
black candidates, of course. This is an important omission, because there
are a number of reasons why black incumbents might get more white
support than black challengers. After all, most candidates, black or white,
are able to garner more votes when they run as incumbents. To address
this issue and, more importantly, to see if racial learning plays a role in
increasing white support for black incumbents, I will look more deeply
at the nature of the white vote in black challenger and black incumbent
elections in the second part of the chapter. If the information model is
correct, we should ¬nd black incumbent contests to be somewhat less
racialized: threat should play a diminished role in the vote of a signi¬cant
portion of white voters, and conventional nonracial factors should re-
emerge as critical considerations. The results of this second test show that
in black incumbent elections, white voting patterns more closely resemble
The Transformation of the White Vote 39

the norm than they do in black challenger elections: turnout declines to
average levels, racial considerations seem to diminish, and traditionally
important electoral in¬‚uences such as endorsements and candidate quality
re-emerge as central factors in the white vote.
The results of both of these tests support the information model. But if
the model is accurate, some of the positive information that whites obtain
from watching black incumbents around the country should be relevant
when new black challengers enter the political arena. Over time, whites
should grow less fearful of new black challengers, and they should be more
willing to vote for these candidates. Since black leadership is certainly
not the only new element affecting white votes over time, my test of
this hypothesis is at best suggestive. Nevertheless, the results of this test,
too, support the information model, as the information whites acquire
about blacks in one place does appear to affect outcomes in other contests
in other locations. Combined, these three tests would seem to indicate
that black leadership does provide many white Americans with critical
information that allays their fears and reduces the signi¬cance of race in
subsequent local elections. Before turning to the tests and analysis of their
results, I want to ¬rst provide a brief review of the existing literature on
white voting in biracial contests.

will whites vote black?
Scholars, activists, and judges have all given enormous attention to the
simple question “Will whites vote black?” For many, it is the central ques-
tion in discussions of minority rights in this country. Despite the attention
that has been given to this question “ or perhaps because of it “ existing
scholarship has provided no clear answer. Some scholars are now con-
vinced that whites will vote black. They cite public opinion surveys in
which white respondents overwhelmingly report that they are willing to
support black candidates, and they point out that the list of African Amer-
icans who have won of¬ce in primarily white states, districts, and cities is
impressive.1 They also emphasize exit poll data from a range of congres-
sional and gubernatorial elections that suggest race is no more an issue in
biracial elections than it is in other electoral contests in America (Highton
2004; Citrin, Green, and Sears 1990). Several scholars have gone so far as

1 When asked if they would vote for a “quali¬ed black candidate for president,” over 90%
of white respondents answer yes (Schuman et al. 1997). The number of whites who say
they would vote black are even higher for other political of¬ces (Williams 1990).
Changing White Attitudes

to interpret these results as a sign that race is largely irrelevant in biracial
contests (Thernstrom and Thernstrom 1997; Thernstrom 1987; Swain
1995). As Abigail and Stephen Thernstrom claimed in 1997, “whites are
voting black “ in increasing numbers “ to a degree that was unimaginable
30 years ago” (Thernstrom and Thernstrom 1997: 295).
But others have reached a very different conclusion. When given the
choice between a white candidate and a black candidate, these scholars
maintain, the vast majority of white voters will embrace the candidate
of their own race. There is considerable evidence to support this view.
Studies that have examined electoral outcomes in hundreds of contests
suggest that in a typical biracial contest one can expect an average of
70 to 90 percent of white voters to vote white (McCrary 1990; Bullock
and Dunn 1999; Henry 1987; Stein and Kohfeld 1991; Loewen 1990;
Bullock 1984; O™Loughlin 1979; Murray and Vedlitz 1978; Black and
Black 1973; Lieske and Hillard 1984; Shef¬eld and Hadley 1984).2 More-
over, these and similar studies ¬nd that the more important the of¬ce, the
fewer whites are willing to vote for a black candidate to ¬ll it (Stein and
Kohfeld 1991; Williams 1990; Bullock 1984). The geographic distribu-
tion of black elected of¬cials also strongly hints at an ongoing aversion
to black candidates on the part of white voters. Today, roughly 80 per-
cent of all black elected of¬cials are elected by majority black electorates
(Canon 1999; Handley and Grofman 1994; Hedge, Button, and Spear
1992; Campbell and Feagin 1984). Even though 70 percent of all blacks
live in state and federal districts that are majority white, only about 1
percent of all majority white districts have ever elected a black of¬cial
(Handley and Grofman 1994). The clear implication is that whites will
not vote black and that race remains central in the minds of white voters.
Accordingly, many scholars believe that race and racial prejudice remain
the primary factor in American politics (Reeves 1997; McCrary 1990;
Huckfeldt and Kohfeld 1989).
How can scholars reach such starkly different conclusions about the
role of race in American politics? The answer is that all black candidates
are not equal in the minds of white voters. A sharp distinction should be
made between black challengers on one hand and black incumbents on

2 Findings in experimental studies are slightly more mixed. Some studies have found that
whites support a black candidate as much or more often than an equivalent white can-
didate (Sigelman et al. 1995; Colleau et al. 1990), but others have found that race, skin
color, and individual racial prejudice all affect voting (Terkildsen 1993). There are also
signs that whites appear to be hiding their intentions. In one experiment, the proportion
of whites stating that they were undecided doubled in biracial contests (Reeves 1997).
The Transformation of the White Vote 41

the other. For white voters, black challengers represent uncertainty, fear,
and, often, a racial threat; if whites do not know how black leadership will
affect them, they will generally choose to vote white. Black incumbents,
by contrast, are likely to be relatively well-known commodities “ they
have a record, and in all but a few cases that record allays fears about the
consequences of black leadership. With this new, reassuring information,
whites become much more willing to support black incumbents. Thus,
it would not be surprising to ¬nd that those studies that reach positive
conclusions about white willingness to support black candidates often do
so because they tend to focus almost exclusively on black incumbents.
Similarly, those that offer more pessimistic accounts of American race
relations and white willingness to vote for African American candidates
do so, perhaps, because they tend to focus on black challengers. Unfortu-
nately, since almost all of the existing research on racial politics has failed
to take into account the important distinction between challengers and
incumbents in data analysis, we do not yet know whether this assumption
is correct.
If this information model is accurate, the critical question is not
whether whites will vote for blacks but under what circumstances they will
vote for blacks. And, more speci¬cally, what difference does black incum-
bency make? Does experience under black incumbents change the way
whites think about black candidates, make them more willing to support
black incumbents, and reduce the role of race in biracial electoral contests?

