New Deal Liberalism and Racial Liberalism in the Mass Public, 1937-1968
Eric Schickler
University of California, Berkeley
March 2012
Abstract
Few transformations have been as important in American political history as the
incorporation of African Americans into the Democratic Party coalition over the course
of the 1930s-60s and the embrace of racial conservatism on the part of many
Republicans. This paper, which is part of a broader book project, focuses on changes in
mass opinion among Democratic and Republican partisans from the late 1930s through
the 1960s. It traces, over time, the relationship between New Deal economic liberalism
and racial liberalism. A key finding is that by about 1940, economically-liberal northern
white Democratic voters were substantially more pro-civil rights than were economicallyconservative northern Republican voters. The relationships among Democratic
partisanship, economic liberalism, and racial liberalism held across a wide range of
demographic groups, but was particularly strong in urban areas. While partisanship and
civil rights views were unrelated among southern whites, southern economic
conservatives were noticeably more racially conservative than their economically liberal
counterparts. These findings suggest that there was a connection between attitudes
towards the economic programs of the New Deal and racial liberalism early on, well
before national party elites took distinct positions on civil rights. Along with grassroots
pressure from African American voters who increasingly voted Democratic in the 1930s40s, this change among white voters likely contributed to northern Democratic
politicians’ gradual embrace of civil rights liberalism and Republican politicians’ interest
in forging a coalition with conservative white southerners.
1
The incorporation of African Americans into the Democratic Party coalition over
the course of the 1930s-60s and the embrace of racial policy conservatism on the part of
many Republicans is widely – and appropriately – viewed as a critical transformation in
modern American party politics. One of the most dramatic manifestations of this change
was the shift among southern whites from the Democratic Party to the Republicans, but
the changes in the south were linked to a broader shift in the partisan landscape, in which
New Deal liberalism came to be identified with racial liberalism and Republican
conservatism became identified with greater opposition to governmental action to redress
racial inequalities.
For several decades, civil rights had been something of a deliberate “dead spot” in
American politics: an issue of central importance to the political system, yet one which
both national parties had little interest in addressing (Katznelson 2012). While
Republicans had pushed to enforce voting rights for African Americans prior to the
1890s, the party soon consolidated a majority that did not depend on African American
votes. Republican President William Howard Taft famously signaled an end to
Republican concern about the South’s treatment of African Americans when he called for
an end to “sectional” disputes and endorsed restrictions on suffrage for the “ignorant” in
his inaugural address. Meanwhile, southerners committed to maintaining the region’s
racial caste system dominated the Democratic Party, and their northern Democratic
counterparts showed no interest in transforming race relations. Yet by the 1960s, civil
rights had risen to the top of the political agenda and northern Democrats had replaced
Republicans as the partisan group most associated with civil rights liberalism.1
1
As the specific issues on the civil rights agenda changed over time, the meaning of “civil rights
liberalism,” “racial liberalism,” and “racial conservatism” also changed. Several of the policies under
2
Understanding the dynamics of this transformation is important in its own right,
while also serving as an excellent site to explore more general core questions about the
workings of the American political system. Under what conditions will political leaders
respond to demands for group rights and incorporation? How does a once-dormant issue
– which both national parties had little incentive to address – rise to the top of the
national agenda? Civil rights offers a test case for understanding the operations of
representative democracy in the United States: how does an implicit deal among national
elites to sidestep an issue unravel only to be replaced by a party alignment in which that
issue holds center stage?
A careful reexamination of the civil rights case points the way to an alternative
view of political change. Rather than a story of elite choice at a critical juncture or a
simple story of bottom-up grassroots pressure, I argue that locally-rooted politicians and
activists played a crucial role as intermediaries between mass-level pressures and elite
decision-making arenas. In this account, federalism and the decentralized system of
electing members to Congress provided key institutional mechanisms to facilitate the
gradual incorporation of civil rights into the mainstream of the Democratic Party,
undermining the implicit deal among national political leaders that had been a key
foundation of the party for decades. Much like abolitionism in the 1830s-40s and the
currency issue in the 1870s-90s, efforts by national party leaders to block a new issue
ultimately failed and party lines were reshuffled.
consideration in the 1960s were not part of the political conversation in the late 1930s or 1940s. For
present purposes, racial liberalism and conservatism are defined strictly in relative terms, reflecting where
actors stand on the policy issues under debate at the particular moment in time. This means that one can
assess changes in the relative positions of the parties and the degree of alignment between civil rights views
and economic policy views, but not the absolute degree of racial liberalism. Future work in this project will
consider the role of agenda change more explicitly.
3
The entry of northern African Americans, Congress of Industrial Organizations
(CIO) unions, Jews, and urban liberals into the Democratic coalition is a crucial starting
point for the analysis. These broad changes at the mass level in the 1930s predisposed
northern Democratic politicians to be more supportive of civil rights than their
Republican counterparts. Most rank-and-file northern white Democratic voters by no
means prioritized civil rights in this early period – and they shared in much of the racial
prejudice that was prevalent even in the north – but the early presence of voters who were
more predisposed towards civil rights policies in the Democratic coalition made northern
Democratic politicians the most likely group to respond as civil rights issues became
more prominent on the agenda.
Meanwhile, civil rights advocates – mainly African Americans, but also including
actors in the labor movement and urban liberals – pushed to make civil rights a more
salient issue. In doing so, they capitalized on external events – such as the rise of Nazism
in Europe, the disruptions of World War II, and the migration of African Americans to
northern cities – and built coalitions with other groups, such as CIO unions and Jews.
One can view these grassroots mobilization efforts as involving a separate timeline from
the mass-level partisan changes: few if any voters became Democrats in the 1930s
because of civil rights, nor is it likely that being a Democrat directly led voters to become
pro-civil rights. However, these two time lines intersected early on, as the sorts of rankand-file voters most likely to be responsive to the grassroots mobilization were those
already part of the New Deal economic coalition.
The nationally-oriented party leaders who had the greatest stake in maintaining
the Democrats’ north-south coalition were generally – though not always – slow to
4
respond to this linkage of civil rights to New Deal liberalism. Instead, locally-rooted
politicians and political organizations were among the first to appeal to the nascent
connection: state Democratic parties outside the south moved to embrace civil rights
policies in the 1940s, just as nonsouthern Democratic members of Congress surpassed
Republicans as the main advocates of civil rights. The crucial point is that the
independent power base of state parties and the election of House members through
separate geographic districts created the space for locally-based politicians to respond to
activist pressure for civil rights, without requiring an immediate showdown with national
party leaders. Civil rights forces gained an institutional foothold through alliances with
state Democratic parties and northern House Democrats. These locally-rooted politicians
then contributed to civil rights activists’ efforts to raise the salience of the issue (see
Sugrue 2008: 111-15). For example, by signing discharge petitions to force civil rights
bills onto the House floor, northern Democrats helped raise civil rights’ visibility on the
national agenda. The implicit “deal” to keep civil rights off the table broke down due to
this intertwining of ground-up and meso-level forces.2 Congress and state parties emerge
from this case as potential vehicles for new interests to gain access; localism and
geographic-based districts are often seen as bastions of conservatism, but in this case they
also provided a foothold for civil rights liberals.3
Changes in the Republican Party illustrate similar dynamics, though with a very
different set of interests gaining the upper hand. Republicans’ association with civil
2
There are potential parallels here to the rise of populism and the currency issue in the 1870s-90s.
National Democratic elites had little interest in moving away from the gold standard in this period, since
“sound money” was crucial to the party’s strategy of winning New York’s pivotal electoral votes. But
those calling for currency expansion gained control of state parties and won the allegiance of Democratic
members of Congress, eventually taking the party over from Grover Cleveland and the gold wing.
3
Of course, localism and geographic-based districts empowered racial conservatives in the South. But the
ways in which these features of the political system allow new interests to gain a foothold has been noted
less often.
5
rights had largely faded by the early 20th century. Even as some party members pushed
the Dyer Anti-Lynching bill through the House in 1922, the measure was a low priority
for most Republicans and little was done to defeat the ensuing Senate filibuster
(Zangrando 1980). Herbert Hoover’s subsequent efforts to promote “lily-white”
Republican party organizations in the South was an early indication that at least some
party leaders saw the potential offered by an alliance with white southerners premised on
dropping the vestiges of the party’s ties to African Americans (see Topping 2008).