does incumbency matter?
To begin to answer these questions, I collected data on white voting pat-
terns in a representative sample of mayoral elections involving black can-
didates. I collected these data with two goals in mind. My ¬rst goal was to
provide as direct an assessment as possible of the impact of incumbency
on the white vote. To do so, I amassed data on white voting patterns
in sets of two mayoral elections in cities that have experienced a transi-
tion to black leadership. For each case, I contrast the white vote in the
¬rst election, in which a black challenger ran successfully against a white
incumbent to become the ¬rst black mayor of the city, with the white
vote in the election immediately following, in which the black mayor
ran for reelection against a white opponent.3 By comparing sets of two

3 Including cases in which two black candidates run against each other would, obviously,
reveal little about white acceptance of black leadership.
Changing White Attitudes

elections that involve the same black candidates, I am able to assess the
effects of incumbency on the white vote directly. I con¬ne my analysis to
general or run-off elections rather than primaries to avoid complications
introduced by multiple candidacies and voter disinterest. To analyze other
aspects of the electoral outcome, I also collected data on overall turnout
and the margin of victory in each election. A detailed account of each of
the variables and its sources is described in the statistical Appendix to
Chapter 2.
My second goal was to be as comprehensive as possible in order to
ensure that the results of the data analysis are representative. Since all
previous studies had considered only a small number of cases, I decided
to create a complete data set that included all relevant cases across the
country. To do this, I compiled a set of the entire universe of cases for
cities with populations of over 100,000 that ¬t the criteria just outlined.
In total, there were ¬fty-two elections in twenty-six cities. While this is
admittedly a small number, it represents two-thirds of the cases of white-
black transition in large American cities. What is happening in this set of
cases, then, should be more or less what happens generally when a white
mayor is replaced by a black mayor in a large American city.
It is also important to note that my selection criteria do not appear
to have created a set of cities with exceptionally liberal or especially
racially tolerant white populations. Although some of the cities, such as
San Francisco, Minneapolis, and Seattle, are generally seen as liberal, oth-
ers, including Memphis, Birmingham, and Houston, would be much more
likely to be labeled conservative, and still others, Durham and Hartford,
for example, fall somewhere in the middle. As we will see, most whites
in these cities were not ready for black leadership and not particularly
racially tolerant when black candidates were trying to win the mayoralty
for the ¬rst time. On the contrary, black challengers in many of these cities
faced nearly unanimous opposition. And, in many cases, whites turned
out in record numbers to try to prevent a black victory. In fact, a compar-
ison of the racial attitudes of white residents in these cities prior to the
election of a black mayor with the racial attitudes of white residents in
other cities using the survey data from the ANES (presented in Chapter 3)
found no consistent or substantial differences in white views. For these
cities, the key to black victory was the black vote, not white support. This
comports with existing research that suggests that the size of the black
community and the resources of the black community are much more
important in determining the success of black candidates than the nature
of the white community (Karnig and Welch 1980).
The Transformation of the White Vote 43

table 2.1 Voting Patterns in Black Challenger and Black
Incumbent Elections

Black Challenger Black Incumbent
White Voters for Black Candidate (%) 30 36
Margin of Victory (%) 12 21
Turnout of Registered Voters (%) 59 52

To illustrate how white voters respond to black mayoral leadership,
Table 2.1 presents a comparison of black challenger and black incum-
bent elections. The numbers tell a fairly clear story: when the same black
candidate runs for reelection for the ¬rst time as an incumbent, the propor-
tion of white voters who support that candidate grows by an average of
6 percentage points, from 30 to 36 percent of all white voters. A six-point
shift in the vote is certainly not unheard of in American elections, and
one could argue that this change represents relatively little movement on
the part of white voters. Yet this relatively small change is clearly impor-
tant, for if whites were reacting to incumbent black mayors as they have
responded to other forms of black empowerment in the past, we would
have seen the opposite: a white backlash characterized by heightened
mobilization and resistance. Similarly, if prejudice were the main factor
behind white opposition to black candidates, we would most likely see no
change at all. The fact that white support grew, even if by a small amount,
is very informative.4
The growth in white support is more impressive when one consid-
ers that whites in these cities had only two or four years (depending on
the length of a mayoral term) to experience black leadership. In cities
like Los Angeles and Newark, where the same black mayor ran repeat-
edly for reelection, white support grew with each election. According to

4 It is also worth noting that the increased white willingness to support the same black
candidate suggests that a fairly large portion of white voters who were motivated by race
in the challenger contest had a change of heart in the incumbent contest. If most of the
white support in the black challenger contests and most of the increase in white support
in the incumbent contests comes from Democrats “ a reasonable assumption given that all
but one of the black candidates are self-described Democrats and most advocated at least
a marginally liberal agenda “ a shorthand calculation (estimating that half of all white
voters in these cities are Democrats) indicates that one-quarter of the white Democrats
who voted “race” instead of party in the black challenger election reverted to voting along
partisan lines in the black incumbent election. In other words, a substantial proportion
of all racially motivated white voters change their minds after just a few years of black
Changing White Attitudes

Sonenshein (1993), Tom Bradley™s white support in Los Angeles grew in
each of his ¬rst four elections. All told, his white support almost doubled
from 32 percent in 1969 to 62 percent in 1985. Thus, the six-point shift
may represent only the ¬rst step in growing white acceptance of black
leadership. In addition, this analysis in some ways understates the excep-
tional nature of the white support that these black incumbents won. I do
not compare the average challenger to the average incumbent but instead
focus only on the most successful black challengers. Most black chal-
lengers lose their electoral bids. Thus, if I had included a cross-section of
all black challengers, the contrast between support for challengers and
support for incumbents would be much greater. The limited data that
are available attest to this point. In an analysis of a series of city council
and mayoral elections in Atlanta, Bullock (1984) found that incumbency
more than doubled white crossover voting. His ¬ndings were echoed in
an analysis of the vote in mayoral and council elections in New Orleans
(Vanderleeuw 1991).
It is worth noting that the black candidates in the sample gained sub-
stantial white support as incumbents despite the fact that they did not get
the boost in electoral resources that most incumbents receive. For most
white candidates, incumbency has enormous bene¬ts: it usually means
more endorsements, more money, and weaker opponents. This is much
less true for the black candidates in my sample.5 Largely because they
needed tremendous resources to be elected in the ¬rst place, the majority
of these twenty-six black candidates garnered few new electoral resources
as incumbents. In 81 percent of the cases, they received no new Demo-
cratic Party endorsement when they ran as incumbents. In 62 percent of
the cases, they gained no new endorsements from local newspapers. These
black incumbents also tended to face strong white challengers. Speci¬-
cally, 62 percent of the incumbents faced opponents who had the same or
a higher level of experience than their opponents in the challenger elec-
tion. The candidates were able to muster only marginally greater ¬nancial
resources as incumbents, and one-third actually raised less money than
they had as challengers. It would make little sense, then, to attribute the
growing white support for these candidates to the conventional resources
of incumbency.
The six-point increase in the percentage of white residents who vot-
ed black was not the only signi¬cant change from the challenger to the

5 Data sources for these comparisons are described in the appendix.
The Transformation of the White Vote 45

incumbent elections. There was an even sharper decline in the abso-
lute number of white voters who opposed the black candidates.6 Across
all twenty-six cities, the number of white votes for the white candidate
declined by 19 percent on average between the challenger and the incum-
bent elections. This result suggests that as many as one-¬fth of all white
voters who opposed black leadership may have changed their minds suf¬-
ciently either to support the black candidate or to choose not to vote at all.
As a consequence of both the drop in voter turnout and the higher level
of support for the black incumbents, the incumbents™ average margin of
victory jumped from 12 percent in the challenger elections to 21 percent
in the incumbent elections, leading to victory for the black incumbents in
all but three cities.
The ¬nal factor to consider is voter turnout. Table 2.1 reveals that
turnout decreased substantially in the black incumbent elections. In a lit-
tle over half of the challenger elections, turnout had reached or exceeded
record levels.7 On average, it exceeded the national average by over 10 per-
centage points (Hampton and Tate 1996). But this mobilization quickly
faded away when blacks ran as incumbents: across the twenty-six cities,
turnout dropped from almost 59 percent in challenger elections to 52
percent in incumbent elections, falling in many cases to average or below
average levels. In Charlotte, for example, where Harvey Gantt faced well-
known white Republican city council members in both of his elections,
voter turnout fell by over 15 percentage points from 50 percent in Gantt™s
challenger run to 34 percent, near the historic norm, in his reelection
bid. From this data, it seems that black incumbency at the mayoral level
transforms extraordinary black challenger elections into more ordinary
contests for reelection.
The opposition that the black challengers faced was by no means totally
erased when they ran as incumbents, of course. The data in Table 2.1

6 This assumes that the decrease in support for the white candidate comes from white voters.
Two facts make this assumption reasonable. First, there was almost no black support for
the white candidate in any of these elections. Exit polls and precinct analysis of the black
vote indicate that on average 95% of black voters supported the challenger and 93%
supported the incumbent. Also, across the cities, there are few voters who are neither
black nor white. Blacks and whites combined make up 92% of the population in these
twenty-six cities.
7 Although Table 2.1 only presents aggregate turnout rates, it is clear that white turnout
rates follow the same pattern. In cities where turnout ¬gures are available for whites
and African Americans separately, the numbers suggest that both groups turn out in
large numbers in challenger elections and in considerably smaller numbers in incumbent
Changing White Attitudes

indicate that large numbers of white voters continued to oppose the black
incumbents. But in the average case, after a few years of black incum-
bency, white Americans became more accepting of black leadership.
Again, the most remarkable aspect of this shift was not its size but the
fact that there was any positive change at all. Peter Eisinger noted, in his
study of Atlanta and Detroit, how sharply the elections of black rep-
resentatives in those cities contrasted with expectations: “What has
occurred is particularly noteworthy when it is set against the history
of race relations in those two cities themselves, against the habits of
racial oppression in American society in general, and indeed against a vir-
tually worldwide tendency to deal with ethno-racial political competition
by violent means” (Eisinger 1980: xxi). In many cities, even city residents
themselves seemed surprised at their mayoral election™s outcome. As one
reporter in Birmingham put it, “This city, once branded by the Rev.
Martin Luther King Jr. as ˜the most thoroughly segregated in America,™
accomplished something Tuesday that many of its residents consider
remarkable: it reelected its ¬rst black mayor with a biracial coalition and
the largest victory margin in city history” (Russakoff 1983). The fact
that whites™ anti-black mobilization declined after only a few years and
signi¬cantly more whites became willing to support black leadership
was not only a positive sign for race relations in these cities “ it was
a positive change that many did not foresee.

A Broader Phenomenon: All Incumbent Black Mayors
The changes in white voter behavior noted above may be unique to the
twenty-six cities in the data set or limited to the ¬rst few years of black
leadership. To assess black incumbency more broadly, I collected data on
the reelection bids that took place in the twentieth century of every black
incumbent mayor in every city with a population over 50,000.8 In each
case, in addition to the outcome of the contest, I obtained information
on the racial makeup of the city, the race of the opponent, the number of
terms the incumbent had been in of¬ce, and the number of black mayors
who had previously served in the city.

8 This data set was compiled using the National Roster of Black Elected Of¬cials, local
newspaper reports in each city, and a data set of mayoral names (Wolman, Strate, and
Melchior 1996), and it includes the race of the mayor, the challenger, and the winner. As
in the ¬rst data set, I focus on general or run-off elections rather than primaries, where
factors such as multiple candidacies, lack of interest, and limited availability of empirical
data complicate empirical analysis.
The Transformation of the White Vote 47

The ¬ndings from my analysis of this larger set of cases echo the results
for the twenty-six cities. First, black incumbents won in the vast majority
of the cases. Since 1965, black mayors have won 78 percent of their
reelection bids (98 out of 126 cases). In fact, depending on the exact
comparison, black incumbents do almost as well as or even better than
white incumbents. Between 1970 and 1985, the only period for which
I was able to obtain equivalent data for both black and white mayoral
incumbents, black mayors were reelected 89 percent of the time (31 out
of 35 cases), a slightly higher rate than white mayors, who were reelected
84 percent of the time (359 out of 429 cases). From these data, it would
seem that black and white incumbents are treated almost equally by the
American electorate.
Second, there was no sharp decline in black reelection rates over time,
and thus little indication that the information provided by black incum-
bents was losing ef¬cacy over time. Although the ¬rst African Americans
to serve as mayors of their cities were particularly successful when they
ran as incumbents for the ¬rst time (winning 83 percent of these reelection
bids), they also did well in subsequent electoral bids, winning 74 percent
of the time. Equally important, there does not appear to be a major dis-
tinction in success rates between the ¬rst black mayor of a city and others
who follow. The overall reelection rate of cities™ ¬rst black mayors (80 per-
cent) by no means dwarfs the reelection rate of subsequent black mayors
(73 percent).
But does the success of black incumbents have anything to do with
white voters? After all, the majority of black mayors represent minority
white cities. Given the fact that black voters tend to favor black candidates
over white candidates, the success of black incumbents could merely be an
artifact of black unity and voting strength and not the result of increasing
white support (McCrary 1990). But this appears not to be the case: if we
con¬ne the analysis to minority black cities, where white voters presum-
ably have a good chance of controlling the outcome of the contests, black
incumbents still do well. Black mayoral incumbents in minority black
cities won reelection over 80 percent of the time, only marginally below
the overall white incumbent reelection rate. Moreover, black incumbents
did not win these contests simply because white voters were forced to
choose between two black candidates. Even in minority black cities in elec-
tions in which black incumbents faced a white challenger, black mayoral
incumbents won reelection 73 percent of the time (19 of 26 cases). In fact,
black incumbents actually did better against white candidates then they
did against black candidates.
Changing White Attitudes