This idea gained more traction in the late 1930s, as southern Democrats’
disaffection with the New Deal generated greater cooperation between northern
Republicans and southern Democrats and talk of a partisan realignment along liberalconservative lines (see, e.g., White 1936; Patterson 1965). But many GOP leaders were
wary of an alliance with southern white conservatives. Thus, in December 1937, Senate
Republican leader Charles McNary sabotaged efforts by Josiah Bailey (D-NC) and
Arthur Vandenberg (R-MI) to put together a Conservative Manifesto by prematurely
revealing the plans to the press (Patterson 1965: 607). Prominent national leaders who
shared McNary’s moderate brand of Republicanism – such as Wendell Willkie and
Thomas Dewey – sought to position their party as relatively liberal on civil rights. More
generally, Republican leaders in urban states in the northeast and Midwest believed that a
moderate or liberal position on civil rights would help in statewide elections and better
position the party in terms of the national electoral map (Topping 2008). To these party
leaders, an explicit alliance with southern whites that involved embracing racial
conservatism would jeopardize the GOP’s ability to appeal to northern moderates and to
keep at least a foothold among African American voters. Thus, the Republican Party’s
6
national platform continued to generally be as liberal as – or more liberal than – the
national Democratic platform for most of the 1930s-50s (see Carmines and Stimson
1989).
But important changes were occurring within the GOP that would ultimately
transform the Party. At the mass level, as shown below, rank-and-file northern
Republican voters had become more racially-conservative than rank-and-file northern
Democrats by the late 1930s, and economically conservative Republicans were especially
likely to be racially conservative. Within the party, many politicians came to recognize
that African American voters were unlikely to come back to the GOP; Willkie and
Dewey – notwithstanding their reputations racial liberalism – lost the black vote by wide
margins in the 1940 and 1944 presidential elections.4 Meanwhile, Republicans who
sought to push the GOP in a more conservative direction overall – such as Senator Karl
Mundt of South Dakota – saw the potential for a realignment premised on a states-rights
platform.5 Indeed, Mundt made a series of high profile speeches across the south in
1950-51 arguing on behalf a realignment.6 Mundt had the backing of Republican
National Committee chair and Robert Taft ally Guy Gabrielson, who also announced a
GOP southern drive to capitalize upon “disaffections over the states’ rights issue” (New
4
Based on Gallup data, Ladd estimates that nearly 70% of African Americans voted for Roosevelt in 1940
and 1944 (Ladd 1975: 158).
5
Republican economic conservatives and southern Democrats began to join forces in Congress in the late
1930s and early 1940s, but this “conservative coalition” initially was most evident in the area of labor
policy, where the two groups cooperated to pursue an aggressive agenda of scaling back the pro-union
Wagner Act regime (see Katznelson et al. 1993; Farhang and Katznelson 2005; Brinkley 1995; Schickler
and Pearson 2009). The relationship between congressional Republicans and southern Democrats on civil
rights policy in that period is more difficult to tease out, but at a minimum rank-and-file Republicans
exhibited less inclination to support efforts to force civil rights initiatives to the House floor by the mid1940s (see Schickler, Pearson, and Feinstein 2010).
6
For example, in a 1950 speech in North Carolina, Mundt,called for the nomination of a Republican
presidential candidate acceptable to the south and the omission from the Republican platform of planks
“repugnant to the South and an insult to your traditions” (Heidepriem 1988: 159).
7
York Times, October 21, 1951, p. 54). Mundt’s effort was endorsed by a handful of other
conservative Republican senators, such as Joe McCarthy (R-WI) and Owen Brewster of
Maine (New York Times, August 1, 1951, p. 47) and garnered considerable coverage in
southern newspapers (Heidepriem 1988).7
Eisenhower’s victory in 1952 stalled the talk of such a realignment by bolstering
the position of the moderate faction within the national GOP. But Eisenhower sought to
build upon his personal popularity in the South by launching a serious party-building
effort in the region (see Bowen 2011; Brennan 1996). This effort paid off in the shortterm with the development of genuine Republican organizations in several states for the
first time in generations. However, these same state organizations were captured in the
mid-to-late 1950s by grassroots conservative activists who rejected Eisenhower’s
moderation.8 Indeed, the new state parties provided a vehicle for followers of Barry
Goldwater to capture the GOP. As early as 1956, southern Republican national
convention delegates were nearly as racially conservative as their southern Democratic
counterparts.9 Four years later, South Carolina and Louisiana’s state Republican
conventions pledged their delegates to Barry Goldwater and a movement to draft the
Arizona Senator for Vice President garnered much of its support in the south (Brennan
7
A high point of Mundt’s drive was a nationally broadcast radio debate with Hubert Humphrey (D-MN) in
which the liberal Democrat and conservative Republican found common ground in their agreement that
conservative southern Democrats and conservative northern Republicans belong in the same party (see
Heidepriem 1988: 162.)
8
Eisenhower’s forces allied with hard core segregationists at times as well (e.g. in Mississippi; see Bowen
2011), but for the most part his strategy was not based on appeals to the fiercest civil rights foes.
9
A reanalysis of Herbert McClosky’s survey of 1956 national convention delegates shows that 57% of
southern white Republican delegates favored decreased efforts on behalf of desegregation, as compared to
61% of southern white Democrats. Just 13% of the southern Republicans favored increased efforts (as
compared to 16% of the Democrats). Consistent with the decline in biracial Republican state parties in the
south, a mere 12 of the 312 Republican southern delegates surveyed were African American.
8
1996; Pearlstein 2001).10 The revived southern state Republican parties thus provided an
institutional foothold for the Goldwater movement, broadening its base beyond the west
and paving the way for Goldwater’s 1964 triumph over the moderates. Thus, even as
national leaders in both parties continued to look with some trepidation at the
increasingly intense debate on civil rights, rank-and-file members of Congress and
several state Republican parties responded to activist pressure to take a clearer stand on
the issue.11
The civil rights realignment thus exemplifies how political transformations can
emerge from the intersection of multiple institutional trajectories (see Orren and
Skowronek 2004). Along the first trajectory, the party system was reshaped in the 1930s40s without regard for civil rights politics, as the Democrats embraced New Deal
liberalism and new coalition partners in response to the Depression, and Republicans
countered with a turn to antistatism. Meanwhile, on an initially separate trajectory,
grassroots activists and groups gradually pushed the civil rights issue onto the national
agenda. Social change – such as the migration of African Americans to the north – played
a key role in this long-term process (see McAdam 1999). Many of these activists and
groups had ties to the New Deal Democratic Party due to the economic policies and
ideological doctrines embraced by the party in the 1930s, but these linkages developed
10
Newsweek claimed that in a truly open 1960 convention, Goldwater would have been the Vice
Presidential nominee (Pearlstein 2001: 83). Pearlstein concludes that by the end of the 1960 convention,
Goldwater’s ranks had swelled to include most of the southern delegations.
11
Outside the south, state party dynamics may have also played an important role in sidelining liberal
voices within the GOP on civil rights. For example, Charles LaFollette of Indiana was a lonely ardent procivil rights voice within the House Republican conference in 1945-46 (see GOP conference meeting
minutes from September 1945). The Indiana state GOP convention overwhelmingly chose William Jenner
over La Follette when the two faced off for the party’s Senate nomination in 1946; Jenner made little secret
of his disdain for civil rights (see, e.g., Chicago Defender, July 27, 1946, p. 4). The liberal La Follette was
out of touch with his party on numerous issues, including labor policy, but he put a premium on the civil
rights issue.
9
for reasons largely independent of the push for civil rights.12 The key is how these two
timelines intersected: when civil rights activists succeeded in pushing the issue onto the
national agenda—despite the resistance of many national leaders in both parties but with
considerable support from rank-and-file Democrats and meso-level party actors—it was
the Democrats who were disposed to embrace the issue because of the changes along the
first timeline. By remaking the Democratic Party outside of the South to be the
representative of CIO unions, African Americans, Jews, and liberal egalitarianism, the
New Deal set the stage for the later realignment on the race issue—though the latter
could not occur until actors on the second timeline forced the issue to the decision stage.
The civil rights realignment was thus shaped by the braiding together of two distinct
political trajectories over time. It was a gradual process that started in the mid-1930s,
gathered momentum in 1940s as the war mobilization created a window of opportunity
for civil rights activists to force fair employment laws onto the agenda, and continued
into the 1950s and 1960s.