It is also worth noting that the pattern of declining turnout that was
evident in the twenty-six black challenger/black incumbent elections can
be found across the wider range of cases. In an analysis of a broad sample
of mayoral elections in large American cities, Hampton and Tate (1996)
found that the mere presence of an African American candidate in the
contest raised turnout by 10 points over the national average. But this
changed when blacks ran as incumbents. The same data show that hav-
ing a black incumbent in the contest actually lowered turnout by 4 percent
(Lublin and Tate 1995). Black challengers spark extraordinary mobiliza-
tion, while black incumbents seem to spark only average interest.

the shifting calculus of the white vote
Across a wide range of cases and on a number of different measures,
black mayoral leadership appears to lead to positive changes in white
political behavior. These positive changes seem to favor the information
model over both the backlash and the prejudice hypotheses, but they do
not themselves demonstrate racial learning on the part of white voters.
There are a number of possible reasons why black incumbents might
be successful and why white voters might change their minds about
black leadership. Most of¬ceholders, whether they are white or black,
get more support when they run as incumbents. To see if race and racial
learning are behind the changes in the white vote observed in the data,
more tests are required. In this section, I begin to examine the nature
of the white vote more closely to see if the change in the vote can be
linked to information. If the information model is accurate, we should
see a distinct pattern emerge: in black challenger elections, the white vote
should be largely based on racial fears; in black incumbent elections, fear
should play a diminished role, and white voters should begin to base
their votes on the track record of the incumbent and the speci¬cs of the
I collected an array of data on the campaigns and candidates for each
of the ¬fty-two elections in the original set of twenty-six cities. To assess
the role of race and fear in each contest, I included two different kinds of
measures. First, I used a measure of the black population size as a proxy
for racial threat. The size of the black population is regularly employed
as a measure of racial threat, and in a wide range of cases white political
choices have been shown to be shaped by the local racial context (Giles
and Hertz 1994; Key 1949). If fears about the consequences of black
The Transformation of the White Vote 49

leadership are in fact driving the white vote in black challenger elections,
we should ¬nd that white voters™ preferences are closely tied to the size of
the black community. The larger the black population and the more likely
it is that blacks could actually gain control of the local political arena, the
more we should see whites fearing black leadership and voting against
black candidates.
Second, I included a measure of the racialization of each black candi-
date™s campaign. If fear about racial change is behind white opposition,
then what black candidates do or say regarding racial policy should also
affect the white vote. The less black candidates talk about serving the black
community and the more they run deracialized campaigns that promise a
race-neutral administration, the less fear there should be in the minds of
white voters and the more likely it should be that white voters will sup-
port black candidates.9 To measure the racialization of a given campaign,
I coded the extent to which the black candidates™ speeches, policy plat-
forms, and mobilization efforts were targeted at blacks, whites, or both.
This is admittedly a subjective measure, but in practice it was fairly easy
to divide campaigns into three categories: campaigns that had any sort
of explicit, pro-black focus; campaigns that addressed the black commu-
nity implicitly through a generally pro-black policy agenda or by actively
mobilizing black voters and speaking before black audiences; and cam-
paigns that never mentioned black interests and were fairly race neutral.10
A comparison of the racialization measure employed in this study with a
similar measure used in Lublin and Tate (1995) suggests that the coding is

9 This type of deracialization hypothesis has been the subject of a lengthy debate, with
some believing that deracialization is a critical strategy for black candidates and others
maintaining that it is impossible for black candidates to effectively downplay the signi¬-
cance of race (C. Hamilton 1977; Henry 1992; Perry 1996; but see Starks 1991; Wright
10 In practice, most black candidates ran dual campaigns, using different tactics and address-
ing different issues depending on the racial makeup of their audiences. Few campaigns
were overtly racial. Fewer still were clearly race neutral. In the end, the range of campaigns
was not that wide. Tom Bradley, who is seen as having the quintessential deracialized
campaign, talked about af¬rmative action and the problems of the black community.
On the opposite end of the spectrum, Harold Washington did make the famous “It™s our
turn” comment, but the vast majority of the time he avoided mentioning black interests
and instead talked repeatedly about white interests, about serving the whole city, and
about such issues as reform that primarily interested white liberals. Very few campaigns
were coded as explicitly racially focused. In black challenger campaigns, there was an
even split between implicitly pro-black and race-neutral campaigns. In black incumbent
elections, race-neutral campaigns were a slight majority.
Changing White Attitudes

valid.11 If the information model is accurate, this measure should have a
bigger effect on the white vote in black challenger elections than in black
incumbent elections.
If whites cease to fear the consequences of a black takeover, conven-
tional nonracial factors that are normally important determinants of elec-
toral outcomes should begin to play a more signi¬cant role in black incum-
bent elections. To see if this is the case, I examine the extent to which three
basic factors of the electoral context affect the white vote in black chal-
lenger and black incumbent elections: candidate quality, political endorse-
ments, and campaign spending. In contests at almost every level of politics,
each of these factors has proven to be critical to electoral outcomes. More-
quali¬ed candidates “ with quality generally measured in terms of polit-
ical experience “ surpass the electoral fortunes of less-experienced can-
didates at both the congressional and local levels (Jacobson and Kernell
1981; Krasno and Green 1988; Lieske 1989; Krebs 1998).12 Similarly,
major endorsements have been shown to play a primary role in most
local contests. In particular, both political party endorsements and city
news-paper endorsements affect voting in local elections (Bullock 1984;
Lieske 1989; Krebs 1998). Finally, campaign spending has been closely
linked to the electoral fortunes of candidates from presidents all the way
down to city council members (Arrington and Ingalls 1984; Krebs 1998;
Green and Krasno 1988; Jacobson 1980; Cox and Munger 1989; Gierzyn-
ski 1998). Candidates who are able to outspend their opponents by wide
margins seem to be much more likely to win at the polls.13 Measures
of each of these factors are fairly straightforward and are detailed in
Appendix A.