This paper takes up one element of this broader argument, focusing on the
dynamics of mass opinion concerning civil rights in the late 1930s through the late1960s. An important question is whether the shifts in Congress and in state parties noted
above corresponded to changes in mass opinion on civil rights issues. At the mass level,
did New Deal liberalism come to incorporate racial liberalism in the 1930s-40s or did
economic liberalism only become joined with racial liberalism much later, after the rise
of the civil rights movement and the crystallization of elite positions? Were Republican
12
It is also doubtful that the correlation between economic conservatism and racial conservatism observed
in the north in the late 1930s and 1940s is due to racial prejudice causing economic conservatism. Civil
rights were simply not prominent enough on the agenda at the time to be a likely source of changing
economic policy views or partisanship for northern whites.
10
elites ignoring – or responding to – their constituents as they began edging away from
racial liberalism in the 1940s and 1950s?13 The evidence presented below suggests that
Republicans’ willingness to cooperate with southerners in blocking civil rights measures
in the mid-1940s coincided with a shift at the mass level in which nonsouthern
Republican partisans displayed markedly less support for civil rights policies than did
nonsouthern Democrats.14
To assess the development of public opinion towards civil rights, this paper draws
upon the earliest available mass survey data. Starting in the mid-1930s, polling
companies surveyed the public about important issues on a monthly basis (see Converse
1987; Igo 2007). Questions on civil rights are spottier than one would like in these early
polls, but as discussed below, there are useful survey items assessing civil rights policy
attitudes that go back to 1937.
While opinion polls conducted in the 1930s and 1940s have numerous problems
that have limited their use by scholars, the National Science Foundation has funded an
extensive effort to make the data suitable for analysis (Berinsky and Schickler 2011). We
recoded the datasets, which has involved ferreting out and correcting many errors and
inconsistencies. We have also put together a series of post-stratification weights that
partially address the problems introduced by the quota-controlled sampling techniques
used in the 1930s-40s (see Berinsky 2006; Berinsky and Schickler 2006; Berinsky,
Powell, Schickler, and Yohai 2011). The recodes and weights are being made publicly
13
One indicator of this edging away was the increased reluctance of Republicans to sign discharge petitions
to force civil rights bills to the House floor (see Schickler, Pearson, and Feinstein 2010).
14
As noted below, I do not argue that nonsouthern Democrats were, on the whole, intense in their support
for civil rights in the late 1930s and 1940s. But they were to the left of their GOP counterparts, with
economically-liberal Democrats particularly likely to support civil rights policies. The rise of the grassroots
civil rights movement was a necessary precondition for this nascent Democratic liberalism on civil rights to
become a potent political force – but the early mass alignments make it clear that the Democratic Party was
the more likely “home” once civil rights activists forced the issue to the top of the agenda.
11
available through the Roper Center. This paper makes use of the early survey data, along
with civil rights-related questions on surveys through 1968.15
Public Opinion and Civil Rights Liberalism
The conventional story regarding the parties’ handling of civil rights has been that
nonsouthern Democratic and Republican elites occupied similar positions on the issue until
the 1960s, and that partisans at the mass level only diverged after elites – such as Lyndon
Johnson and Barry Goldwater – sent clear signals of their respective party’s new position
(Carmines and Stimson 1989). From this perspective, the connection between racial
liberalism and New Deal liberalism is quite limited: the ideological logic and coalition
alignments associated with New Deal liberalism are not inherently associated with civil
rights liberalism. Instead, the connection was largely an elite construction, years after the
initial rise of New Deal liberalism.
However, as noted above, considerable empirical evidence has emerged that
nonsouthern office-holders and state parties began to polarize on civil rights in the mid1940s, with Democrats adopting the more liberal position (see Chen 2009; Karol 2009;
Feinstein and Schickler 2008). Figure 1 – drawn from collaborative work with Brian
Feinstein and Kathryn Pearson – shows the pattern with respect to signing discharge
petitions targeting civil rights bills in the House of Representatives (see Schickler, Pearson,
and Feinstein 2010). Discharge petitions were the primary mechanism through which
members sought to force floor action on civil rights bills in this era, given the opposition of
the powerful House Rules Committee. The figure is derived from a logit model predicting
discharge signatures. The identical model is estimated separately for each civil rights
discharge petition. The model controls for seniority, holding a committee leadership
15
While I have attempted to gather all of the civil rights related policy questions spanning 1937-68 for
which individual-level survey data is available, there are a handful of surveys – particularly in the mid-tolate 1960s – that I have not yet coded and incorporated.
12
position, and membership on the committee targeted by the petition. The figure shows the
first difference and 95% confidence interval when one shifts from a northern Republican to
a northern Democrat, holding the remaining covariates at their mean or mode. While
nonsouthern Republicans had been more likely to sign discharge petitions for civil rights
bills than nonsouthern Democrats in the 1930s, by the mid-1940s the roles had been
reversed and nonsouthern Democrats were substantially more likely to sign (Schickler,
Pearson, and Feinstein 2010). The first difference estimates indicate that a typical northern
Democrat in the mid-to-late 1940s was about 30 to 40% more likely to sign a civil rights
discharge petition than a typical northern Republican. Evidence from party platforms also
indicates that northern Democratic state parties became more liberal than their GOP
counterparts in the mid-1940s (Feinstein and Schickler 2008).
Interestingly, the early public opinion polls indicate that nonsouthern white
Democratic voters moved to the left of their Republican counterparts on the civil rights
items available in the surveys in the late 1930s – before the elite movement became
evident. In addition, support for economic liberalism was tied to racial liberalism in the
nonsouth – and even, though to a lesser extent, in the south – by the late 1930s. These
results suggest that once the civil rights movement mobilized sufficiently to force civil
rights to the top of the agenda, nonsouthern Democratic voters were by far the most likely
to be responsive. In turn, this meant that nonsouthern Democratic elites would have a
stronger incentive to take the lead than their GOP counterparts. This incentive would be
reinforced by the presence of African Americans and pro-civil rights unions (i.e. the CIO)
as core members of the Democratic coalition (Chen 2009; Karol 2009; Feinstein and
Schickler 2008). Although nonsouthern Republicans remained much more liberal on civil
rights than southern Democrats (and the few southern Republicans), the connection
between economic conservatism and racial conservatism that emerged at the mass level
was a harbinger of the eventual “southern strategy” adopted by the Republicans.
13
The focus in this paper is on white respondents in order to separate out the impact
of the changing racial composition of the Democratic Party in the 1930s-60s. That is,
while northern African Americans clearly played a crucial role in changing the party’s
stance towards civil rights (see Jenkins et al. 2010), it is important to determine whether
there is also evidence of change among white Democrats and Republicans. The
implications for our understanding of New Deal liberalism would be different if all of the
Democratic mass-level change was attributable to the influx of African American voters. I
am planning a more detailed study of changes in African American partisanship during
this period, but doing so is difficult due to the uneven representation of African Americans
across the polls (see Schickler and Caughey 2011 on African American economic policy
views, as measured in these early polls).
Several patterns emerge from the results. First, among nonsouthern whites, there is a
clear tie between Democratic partisanship and economic liberalism on the one hand, and
support for the major civil rights initiatives on the agenda in the late 1930s and 1940s: antilynching legislation, a ban on the poll tax, and fair employment practices legislation. This
connection predates Harry Truman’s very public embrace of civil rights liberalism in the 1948
campaign. Second, although Democratic partisanship is generally unrelated to civil rights
views among southern whites, economic liberalism does appear to be related to less
conservative views on lynching, the poll tax, and fair employment practices. Third, the tie
between civil rights liberalism and Democratic partisanship in the north is less clear cut when it
comes to racial prejudice and social segregation than when it comes to lynching, the poll tax,
fair employment, and to the more general idea of government action to counter discrimination
against African Americans and other minorities. Interestingly, economic liberalism is more
consistently related to support for school and housing desegregation than is partisanship.
Northern Democrats’ views in the 1930s-50s – more supportive than Republicans when it
comes to many civil rights policies, but not so much when it comes to policies that encourage
14
more intimate social mixing – presage the ambivalence that northern Democrats would exhibit
towards busing and related measures in the 1970s and beyond.
The discussion is organized as follows. I first trace the response patterns on questions
relating to three leading civil rights related issues on the agenda before the end of World War
II: anti-lynching legislation, the poll tax, and military integration. I then turn to fair
employment legislation, which emerged as a major issue during World War II and was
arguably the top legislative priority of civil rights groups into the 1960s. The next section
discusses some of the limitations in white Democrats’ civil rights liberalism, particularly when
it comes to social segregation and prejudice. Finally, I consider the sources of the linkages
between New Deal economic liberalism and racial liberalism and the relationship between
mass opinion and elite-level decision-making.
Questions on Lynching
Unfortunately, the first poll with a civil rights attitude item was not conducted until
January 1937, as Congress considered an anti-lynching bill. Gallup asked about this legislation
on six surveys conducted from 1937-40; an additional six Gallup surveys in 1947-50 asked
more generally about whether the federal government should have the right to “step in and deal
with the crime” when a lynching takes place, or whether this should “be left entirely to the
state and local governments.” The earliest surveys – conducted from January through
November 1937 -- showed roughly 60% of Americans in favor of the lynching bill and just 2325% opposed. However, when the question wording was changed in December 1937 and
January 1940 to include specific information about the punishment for counties that
countenance lynchings, the respondents leaned only slightly in favor of the legislation (44%40% in 1937 and 49%-42% in 1940). The 1947-50 items on the federal government’s role also
generally revealed a closely divided national public (see Appendix Table 1 for question
15
wording and marginal totals for each question, as well as responses broken down by party and
region).16
The surveys conducted in January to November 1937 show little or no relationship
between Democratic vote choice and support for the lynching bill among nonsouthern whites.
Starting in December 1937, however, there is a clear relationship: nonsouthern white
Democratic voters are about 10 points more supportive than are their Republican counterparts.
For example, in the December 1937 Gallup survey, Democrats back the lynching bill by a
50%-33% margin, while Republican voters are split, 43%-43%. Three years later, Democratic
voters favor the bill 52%-39% while just 42% of Republicans favor it compared to 50%
opposed.17
In order to assess the relationship between support for action against lynching and
partisanship across the items, Figure 2a presents the results of separate bivariate OLS models
run for each survey; the dependent variable is a dummy variable for support for federal action
against lynching and the independent variable is a dummy variable for Democratic presidential
vote choice in the last presidential election.18 The sample in each case is nonsouthern whites
who voted either Democratic or Republican in the preceding presidential election. The results
again suggest that the party gap opened up in December 1937 and remained about the same
over the next thirteen years. The point estimates indicate that a change from Republican to
Democratic vote choice is associated with a .15 shift in support for action against lynching on
the 0 to 1 scale. While Harry Truman was the first Democratic President to embrace racial
liberalism explicitly – particularly in his 1948 campaign – the inter-party gap appears about the
same before and after this transition. When demographic controls are added to each model, the
16
The question wording for these items varied slightly across surveys. The support for federal intervention
is higher in the June 1947 survey than in the other polls. This is likely because the question was placed
after a handful of items on the recent (and notorious) Greenville, South Carolina lynching case.
17
I use presidential vote choice because party identification is not included in these early surveys.
18
The results look substantively the same if one instead estimates the models using logit. Since the level of
support for the various lynching items generally ranges from about 30%-70% (rather than being near the
extremes), I focus the presentation on the easier-to-interpret OLS coefficients. [NOTE: I will redo using
logit model and will present first differences]
16
results remain robust.19 As discussed in more detail below, the consistency of the results when
demographics are included suggests that changes in the demographic composition of the
parties do not account for the observed relationship between Democratic voting and civil rights
views.
Figures 2a and 2b presents the results when economic liberalism is used to predict
views on lynching among northern and southern whites respectively. In coding economic
liberalism, I identified all questions that asked about the government’s role in the economy
(e.g. business regulation, government ownership of industry, government spending on relief
and other social programs, labor policy).20 To facilitate comparisons with the partisanship
measure used above, the economic liberalism measure also ranges from 0 to 1, with
conservatives scored 0, moderates scored .5, and liberals scored 1. Respondents are coded as
economically-liberal if they provide the liberal response more often than the conservative
response to the set of economic policy items in the survey.21 The results for economic
liberalism need to be interpreted with some caution, as the specific items used to construct the
economic liberalism measure change across the surveys; therefore, one cannot assume that the
measure taps the exact same concept in the same way across each data point. However, close
inspection of the measures suggests that there is little reason to believe that there is an overall
trend in their quality; as a result, the consistency of the results over time is likely a meaningful
indicator of a significant and relatively stable relationship.
Indeed, there is a consistent positive association between economic liberalism and
support for action against lynching from late 1937 as well. This holds in both the north and
19
Controls for demographics (which are used in a consistent manner across the surveys discussed below)
include: age, gender, occupation / class (professional, labor, poor), region (Northeast, Midwest, West),
urban and farm residence, phone ownership / car ownership, and, when available, union membership and
education. Note that statistical significance tests are not necessarily appropriate in this context, due to the
sampling issues in the surveys. See Berinsky (2006) for a discussion of this issue.
20
Questions about farm and defense policy were not included because it was not clear what the
economically “liberal” position was on these items, particularly in the context of the 1930s-50s.
21
I will attempt to develop more precise measures of economic liberalism in future iterations of the paper.
17
the south (see Figures 2b and 2c). 22 While the magnitude of the point estimates varies
somewhat across surveys, a shift from economic conservatism to liberalism is typically
associated with a .1 to .2 shift in lynching views (on the 0-1 scale) in both the north and the
south. The southern results are particularly striking because there is no relationship between
vote choice and support for lynching legislation in the essentially one-party region. It is also
important to emphasize that economic items that seem unlikely to have even a remote
inherent connection to racial attitudes – e.g. government ownership of the railroads – are as
closely tied to views on lynching as are items on issues that at least potentially have racial
implications (e.g. government help for those without money).23 The relationship between
economic views and support for action against lynching also holds up when a series of
demographic controls are included in the model, including accounting for differences within
the south (e.g. rim south vs. deep south; level of urbanization).
One question these results naturally raise is whether economic liberalism or
partisanship is more relevant to racial attitudes in the nonsouth. The evidence suggests both
matter. If one replicates the model in Figure 2a, controlling for both economic liberalism
and Democratic presidential vote, the estimates for both variables are positive and
statistically significant in nearly all cases from December 1937 on.24
An alternative approach is to classify individuals based on both their partisan vote
choice and ideology. This allows a closer examination of potential interactions between
these traits. For example, Sniderman and Stiglitz (2012) highlight the concept of “sorted
22
The absence of a relationship in the south in the last poll in the series (see Figure 2c) could well be due to
problems with the economic liberalism item in that survey. The only available economic liberalism
measure asked about the appropriate government response to the budget deficit. Respondents were coded
as liberal if they oppose domestic spending cuts and as conservative if they favored spending cuts and no
tax hikes. 72% of southerners were coded as conservatives, with just 13% moderate and 15% liberal. Most
surveys had questions that more clearly tapped into economic liberalism and that had less skewed response
distributions.
23
For example, in the December 1937 survey, the government ownership of the railroads item is actually
more closely related to views on the lynching bill than is the question about government help for the needy.
24
Party identification is not available in the earliest polls, but is in the post-1947 surveys. For those
surveys with party ID, it is also related to views of the lynching bill among nonsouthern whites, though the
relationship tends to be a bit smaller than it is for presidential vote choice.
18
partisans” – individuals for whom ideology and partisanship are in alignment – arguing that
such individuals will behave in distinctive ways as compared to partisans with views that are
out of synch with their party (see Levendusky 2009). Indeed, the gap between economically
liberal Democratic voters and economically conservative Republicans is especially large.
For example, in the December 1937 Gallup survey economically-liberal FDR voters favored
the lynching bill by a lopsided 60%-28% margin, while economically-conservative Landon
voters opposed it, 53%-33%.
To assess these interactions more systematically, I created a series of dummy
variables indicating economically-liberal Democratic voters, moderate Democratic voters,
conservative Democratic voters, economically-liberal Republican voters, moderate
Republican voters, and conservative Republican voters. The first and last categories are
“sorted partisans”: individuals whose economic views and partisanship are in alignment.