11 Coding from the current data set was compared to the coding of a similar variable from
the Race and Urban Politics Data set (RUPD), a data set covering 315 mayoral elections
in twenty-six large cities. The RUPD variable, which purportedly measured the racial
character of a candidate™s campaign, correlated at 0.97 with the racialization measure
used here (see Lublin and Tate 1995 for a detailed description of the data set).
12 The exact causal relationship between candidate quality and electoral outcomes is, how-
ever, likely to be somewhat complex. Strong incumbents often deter quality challengers
from entering the contest in the ¬rst place, which makes an assessment of the causal ties
dif¬cult (Bond, Covingten, and Fleisher 1985; Jacobson and Kernell 1981).
13 The relationship between money and outcomes is also somewhat complex. Some feel that
challengers can bene¬t more from spending than can incumbents (Jacobson and Kernell
1981). Others see reciprocal causation: how well a candidate is expected to do affects
the candidate™s fundraising ability (Goidel and Gross 1994). In light of this ¬rst claim, I
look at the spending of each individual candidate separately and not just at the spending
advantage of one candidate over another.
The Transformation of the White Vote 51

table 2.2 Determinants of the White Vote in Black Challenger

White Support for the
Black Candidate
’0.72 (0.13)———
Percentage Black of City Population
’0.58 (0.26)—
Racialization of Black Candidate™s Campaign
Candidate Quality
Quality of White Opponent 0.01 (0.08)
’0.09 (0.07)
White Incumbent Running
’0.05 (0.08)
Quality of Black Challenger
Democratic Party Endorsement 0.02 (0.05)
Local Newspaper Endorsement 0.01 (0.05)
0.64 (0.11)———
Adj. R2 0.67
N 25
Note: OLS regression. Figures in parentheses are standard errors.
——— p < 0.01
—— p < 0.05
— p < 0.10

The Importance of Race in Black Challenger Elections
In Table 2.2, I begin to test these propositions by analyzing the aggregate
white vote in black challenger elections. Although the number of cases
is relatively small and is not necessarily representative of all American
cities, the table does reveal a stark, clear pattern. As predicted by the
information model, when black candidates challenged for the mayoralty
for the ¬rst time, the aggregate white vote was tied almost exclusively to
racial fears.
The ¬rst measure of racial fear indicates that the larger the black
population in the cities in the sample “ and hence the greater the perceived
threat that blacks would gain some measure of control over the local
political arena “ the less willing whites were to support a black challenger.
The size of the black population accounts for the bulk of the variation in
aggregate white behavior; by itself, it accounts for 60 percent of the vari-
ation in white vote choice. Even considering the selection bias inherent in
these cases, it is impressive how closely the white vote was tied to the size
of a city™s black population. In the ¬ve cities with the highest proportion
of African Americans, Baltimore, Birmingham, New Orleans, Memphis,
Changing White Attitudes

and Newark, on average only 16.9 percent of whites supported the black
challenger. In contrast, in the ¬ve cities where blacks represented the
smallest proportion of the population and thus the smallest threat, a slim
majority of white voters (on average 50.4 percent) supported the black
candidate. Overall, the regression results indicate that a 10 percentage
point increase in the proportion of a city™s residents who were black led
to a 7.2 percentage point drop in white support for the black mayoral
The importance of racial threat seems to suggest that these challenger
elections were less about the candidates or the speci¬cs of the election
than they were about the size of the threat of a black takeover “ a conclu-
sion that is echoed over and over again in accounts of the elections. One
of the most well-known accounts of Birmingham™s election, for example,
concluded that “whites worried not so much about Richard Arrington
Jr. [the black challenger], but about blacks, the group they believed he
represented. Had that day now come when ˜the last shall be ¬rst, and the
¬rst shall be last?™” (J. Franklin 1989: 172). The transition was viewed
very similarly in Atlanta, where Peter Eisinger found that “the change was
understood not in terms of a turnover in the personnel of city hall but
as a loss by one race to the other” (1980: 154). Wilbur Rich™s account of
Detroit reached the same conclusion: “Many white residents of Detroit
responded to the 1973 election of Coleman Young with intense apprehen-
sions and fear. . . . Many whites saw the race as the last stand before the
takeover by the onrushing black majority” (1987: 208).
The role played by the black candidate™s campaign, and in particu-
lar white voters™ aversion to racially focused campaigns, also serves to
con¬rm the critical importance of racial fears in these black challenger
elections. What black candidates did or did not say about the interests
of blacks apparently in¬‚uenced the white vote in these contests. All else
being equal, black challengers who ran essentially race-neutral campaigns
garnered almost 60 percent more of the white vote than challengers who
ran racially explicit campaigns. Even though all of the twenty-six black
challengers tried in some way to assure white residents that they would not
be ignored, white voters seemingly keyed in on small differences between
campaigns. Thus, a candidate like Harold Washington probably lost sub-
stantial white support as a result of telling a black audience “It™s our
turn,” even though most of his campaign was race neutral. And, at the
other end of the spectrum, candidates like Thirman Milner, who emphat-
ically told white voters that “there is no such thing as black legislation”
and who often repeated his desire to be “mayor of all of Hartford,” seem
The Transformation of the White Vote 53

to have been rewarded with additional white votes. Candidate Charles
Box recalled that “The key . . . was to take the fear of the unknown out of
the equation” (quoted in Colburn and Adler 2003). Box was so concerned
about racial fears that he centered his campaign in Rockford on having
personal interactions with as many white voters as possible.14
In addition to supporting the information model, these ¬ndings also
contribute evidence toward the resolution of two ongoing debates in the
literature on American racial politics. First, the clear negative relationship
between white voting behavior and the size of the local black population
in these mayoral elections reaf¬rms the important role that racial context
plays in American race relations. Existing studies have reached very dif-
ferent conclusions about how the increasing presence of racial and ethnic
minorities affects the white population. Although most studies have found
that a larger black population is associated with greater racial antagonism
(Key 1949; Murray and Vedlitz 1978; Giles and Hertz 1994; Alt 1994;
Taylor M. 1998; Fossett and Kiecolt 1989), several recent works have
concluded either that there is no relationship at all or that the relation-
ship is positive (Welch et al. 2001; Kinder and Mendelberg 1995; Bledsoe
et al. 1995; Carsey 1995). The results reported here support the position
that a proportionately larger black population does represent a racial
threat to white voters.
In addition, the relationship between the racial focus of a campaign
and the white vote seems to suggest that deracialization can lead to
increased white support. Again, there has been considerable debate on
this point. Though many have maintained that black candidates can gar-
ner white support by deracializing their campaigns (Perry 1991; Nichols
1990; C. Hamilton 1977), others disagree. In one of the most exten-
sive studies, Wright found that “black [mayoral challengers] in Memphis
were unable to garner signi¬cant white crossover support regardless of
their use of deracialized strategies” (1996: 151). Similarly, Starks contends
that “There is no way in which a contemporary American campaign can
utilize a deracialization electoral strategy and hope to eliminate race as a

14 This phenomenon does not seem to be con¬ned to mayoral candidates. Gubernatorial
candidate Douglas Wilder proclaimed, “I have never been a civil rights activist of any
kind” (quoted in Jeffries 1999). Harvey Gantt tried hard to run a nonthreatening, dera-
cialized contest in his senatorial bid against Jesse Helms (Wilson 1993). Alan Wheat, one
of the most successful black crossover candidates, was also well aware of white fears
during his Senate campaign. His campaign coordinator stated, “We knew and they knew
that there is just enough racial discomfort among whites that just showing his face was
enough of a message” (quoted in Sniderman, Swain, and Elms 1995: 10).
Changing White Attitudes

factor in that campaign” (1991: 217). But the documented results seem to
indicate that whites can be quite sensitive to the kinds of campaigns black
candidates run. When black candidates move from a racially explicit cam-
paign to a less racially focused campaign, they are able to attract greater
white support. This might lead some to recommend deracializing black
campaigns as an effective strategy to increase white support and expand
black representation. It is important to consider, however, whether any
gains in white support are large enough to offset a possible erosion of
black support and black turnout “ to say nothing of the restraints on
policy changes “ that most likely accompany deracialized campaigns.15

The Irrelevance of Conventional Political Factors in
Black Challenger Elections
Up to this point, I have detailed the factors that do matter in black chal-
lenger elections to show that the uncertainty and racial concerns high-
lighted by the information model appear to govern white reactions to
black challengers. An account of the factors that do not matter in these
contests is, however, just as telling. Looking back at Table 2.2, it is appar-
ent that none of the factors generally viewed as being critical to local
electoral outcomes substantially affects the outcome of elections in which
blacks challenge for control of a city™s mayoralty for the ¬rst time.
The ¬rst of the normally critical electoral variables tested is candi-
date quality. The evidence presented here suggests that candidates™ expe-
rience did not play a signi¬cant role in these black challenger elections:
whether black challengers faced an unknown and untested white oppo-
nent or someone who was well known and had held a prior of¬ce seems
not to have signi¬cantly affected their prospects of winning white sup-
port. Chicago provides one of the starkest examples: even though Harold
Washington faced an unknown Republican, Bernard Epton, who began
his bid with limited campaign experience, no campaign funds, and no

15 Supplementary analysis suggests that the kinds of campaigns black candidates run may
also be related to the policies that black mayors enact and the voting behavior of the black
community. Among the elections that I looked at, black candidates who deemphasized
race were also less apt to enact such policies as af¬rmative action and redistributive
spending that might bene¬t the black community. Deracialized campaigns also appeared
to be linked to lower black turnout. Neither of these relationships is particularly robust
but it may be that by shying away from race in their campaigns, black candidates lose
black votes and limit the kinds of policies that they ultimately can enact. In short, black
candidates may have to walk a ¬ne line between trying to get elected and trying to serve
the black community.
The Transformation of the White Vote 55

organization, 80 percent of white Chicagoans voted for Epton; presum-
ably this means that many of them were simply voting against Washington.
The election in Newark also appears to illustrate how little white voters
cared about the quali¬cations of the white candidate. In that city, 90
percent of white voters supported the white candidate, Hugh Addonizio,
even though he had recently been indicted on multiple counts for having
ties to organized crime and was on trial the day of the election. Perhaps
even more extraordinary is the fact that the white vote in these biracial
elections was unrelated to whether or not the black challenger faced an
incumbent. In the nineteen cases in which black challengers ran in an
open-seat contest, they did not do signi¬cantly better than in the other
six cases, where they faced a white incumbent.16
Just as white voters were not deterred by the poor quali¬cations of
white candidates, they seemed not to have been positively in¬‚uenced by
the quali¬cations of the black challengers. Having held prior of¬ce as a
state legislator or city council member did little to improve black chal-
lengers™ prospects of success. Despite the fact that Ernest Morial had been
a state legislator in New Orleans for years and Wilson Goode had been
city manager in Philadelphia in the three years prior to his mayoral bid,
both failed to garner more than a quarter of the white vote in their cities.
At the opposite end of the spectrum, black challengers such as James
Sharp in Flint and James Usry in Atlantic City, both of whom had held no
previous elected of¬ce, did only marginally worse with white voters than
their more quali¬ed peers.17
Political endorsements “ the second traditionally important factor
tested in Table 2.2 “ also do not appear to be signi¬cantly tied to the
white vote in these black challenger elections. Despite the fact that past
research has found that newspaper endorsement and party endorsements
greatly in¬‚uence the outcomes of local elections, neither had much impact
on the white vote when white voters were faced with the prospect of a per-
ceived black takeover (Bullock 1984; Lieske 1989; Stein and Fleischmann

16 Other tests also indicate that open-seat contests were in most respects no different
from elections in which the black challenger faced a white incumbent. Racial factors
were equally important in both types of contests, and nonracial factors were equally
17 Thus, Niemi and Weisberg may be right about most elections in America when they echo
the wisdom of political science and claim “It makes a difference who the candidates are”
(1993: 217), but when one candidate is black and vying to take over a position previously
held by whites, the only thing that seems to matter about the two candidates is the color
of their skin and the fear that it invokes.
Changing White Attitudes

1987; Krebs 1998). Both Kenneth Gibson in Newark and Wilson Goode
in Philadelphia won the endorsement of their city™s Democratic Party
organization and its major newspapers, usually a sure sign of widespread
white support in these strongly Democratic cities, but in each case the
endorsements were followed by strong white opposition. The same pat-
tern held in Detroit. As one reporter described the campaign, the focus
was not on endorsements or campaigns but on race: “The unions were
for [Coleman Young]. The business and political establishments were for
him. The liberal, money-laden suburbs were for him. But when they tal-
lied the votes, none of that mattered. The only issue that counted when
Coleman Young became Detroit™s ¬rst black mayor in 1973 was race”
(Cantor 1989).
Alternate tests suggest that even the policy platform of the black candi-
date™s campaign was not as important as it is normally.18 The liberalism of
the black candidate™s campaign, as measured by statements about overall
policy direction and speci¬c policy initiatives in newspaper reports and
the candidate™s speeches, had no clear effect on white support. Whether
policy liberalism was included in the model in Table 2.2 or examined
alone, it appeared to have no impact on the decisions of white voters in
these challenger elections.19
The impact of conventional factors is not simply masked by the inclu-
sion of other variables: even when both of the racial fear variables are
omitted from the regression, none of the remaining independent variables
is signi¬cantly related to white voter behavior in black challenger elec-
tions. Even simple bivariate tests reveal no clear link between any of the