Figure 3 presents results from a simple regression model in which support for federal action
against lynching is modeled as a function of these dummy variables (the excluded category
consists of economically-conservative Republican voters), along with a dummy variable for
each survey.25 The analysis is limited to nonsouthern whites and begins with the December
1937 survey in which the gap between the parties first becomes substantial. The results
suggest that economically-liberal Democratic voters are nearly 22 points more likely to
support action against lynching than are economically-conservative Republican voters.26
25
Pooling across surveys creates potential problems, particularly if the effect of the independent variable of
interest is changing over time. I plan to explore more elaborate models that allow for time trends in
estimates in future iterations of this paper. Figure 2 does give at least some reason to believe that pooling
the post-December 1937 polls is a reasonable move in this case. Inclusion of a separate dummy variable
for each survey accounts for overall changes in support and for differences in question wording.
26
In interpreting the substantive magnitude of the estimates, it is useful to note how they compare with the
estimates for demographic predictors of views on civil rights. If one estimates the same baseline pooled
model (with a separate intercept for each poll), education has only a modestly sized impact (attending at
least some college is associated with a 4 point increase in support for anti-lynching legislation (point
estimate = .04; SE=.01), as compared to a grade school education; completion of high school is also
associated with a .04 increase (SE=.01), again compared to the baseline of a grade school education. If one
instead uses urban residence as a predictor, it is associated with a more substantial .10 increase (.01) in
support, which is still a bit less than the estimate for Democratic presidential vote choice in an analogous
model.
19
Notice, however, that economically-liberal Republicans look very similar to economicallyconservative Democrats. Indeed, within each partisan category, one observes a clear tie
between economic ideology and support for federal action. At the same time, for each
ideology level, Democrats are more likely to support federal action than are Republicans.
When one replicates the analysis adding controls for a full range of demographic variables,
the partisan and ideological gaps remain robust.27
The bottom line from these analyses is that support for federal action against
lynching was higher among Democratic voters and economic liberals in the north starting in
late 1937. As of 1937, however, northern Democrats in the House of Representatives
appear to have been less clear in their support for lynching legislation than were their
northern Republican counterparts. For example, they were less likely to sign the discharge
petition promoting the legislation and they were more likely to support a watered-down
alternative to the bill favored by the NAACP.28 However, by the mid-1940s, northern
Democrats in the House had surpassed their Republican counterparts in their willingness to
support action against lynching. While mass-based pressures may not explain this shift, the
opinion data suggest that northern members of the House came to act in ways that were
consistent with their constituents’ relative levels of support for federal action against
lynching.
The Poll Tax
Gallup asked a national sample of respondents about banning the poll tax on six
occasions from December 1940 through February 1953. Respondents in several southern
27
The multivariate models indicate that females, professionals, laborers, Northeasters, young people, and
urban residents are each more likely to back action against lynching. For polls with education, more
educated respondents are more supportive of action against lynching.
28
See “House Prepares for Passage of Anti-Lynching Bill,” Washington Post, April 8, 1937; and “House
Sidetracks Anti-Lynching Bill,” New York Times, April 8, 1937. In the Senate, however, Republicans
proved somewhat less willing than Northern Democrats to back cloture—though in other cases in this
period, Senate Republicans provided greater support than their Democratic counterparts (see Jenkins et al
2010).
20
states were asked about eliminating the poll tax “in this state” on two additional occasions
in 1941. A substantial majority of the national population favored banning the poll tax
throughout this period: the smallest margin was 63%-26% in 1940; the most lopsided was
72%-21% in 1953.
This item is trickier to interpret than the lynching questions, since the poll tax
undermined poor whites’ voting rights in the south, as well as African Americans.
Nonetheless, the results are consistent with the anti-lynching findings. From the start in
1940, nonsouthern white Democratic voters are more supportive of banning the poll tax
than are nonsouthern white Republican voters.29
A simple OLS model predicting support for the poll tax ban among non-southern
whites, estimated separately for each survey, suggests that Democratic vote choice is
associated with support for the ban across each survey from 1940 to 1953, with little
variation in the size of the association (see Figure 4a). The point estimates are generally a
bit smaller than in the lynching case; one interpretation of this difference is that the
lopsided national margin in favor of banning the poll tax generates an extreme cutpoint that
does less to separate weak from strong civil rights supporters (i.e. even weak civil rights
supporters back a poll tax ban). Indeed, even southern white respondents lean in favor of
the poll tax ban, though less decisively than their northern counterparts. The results for
economic liberalism are somewhat more variable – perhaps due to the varying quality of
the liberalism measures – but nonetheless suggest that there was a significant association
between economic liberalism and support for an end to the poll tax among whites in both
the north and the south (see Figures 4b and 4c).
29
One interesting wrinkle is that the December 1940 survey suggests that 1936 vote choice is less strongly
related to supporting the poll tax ban than is 1940 vote choice. Indeed, a full 74% of new FDR voters in
1940 (after having either not voted or voted for another candidate in 1936) backed the poll tax ban, while
just 60% of new Willkie voters backed the ban. This suggests that the relationship between support for
civil rights and vote choice had tightened over the course of 1936-40. In a multivariate model with controls
for demographics, if one includes both 1936 vote and a variable for change in vote from 1936-40 (scored
one for new FDR voters; 0 for new Willkie / GOP voters; and .5 for everyone else), the change vote
variable is positive and generally statistically significant, while the 1936 vote item is substantively and
statistically insignificant.
21
As with the lynching items, the results suggest that partisanship and economic
views continue to be associated with civil rights views when both independent variables
included in the same model. When one pools the data for nonsouthern whites, and
estimates a single model predicting support for the poll tax ban – with a separate intercept
for each survey – both economic views and partisanship have a significant impact, though
the size of association is smaller than for the lynching items.30 These results again hold up
when controls are added for a full range of demographic variables.
When one examines potential interactions between partisanship and economic
views, it once again appears that “sorted partisans” are particularly far apart. Economically
liberal Democrats are a full 12 points more in favor of the poll tax ban than are conservative
Republicans (see Appendix Figure 1). By contrast, moderate Democrats are just over 7 points
more favorable than conservative Republicans, and economically conservative Democrats and
liberal Republicans are both about 4 points more favorable than are economically conservative
Republicans.31
Integrating the Military
While lynching and the poll tax were the main racial policy items in the early surveys –
coinciding with their top spots on the sparse congressional civil rights agenda in the late 1930s
and early 1940s – there were a handful of additional questions. A June 1942 Gallup survey
30
The point estimate for Democratic presidential vote choice is .063 (with a standard error of .010) when it
is the sole independent variable (other than the intercept for each survey) and falls slightly to .050
(SE=.011) when economic liberalism is added to the model. The point estimate for economic liberalism is
.062 (SE=.014) in the latter model.
31
Once again, the results are very similar if one adds demographic controls. The estimated effects for
partisanship and ideology are by no means huge, but compare reasonably favorably to the associations
observed for demographic predictors of civil rights views. When one estimates the same baseline pooled
model using alternative demographic variables as predictors instead of partisanship and ideology, one finds
that urban residence is associated with a .11 increase (SE=.01) in support for the poll tax ban. By contrast,
education has only a modest impact in an analogous model (again, only including an intercept for each
poll): a college education is associated with a .03 increase in support for the poll tax ban (SE=.01) as
compared to a mere grade school education.
22
asked respondents about integrating the U.S. military.32 Nonsouthern white FDR voters were
just six to seven points more supportive than nonsouthern white Willkie voters (Willkie voters
opposed integration by a 53%-38% margin; FDR voters opposed it by 47%-45%).33 But when
one isolates economic liberals who voted for FDR and compares them to economic
conservatives who voted for Willkie, there is a much bigger gap, with economically-liberal
Democrats backing military integration by a 56%-40% margin and conservative Republicans
opposing it 58%-35%. Thus, six years before Truman’s executive order desegregating the
military, nonsouthern white Democratic economic liberals backed integration at the mass level,
while their economically conservative Republican counterparts opposed it. It is worth
emphasizing that the integration question – like the earlier poll tax and lynching questions –
made no mention of the position of any Democratic or Republican elites. When Gallup asked
again about integrating the military in May 1948 – amidst Truman’s civil rights initiatives – the
gap between Democratic and Republican voters is about ten points – so a bit larger than in
1942, but not by much (see Appendix Table 2). As in 1942, economic liberals are more
supportive of integration than are economic conservatives, but the later survey had only limited
economic policy questions.34
Fair Employment Policy
The single civil rights issue that became most prominent in the mid-1940s was fair
employment practices (Chen 2009). Following Roosevelt’s creation of a Fair Employment
Practices Committee during the war – due to pressure from African American civil rights
32
The text read: “Should negro and white soldiers serve together in all branches of the armed forces?” See
Appendix Table 2, item 2 for the overall marginal distribution.