18 Due to the collinearity between campaign liberalism and the racialization of the black
candidate™s campaign, the liberalism measure was not included in the ¬nal model.
19 Chicago and Philadelphia provide perhaps the starkest example of the irrelevance of
what one might call conventional politics in the face of a black mayoral challenge. Even
though the tone and content of the campaigns in the two cities were almost polar oppo-
sites, the outcome in both cases was almost exactly the same. In Chicago in 1983, in
what has been described as “one of the most racist campaigns outside of the Southern
states” (Alkalimat 1986: 7), Harold Washington™s white opponent, Bernard Epton, ran a
highly charged campaign that centered on the slogan, “Epton Now, Before It™s Too Late.”
In sharp contrast, in Philadelphia in the same year the general election had few explicit
racial references. Wilson Goode™s white opponent didn™t mention race. Throughout the
campaign, “Philadelphians congratulated themselves, while accepting the plaudits of the
national media, because their contest was free of the open anger that marked Chicago™s”
(Kleppner 1985: 250). Yet, despite the contrasting tones, the end result was almost iden-
tical. Exit polls indicated that white support for the two black challengers differed by
only 3 percentage points. The vast majority of white voters in both cities chose not to
vote for the black candidate.
The Transformation of the White Vote 57

conventional nonracial factors and the white vote.20 With the possible
exception of campaign spending, where there is some evidence that white
voting is related to the white opponent™s spending, the data suggest that
when a black candidate challenges to become the ¬rst black mayor of
a city, all that seems to matter is race and the fear that accompanies a
perceived black takeover.21
These black challenger elections stand out not only because they are
different from the typical urban contest but also because they are very dif-
ferent from previous elections in the same cities. A comparison of these
challenger elections with previous contests in the same cities indicates
that in many of these cities voting patterns that had held ¬rm for decades
were shattered when the possibility of an African American winning the
mayor™s of¬ce became real. Cities with apathetic electorates suddenly wit-
nessed record-high turnout levels. In Birmingham, turnout jumped from
48 percent to 68 percent when Richard Arrington ran for mayor. Simi-
larly, in Seattle, the proportion of registered voters going to the polls rose
from 41 percent to 60 percent when Norm Rice sought the mayoralty. In
Charlotte, when Harvey Gantt ran to become the city™s ¬rst black mayor,
turnout more than tripled from the previous election, growing from 14
percent to just over 50 percent. Amazingly, in roughly half of the cities in
the sample, including Birmingham, Chicago, Charlotte, Cleveland, Gary,
New Orleans, Newark, Memphis, and New Haven, turnout exceeded or
came very close to passing the previous highest level in the city.
Perhaps even more unusual is the way that party allegiances were
ignored when white voters faced the possibility of black leadership. In
many cases, vast numbers of white residents ended decades of support
for the Democratic Party to support a white Republican (Nelson and
Meranto 1977). In Chicago “ a city that had not elected a Republican for
¬fty-six years and had given white Republican nominees on average less
than 5 percent of the vote in previous decades “ white voters gave Bernard
Epton, Harold Washington™s white Republican opponent, 80 percent of
their votes (Lewis, Taylor, and Kleppner 1997). In Cleveland, 70 percent

20 As this implies, collinearity between the independent variables is not a problem here.
Although the number of cases is small, ordinary least-squares regression still represents
the best way to analyze the impact of each of these factors on the vote.
21 Neither spending by the black candidate nor the relative spending advantage of the white
opponent over the black challenger was signi¬cantly related to white vote choice but in
some speci¬cations, spending by the white opponent was negatively related to white
support for the black challenger. Since campaign spending data are available in only
two-thirds of the cases, spending was not included in the ¬nal model.
Changing White Attitudes

of all white Democrats switched over to the white Republican candidate
rather than support the black challenger (Cho 1974). The fact that 70
percent of all white voters in the twenty-six black challenger elections in
the data set opposed the black candidate and voted white, even though
most of the black challengers were Democrats running in overwhelm-
ingly Democratic cities, strongly suggests that party was less important
than race in these contests.22
Overall, the story that these aggregate electoral data tell is clear. As
predicted by the information model, when blacks challenge to take over
of¬ces they have never held, uncertainty and racial concerns end up dom-
inating the white vote while other more conventional, nonracial factors
play at best a secondary role. In almost every way “ from the central role
played by the size of the black community to the wholesale abandonment
of the Democratic Party by white Democratic voters “ the perceived threat
of a black challenger leads to an exceptional kind of election.

Individual Voices
The regression results in Table 2.2 are important in that they help substan-
tiate the role that race and racial fears played in the challenger contests.
They do not, however, allow us to get into the minds of white residents. If
we focus on the vote and the actions of white residents, we can only infer
what they are thinking when faced with the prospect of black electoral
victory. To know more about what whites are feeling and exactly what it
is they fear, we need to listen to the voices of individual white residents.
These voices echo my analysis of the aggregate white vote.
Although whites in the cities under consideration expressed an array
of feelings about the prospect of black mayoral leadership, and although
some did end up supporting black challengers, newspaper coverage sug-
gests that many white residents were intensely concerned about the pos-
sibility of a black takeover. In Chicago, a white voter expressed his fears

22 Although cities do not report voter registration data by race and party, one can very
roughly calculate the proportion of whites who are Democrats by assuming that nearly
all African Americans in a city are Democrats and that Latinos and Asian Americans (who
are generally a very small proportion of the population during these elections) are evenly
divided between Democrats and Republicans. When an estimate of the proportion of
whites in each city who are Democrats is added to the regression in Table 2.2, it is signi¬-
cantly related to white support for the black challenger, suggesting that white Democrats
are more likely to vote for black challengers. Thus, although there is little doubt that
white Democrats abandoned their party at high rates in these elections, white Democrats
were probably more likely than white Republicans to support black challengers.
The Transformation of the White Vote 59

about a black victory this way: “There will be turmoil in this city. What are
the blacks trying to do, win the whole United States from us? I™m scared”
(Peterson 1983b). The fears of many whites are, in retrospect, quite fan-
tastic. In Los Angeles, one resident viewed a black victory this way: “If
we have a colored mayor we™ll have colored people pushing us out of the
city. The whole city will be black if Bradley wins “ all those people will
be moving up from the South” (Reich 1973a). An elderly parishioner was
quoted in the Chicago Tribune as saying “if that black man gets elected,
no white woman would be safe on the streets” (Coleman 1983a). One
possibility “ neighborhood integration “ was foremost in many whites™
minds. Democratic Committee member Aloysius Majerczyk explained
that his constituents were “afraid of scattered-site housing” and “con-
cerned about the stability of [the city™s] neighborhoods” (quoted in Rivlin
1992: 185“6).
The importance of these racial concerns also came through in analyses
of exit poll data. In the few cities where whites were polled about both
their racial concerns and their preferences in the mayoral election, white
individuals™ support for the black challenger was closely tied to whether
or not they indicated that they had strong racial concerns (Kaufman 1998;
2004; Kinder and Sanders 1981; Pettigrew 1972; Jeffries and Ransford
1972; Sears and Kinder 1971). Moreover, these same studies found that
demographic factors, such as political ideology, partisan identi¬cation, or
socioeconomic status “ normally powerful predictors of individual vot-
ing preferences “ played unusually limited roles (Kaufman 1998; Halley,
Acock, and Greene 1976).
Black candidates themselves were all too well aware of the enormous
uncertainty surrounding their candidacies. In addition to running largely
deracialized campaigns “ almost half ran totally race-neutral campaigns,
and only two explicitly talked about favoring the black community “ many
openly discussed white fears after the campaigns. Richard Arrington, the
black mayor of Birmingham, was clear on this point: “The transition
creates uneasiness. I understand that” (quoted in Curry 1979). David
Dinkins™s slogan in New York in 1989, “Vote your hopes, not your fears,”
stressed the same point.23

23 Media accounts echoed the sentiments of black candidates. In New York, for example,
pollster Geoffrey Garin noted, “There is still a nervousness about any black candidate”
(quoted in Borger 1989). Even in some of the less racialized contests, like the one that took
place in St. Louis in 1993, reporters talked about white fears: “People are apprehensive
and uncomfortable about what might happen in this city. There are people who think
the city hall will collapse the next day” (Casmier et al. 1993).
Changing White Attitudes

table 2.3 The Transformation of the White Vote Between Black
Challenger and Black Incumbent Elections

White Support for the
Black Candidate
Candidate Quality
’0.13 (0.07)—
Quality of White Opponent
Quality— Black Incumbent Election ’0.23 (0.11)——
Democratic Party Endorsement 0.02 (0.05)
Party Endorsement— Black Incumbent Election 0.18 (0.08)——
Local Newspaper Endorsement 0.02 (0.04)
Newspaper Endorsement— Black Inc. Election 0.23 (0.12)—
’0.07 (0.00)———
Percent Black of City Population
Percent Black— Black Incumbent Election 0.00 (0.00)
’0.13 (0.06)——
Racialization of Black Candidate™s Campaign
Racialization— Black Incumbent Election 0.01 (0.09)
’0.30 (0.17)—
Black Incumbent Election
0.64 (0.11)———
Adj. R2 0.72
N 48
Note: OLS regression. Figures in parentheses are standard errors.
——— p < 0.01
—— p < 0.05
— p < 0.10

In short, by three different measures “ the aggregate white vote chief
among them, but also the individual sentiments expressed by white resi-
dents, and the concerns voiced by black candidates themselves “ there is
evidence that racial fears played a critical role in these black challenger

A Different Calculation in Black Incumbent Elections
But what happens the second time around? Is there, as the information
model predicts, a real transformation in the nature of the white vote in
black incumbent elections? A comparison of the aggregate vote in chal-
lenger and incumbent elections suggests that there is. Table 2.3 combines
the results of the same set of black challenger and black incumbent elec-
tions and includes a series of interactions to directly determine if different
The Transformation of the White Vote 61

factors matter more or less in the latter.24 In the table, each variable that is
not interacted with black incumbent elections measures the effect of that
variable in black challenger elections. Each interaction directly assesses
how much more or less that variable matters in black incumbent elections.
Thus, reading down the table, the signi¬cant interactions and the largely
insigni¬cant individual variables indicate that it is only after black incum-
bents have been given a chance to prove themselves that conventional fac-
tors begin to play an important role. As experience with black leadership
grows and fear about its consequences declines, “politics” begins to play
a primary role in voters™ choices.
More speci¬cally, Table 2.3 reveals that conventional factors such as
candidate quality and political endorsements matter much more in black
incumbent elections than they do in challenger elections.25 As evidenced
by the signi¬cant interaction between candidate quality and incumbent
elections, the weight that white voters put on the quality of the white
opponent grew sharply from the challenger to the incumbent elections.
White voters may not have cared who the white candidate was when he
or she faced a black challenger; at that point, any white would do. But
in black incumbent elections, white voters gave white candidates with
experience in citywide of¬ce almost 20 percent more votes than candi-
dates with no experience in political of¬ce. As one white politician put
it, “Race is not as much of a litmus test as it once was. The issue now
is who is the best quali¬ed man” (Sun Reporter 1993). The reduction in
white fears also appears to have increased white voters™ attention to news-
paper endorsements. These endorsements were essentially meaningless in
black challenger elections, but endorsement by the main local newspa-
per increased white support in the average incumbent election by another
20 percent. As well, party endorsements helped in the black incumbent
elections, even though they did not in the challenger elections. The local

24 This type of ordinary least-squares regression clearly cannot tell us about the causal direc-
tions of these relationships. Given that Jacobson (1980) and others have demonstrated
the endogeneity of campaign spending, challenger quality, and other resources in cam-
paigns, it is possible that the causal arrows could be reversed. For the present purposes,
however, all that matters is that conventional politics plays more of a primary role in
black incumbent elections than in challenger elections.
25 It is also worth noting that endorsements and candidate quality do not become more
important in black incumbent elections simply because there is more variation on each
measure in incumbent elections. As Appendix A shows, variance actually declines for
most of these measures in black incumbent elections.
Changing White Attitudes

Democratic Party™s endorsement delivered an additional 16 percent of the
white vote on average when blacks ran as incumbents.
A second important conclusion to draw from Table 2.3 is that race
still mattered in these elections. The fact that interactions with both of
the racial fear variables are insigni¬cant indicates that the size of the
black population and the racial focus of the black candidate™s campaign
remained important to white voters. This is not surprising; even a brief
review of these elections reveals that many of them were highly racialized.
Chicago, New York, and New Haven, in particular, represent cases where
the general trend toward increased white support and diminished racial
tension did not apply.
At the same time, there is evidence that race and racial fears were gen-
erally less powerful in the incumbent elections. Further analysis indicates
that racial fear lost half of its explanatory power: whereas the size of the
black population and the racial focus of the black candidate™s campaign
alone account for 67 percent of the variation in the vote in black chal-
lenger elections, these two variables account for only 36 percent of the
variation when white residents voted in black incumbent elections. As
one reporter put it in Chicago, “Something has changed. The paranoia
and ugly racism that ripped the city apart [four years ago] are largely
absent this time” (Bosc 1987). Another observer of several black incum-
bent elections in 1993 simply stated, “Race has faded in many places”


. 2
( 7)