33
The unweighted gap between FDR voters and Willkie voters is greater than the weighted gap in this
survey. If one compares consistent Democratic voters – i.e. those voting for FDR in 1940 and intending to
vote Democratic for Congress in 1942 and for FDR in 1944 -- to consistent GOP voters (defined
analogously), the gap is also more substantial: core Democratic voters split evenly on integrating the
military (46%-46%), while Republicans opposed integration by a substantial 57%-37% margin.
34
The survey did include items tapping attitudes towards Russia and domestic communists. Among
nonsouthern whites, those favoring a softer line on Russia and opposed to cracking down on domestic
communists were more supportive of integrating the military. Southern whites were virtually unanimous in
opposing military integration.
23
leaders, most notably A. Philip Randolph – there were repeated efforts to enact FEPC
legislation at both the national and state level. Gallup first asked about fair employment
practices in 1945, one year after Republicans adopted a national platform containing a general
endorsement of fair employment legislation while the Democratic platform was silent on the
issue. The Gallup question focused on state laws, rather than federal legislation, which is a
potential advantage since it separates out concerns about federalism. Gallup worded the
question in two ways on the survey: half the respondents were asked about a state law barring
discrimination by employers; the other half were instead asked about a state law requiring
“employees to work alongside persons of any race or color.”35 Respondents split evenly (44%44%) on the ban on employer discrimination, while opposing requiring employees to work
alongside people of other races by a 57%-34% margin. But in both cases, FDR voters were
substantially more supportive than were Dewey voters. Nonsouthern white FDR voters
supported a state ban on employer discrimination by a 52%-31% margin, while Dewey voters
opposed the measure by a 51%-39% margin. While FDR supporters were much less
supportive when the wording focuses on integrated workplaces, they still were significantly
more likely to back the proposal than were GOP voters.36 These relationships also hold up
when demographic controls are included.
Economic liberalism is also strongly related to support for state fair employment
practices at both the bivariate level and when demographics are included in a multivariate
model.37 When one isolates economically liberal FDR voters and compares them to
economically conservative Dewey voters, the gap is even bigger. Thus, economically-liberal
FDR voters back a ban on employer discrimination by a 59%-26% margin, while economically
35
Neither question mentions religion or ethnicity. This makes it less likely that responses were driven by
attitudes towards Catholics, Jews, Italians, etc.
36
Dewey voters opposed requiring integrated workplaces by a decisive 61%-32% margin, while FDR
voters opposed it by a more modest 48%-41% margin.
37
The two forms had different economic policy questions. The form with the item about employer
discrimination asked about government ownership of the railroads and helping the unemployed. The form
with the item on integrated workplaces asked about government ownership of electric utilities and about
government payments for the unemployed.
24
conservative Dewey voters oppose the ban 59%-31%.38 Given that Dewey himself was a key
advocate of New York state’s fair employment practices law – and that the legislation
mentioned in the survey question focused on state, rather than national law – it is striking that
white Republican voters were so much more opposed than white Democratic voters at this
early stage in the civil rights process (see Chen 2009).
Southern whites’ opposition to fair employment practices was overwhelming in the
1945 survey and did not differ appreciably between the two parties. Nonetheless, southern
economic liberals were about 10 points more likely to back a ban on discrimination than were
southern economic conservatives. For example, 24% of the liberals backed the ban on
employer discrimination, as compared to 14% of moderates and 13% of conservatives.39
While this degree of support among white southern economic liberals is far from impressive, it
does reinforce the more general message that while party and racial attitudes were not aligned
in the south, economic views were related to racial attitudes, even among white southerners in
the Jim Crow era. The same questions regarding state fair employment laws were asked again
in July 1947, eliciting a similar pattern of responses.
Gallup continued to ask about fair employment laws on several surveys in the late
1940s and 1950s but with the focus now on federal rather than state legislation. The results tell
much the same story as the survey evidence from the mid-1940s: there is a fairly substantial
gap between nonsouthern Democrats and Republicans in their support for government
intervention against employment discrimination. The gap varies in size across surveys but
without an evident trend. For example, a March 1948 Gallup survey asked how far the federal
government ought to go “in requiring employers to hire people without regard to their race,
religion, color, or nationality” (Appendix Table 3, item 4). The survey was taken shortly after
Truman announced his civil rights program. As a result, it is more vulnerable to the concern
38
The difference is somewhat smaller when the civil rights item focuses on employees rather than
employers.
39
Similarly, 19% of southern liberals backed requiring employees to work in integrated workplaces, as
compared to 11% of moderates, and 6% of conservatives.
25
that elite cues (i.e. attitudes toward Truman) were driving the responses. But this survey’s
results are much the same as the earlier surveys, reinforcing the general message that the masslevel connection between New Deal liberalism and racial liberalism had become reasonably
strong even as national elites were just beginning to grapple directly with the issue.40 While
white nonsouthern Dewey voters overwhelmingly opposed the federal government requiring
non-discrimination (59%-27%), FDR voters narrowly backed it (41%-39%).
Figure 5a presents the same baseline OLS model estimated separately for each poll with
a fair employment question from 1945-72, using presidential vote choice to predict support for
action to prevent job discrimination among nonsouthern whites. While the early items are
entirely from Gallup, Roper and the NES began to ask about job discrimination policy in 1952.
In each case, the dependent variable is recoded to range from 0 to 1. With few exceptions, the
results suggest a strong, consistent relationship between voting Democratic and support for
action against job discrimination. Democratic vote choice is typically associated with a .1 to .2
increase on the 0 to 1 scale in support for policies to combat racial discrimination in
employment. When economic liberalism is substituted for presidential vote choice, the results
are similar in the nonsouth, with a consistent relationship evident in nearly every poll (see
Figure 5b). For southern whites, economic views also appear to be tied to views on action
against job discrimination; though the relationship falls short of statistical significance in
several of the polls, it is positive in all but one case (see Figure 5c). On balance, the evidence
thus suggests that economic conservatism and racial conservatism were connected at the mass
level in the south long before Barry Goldwater entered the scene (see Lowndes 2009).
When one pools the data across multiple surveys and allows for interactions between
partisanship and economic views, the results are again similar to the lynching and poll tax
40
The March 1948 survey asked respondents whether they had heard about Truman’s program, what they
believed it included, and whether they supported specific civil rights measures. When asked about the
program as a whole, the national public was closely divided, with many not registering an opinion or not
having yet heard about the package. Still, white nonsouthern FDR voters from 1944 were more supportive
of the package than were white nonsouthern Dewey voters (29%-13% in favor, as compared to 25%-22%
opposed).
26
cases. Due to the differences in the kinds of questions asked by the different survey houses, in
pooling the fair employment questions I focus on the Gallup polls, which span 1945-1953.
Regression models with a series of dummy variables for each partisan/ideological group
indicate that economically liberal Democratic voters are nearly 25 points more supportive of
government action to combat job discrimination than are economically conservative
Republican voters (see Figure 6).41 Economically conservative Democrats are about 12 points
more supportive than are conservative Republicans. Among Republicans, economic liberals
are once again more supportive of federal action than are their conservative counterparts.
These patterns hold up when controls for demographics are added to the model.42
The partisan and ideological gap was not confined to anti-lynching legislation, the poll
tax, and fair employment practices – though these were the most discussed issues on the civil
rights legislative agenda in the 1940s. At a more general level, when Roper asked in 1947 how
certain “racial and religious” groups are treated in this country, 37% of northern white
Democrats agreed that “strong measures” are needed to help these groups, as compared to 24%
of Republicans.43 Furthermore, 41% of economically liberal Democrats favored strong
measures, as compared to 21% of economically conservative Republicans.
Segregation, Prejudice, and the Partisan Gap on Civil Rights
41
A dummy variable provides a separate intercept for each survey item. The gap is 21 points if one uses
party identification instead of vote.
42
When one estimates the same baseline model using alternative demographic variables as predictors
instead of partisanship and ideology, the generally small point estimates highlight the relative importance
of Democratic vote choice and economic liberalism. Education has only a modest impact, with a college
education associated with just a .04 increase along the scale (SE=.01), as compared to a grade school
education. Union membership is associated with a .06 increase in support (SE=.01). Professionals are
slightly less likely to back fair employment practices (-.045; SE=.02). Urban residence is associated with a
substantial .14 (SE=.01) increase in support for fair employment practices (and farm residence is associated
with a .11 decrease; SE=.01).
43
The question wording was: “Opinions differ as to how certain racial and religious groups are treated in
this country. Which of these three ideas comes closest to expressing your opinion of what the real situation
is? A) Racial and religious groups are, on the whole, as well treated as they should be; (B) While certain
racial and religious groups in this country are sometimes not treated as well as they should be, we are now
improving the situation as fast as is practical; (C) Certain racial and religious groups in this country are
treated very badly, and some strong measures should be taken to improve the situation.” Republicans were
also more likely to say that these groups are treated as fairly as they should be treated (response option A).
27
The survey items discussed above suggest that Democratic partisanship, economic
liberalism, and support for several civil rights policies came into at least partial alignment at
the mass level among northern whites in the late 1930s-early 1940s. For southern whites,
economic liberalism and lower levels of racial conservatism also were related during this
period, even as partisanship was unrelated to civil rights views. But two areas in which the
partisan gap in the north was less clear are perhaps indicative of the challenges that the
Democratic Party would face down the road. First, the evidence with respect to racial
prejudice among whites is more mixed than is the evidence concerning civil rights policy.
That is, white nonsouthern Democrats were only slightly less prejudiced than were white
nonsouthern Republicans during this period, even as they were more likely to endorse
government action to combat discrimination.44 This finding is consistent with Sniderman and
Carmines’ (1997) results showing only a modest correlation between racial prejudice and
measures of ideology and partisanship in the 1990s.
Second, and perhaps not surprisingly given the low correlation between prejudice and
partisanship, the partisan gap on civil rights policy in the 1940s-50s – as well as later – appears
smaller and less consistent when it comes to issues that involve close social contact between
the races. This is evident in early surveys that ask about eating at integrated restaurants or
having an African American nurse.45 It also persists in the late 1940s when Gallup asks about
integrating bus and rail travel: there is generally only a small relationship with partisanship
among northern whites.46 Similarly, evaluations of the Brown decision on school segregation
show little, if any, tie to partisanship. While some items relating to social segregation do show
44
This is based on a preliminary analysis of the data. Perhaps the best indicator of prejudice is a question –
included on several surveys – asking whether African Americans are less intelligent than whites. This item
generally shows northern white Democratic voters were a bit less prejudiced than their Republican
counterparts, but the gap is much smaller than for policy issues such as fair employment. See Sugrue
(2008) on the persistence of prejudice and discrimination in the urban north during the 1940s-60s.
45
A NORC 1944 survey included numerous items that tap into this dimension. White FDR voters look
very similar to their GOP counterparts when it comes to such items as eating in integrated restaurants,
having an African American nurse, or having an African American move in next door.
46
The same surveys show a much bigger gap in views when it comes to lynching, fair employment, the poll
tax, and Truman’s civil rights program as a whole. The wording of the travel desegregation item is
somewhat confusing, suggesting that some caution ought to be used in interpreting its results.
28
a significant partisan gap – such as the soldier integration items in 1942 and 1947 – for the
most part the tie between partisanship and civil rights views is smaller on this set of issues.47
This difference may also explain one wrinkle worth noting in the fair employment
analyses presented above: when questions about job discrimination also refer to housing – as in
the NES 1956-60 surveys, which ask about “jobs and housing” (unlike the 1952 and 1964-1972
NES) – the partisan gap in support for government action is smaller than in the other surveys
(see Figure 5a). While it is possible that the relatively small partisan gap in 1956-60 is due to
the muddled national party messages on civil rights in the late 1950s, it is worth noting that the
Roper item on fair employment from 1957 – which asks only about jobs but not housing –
generates the same substantial inter-party gap as the 1952 NES (and earlier and later surveys
that also focus on jobs). As a result, it is at least plausible that “housing” signified something
very different for respondents from “jobs,” generating the different results.
Interestingly, when one turns to surveys conducted in the years following the 1964
election, the difference between questions focused on job discrimination and inter-racial social
contact persisted. For example, in the 1972 NES, white northern Democratic identifiers are
clearly more supportive of federal action to ensure equal job opportunities for blacks than are
white northern Republicans (39%-36% in favor among Democrats, as compared to 40%-32%
opposed among Republicans). By contrast, Democrats and Republicans are very similar in
their general views on segregation (40% of Democrats favor desegregation over “strict
segregation” or “something in between,” as compared to 39% of Republicans).48
47
The few state-level polls that ask about civil rights also show a clear tie between partisanship and support
for fair employment legislation, while showing mixed evidence when it comes to social segregation. For
example, a 1945 Minnesota Poll asked whether the respondent would mind if an African American family
moved into your neighborhood. The results showed a clear party gap, with Democratic-Farmer Labor
identifiers more liberal than Republicans. Two years later, another Minnesota Poll found a strong tie
between partisanship and support for fair employment legislation, but a much smaller relationship between
partisanship and willingness to work next to people of different races. The same 1947 survey finds a
modest connection between partisanship and willingness to live next to people of different races.
48
The question asked, “What about you? Are you in favor of desegregation, strict segregation, or
something in between.” The same 1972 survey shows that northern Democrats are somewhat more
supportive of school integration than Republicans, but are slightly less supportive of fully integrated
housing.
29
Despite the persistent similarities between northern Democrats and Republicans on
issues relating to segregation in the 1940s-70s, economic liberalism was, with few exceptions,
clearly tied to support for policies to promote integration. For example, while partisanship was
less related to views on the NES items relating to “jobs and housing” than to NES items
relating solely to jobs, economic liberalism had a strong relationship to both sets of items (see
Figure 5b).49 Similarly, when the NES asked about whether the federal government should stay
out of the issue of school integration in 1956-60, economic liberalism was strongly tied to
respondent attitudes even as partisanship bore only a weak, inconsistent relationship.50 This
relationship with economic liberalism held up in later surveys on school segregation and other
forms of integration.
A Gallup series starting in May 1962 asking whether the current administration is
pushing “too fast” or “too slow” on racial integration provides a final window into these
linkages. As with the other items, economic liberalism is clearly related to support for moving
faster on integration throughout the series. When it comes to partisanship, nonsouthern white
Republicans are much more likely to see the Kennedy (and later Johnson) administrations as
pushing too fast for integration, with few Republicans criticizing the administration as “too
slow.” Nonsouthern white Democrats, by contrast, are far less likely than Republicans to see
the administration as pushing too fast, while being a bit more likely than Republicans to see the
administration as moving “too slow.”51 As Figure 7 shows, the size of the inter-party gap
appears stable throughout the 1960s (unfortunately, the question is not asked after 1968). At
the same time, it is worth noting that while Democrats are more supportive of integration than
49
A simple OLS model predicting support for federal action on either “jobs” (1952) or “jobs and housing”
(1956-60), coded as a dummy variable, shows a point estimate of .21 for economic liberalism (coded 0-1)
in 1952 and point estimates ranging from .16 to .21 in 1956-60 (see Figure 7b). (The model is estimated
separately for each survey).
50
For example, in the 1956 NES, just 33% of northern white economic liberals agreed that the federal
government should stay out of school segregation, while 49% of economic conservatives agreed (49% of
the liberals disagreed, as compared to just 34% of the conservatives).
51
A plurality of Democrats see the Administration’s pace as “about right.” Given the tendency for
partisans to support their own party’s president, I focus mainly on the ratio of “too fast” to “too slow”
responses as a first cut in analyzing this data.
30
Republicans, they are hardly clamoring for more aggressive federal action. Even when the
Kennedy administration was moving slowly on integration in 1962 and early 1963, northern
Democrats were not likely to say that the administration was moving too slowly—but they
were much less likely than Republicans to see the administration as moving too fast.52
This Gallup series also provides some interesting additional hints about developments
in the South in the early 1960s. Barry Goldwater’s path to the GOP nomination ran through
the South, where he picked up many delegates through his supporters’ efforts to capture the
local Republican Party machinery (see, e.g., Brennan 1996). Well before Goldwater cast his
famous vote against the 1964 Civil Rights Act, the Gallup data suggest that southern whites
supported the Arizona senator in large numbers for the GOP nomination – and that those
backing Goldwater were particularly likely to believe that the Kennedy administration was
pushing too fast on racial integration. For example, in the early Gallup polls on the GOP
nomination race (those conducted through July 1963), Goldwater was the first choice of 33%
of southern whites, as compared to just 21% of northern whites. An astounding 83% of those
southern Goldwater supporters interviewed before July 1963 claimed that the administration
was pushing too fast on integration, as compared to 61% of southern whites who did not back
Goldwater at the time.53
In sum, the partisan linkage to views on prejudice and segregation among northern
whites is less consistent than is the linkage on such civil rights issues as lynching, the poll tax,
and job discrimination. To the extent that battles over open housing and busing would reveal
the limits of many northern Democrats’ commitment to civil rights, these limitations reflected
enduring features of the party-civil rights alignment rather than a result of a post-1964
52
For example, if one aggregates the surveys from May 1962 through July 1963—a period where the
administration was doing relatively little on integration—17% of northern white Democrats with an opinion
thought the Administration was moving too slow, as compared to 34% who thought it was moving too fast.
By contrast, 14% of Republicans thought the administration was moving too slow, while 53% thought the
administration was moving too fast.
53
The relationship between Goldwater support and views on integration was also significant in the north in
these early polls, though less strong than in the south. The relationship appears to have strengthened over
the course of the nomination campaign in the north while remaining stable in the south.
31
backlash. At the same time, economic liberalism has been tied to northern whites’ views
across the full range of civil rights issues throughout the time period examined here. It is not
obvious why both partisanship and economic views appear to be related to support for some
civil rights policies, while only ideology is strongly tied to others.54 But even as one
recognizes northern white Democrats’ ambivalence on questions of integration, northern
Republicans appear to have entered the 1960s primed to be receptive to racially conservative
appeals. GOP voters had long taken the less racially liberal stance across a range of civil rights
issues than their northern Democratic counterparts, and they were far more eager to criticize
the Kennedy administration from the right (“pushing too fast”) than the left (“pushing too
slow”) on racial integration.55 Furthermore, the southern Republican state parties that the
Eisenhower Administration and the Republican National Committee had helped foster in the
1950s were quickly becoming a power base for the growing Goldwater movement (see Bowen
2011; Brennan 1996), which the Gallup data suggests already had considerable appeal to the
most racially conservative southern whites.
Explaining the Linkages
An obvious important question that emerges from the evidence presented thus far is
what explains the alignment between Democratic partisanship, economic liberalism, and racial
liberalism among northern whites that took shape in the late 1930s. Obstacles to providing a
definitive answer to this question include the absence of racial policy survey questions prior to
1937, the lack of panel data, the limited number of racial attitudes items that are repeated on
multiple surveys, and the limited number of potential explanatory variables that are included
54
This disjuncture hints that different mechanisms may link partisanship and civil rights views as compared
to ideology and civil rights views. Teasing out these dynamics remains a challenge for future work.
55
An item included in the 1948 NORC election study – which surveyed respondents in California, New
York, and Illinois – suggests that this Republican inclination to criticize Democrats as moving too fast on
civil rights existed even at that early date. NORC asked: “Do you think the President has gone too far, or
not far enough, in helping Negroes and other minority groups.” Democratic voters and economic liberals
were each significantly more likely to take the pro-civil rights position on this item than were Republican
voters and economic conservatives.
32
on a consistent basis in the early surveys. Given these limitations, considerable caution is
required but some tentative conclusions can be reached.
At a minimum, one can reject several seemingly plausible explanations. First, as noted
above, it is unlikely that a simple prejudice story accounts for the results. The gap between
northern white Democrats and Republicans in racial prejudice is quite small – both in the
1940s-50s and in later surveys.56 While racial prejudice has a powerful impact on racial policy
attitudes, that impact does not account for the association between partisanship and racial
policy views.57
Second, one can reject the idea that the partisan divide only emerged when civil rights
issues affected the north. Some of the issues discussed above directly affected only the south
(e.g. lynching, the poll tax), while others challenged northern race relations (e.g. fair
employment practices). The consistency of the partisan and ideological gap across both sets of
issues suggests that a narrow explanation – e.g. that Republicans were willing to target
discrimination in the south but not the north – does not hold up.58
Third, it is difficult to argue that cues from national party elites generated the masslevel connections discussed here. The most visible national Democratic leader before 1945,
Franklin Roosevelt, kept quiet on civil rights for fear of alienating southern Democrats in
Congress. While there were several prominent Democrats who advocated for civil rights –
including Eleanor Roosevelt and Robert Wagner (D-NY) – most national leaders avoided the
issue, including the top Democratic leaders in both the House and Senate. In the 1936-44
presidential elections, the Republican national platform said more about civil rights and
advocated a more liberal position than the Democratic platform (Carmines and Stimson 1989).
56
See Sniderman and Carmines 1997 for evidence that the correlation between partisanship and prejudice
was limited even in the 1990s.
57
The earliest surveys do not include both civil rights policy and prejudice items on the same poll.
However, in later surveys, the relationship between New Deal liberalism and civil rights support is robust
to controlling for prejudice measures.
58
Republicans’ pro-business orientation surely encouraged resistance to fair employment practices
legislation, which was viewed as interfering in the employer-employee relationship. But partisanship and
economic views were also related to support for such civil rights proposals as anti-lynching legislation, the
poll tax ban, and integrating the military, none of which posed a threat to business interests.
33
National Democrats did eventually send clear signals of their civil rights liberalism,
particularly in 1948 when Truman made a major push for civil rights legislation and the party
adopted a liberal civil rights plank at its national convention (though Truman’s forces sought a
weaker platform plank in order to prevent a southern bolt). The elite cues perspective would
thus lead us to expect a much sharper tie between partisanship and civil rights views starting in
1948. But the inter-party gap was about as big before 1948 as following the campaign (see
Figures 2, 4, and 5).59 Indeed, on all 12 survey items asking about lynching, the poll tax,
soldier integration, and fair employment practices from December 1937 to December 1947,
there is a clear gap separating northern white Democrats and Republicans, with Democrats
taking the more liberal position.60 It simply does not appear that the relationship between civil
rights views and partisanship (or ideology) tightened amid the 1948 campaign or in its
immediate aftermath.61
Finally, changes in the demographic group composition of the parties fails to account
fully for the relationship between New Deal Democratic liberalism and racial policy liberalism.
Across the vast majority of civil rights items, controlling for a wide range of demographic
variables does not eliminate – and generally barely changes – the estimated relationships
among the key variables of interest.
This still leaves open the possibility that the partisan differences on civil rights are
particularly sharp among certain subgroups in the electorate – and such subgroup differences
could provide clues about the origins of the partisan differences. To explore this possibility, I
estimated a series of models pooling data across surveys, analyzing nonsouthern whites’
59
Changes in question wording over time require some caution in interpreting the relative size of the
estimates. But close inspection of the question wording and response distributions over time does not
provide a basis for thinking that the later questions somehow artificially reduced the party gap relative to
the early questions (e.g. by having a more lopsided response distribution). However, additional work on
scaling the items may help address this potential concern.
60
The 1947 Roper item on whether strong measures are needed to help minorities also shows the identical
pattern (see discussion above). The twelve items are spread across nine different surveys.
61
While the national Democratic Party did take a clear civil rights position in 1948, this clarity was shortlived. National leaders succeeded in adopting a weaker national platform in 1952 and 1956, while nominee
Adlai Stevenson presented a muddled civil rights message.
34
support for each set of policies with numerous repeated items (anti-lynching policies, the poll
tax ban, and fair employment practices) as a function of Democratic presidential vote. This
time, however, the models are estimated separately for a variety of demographic groups.62
Figure 8a presents the results when support for action against lynching is the dependent
variable.63 The results are remarkably consistent across groups: Democratic presidential vote
was significantly associated with support for federal action for both urban and non-urban
residents; farm residents and non-farm residents; small-town residents; professionals and nonprofessionals; laborers and non-laborers; women and men; those under age 35, age 35-50, and
50 and above; those with only a grade school education, those who completed part of high
school, high school graduates, and those with at least some college education; union and nonunion members; phone-owners and those without a phone at home; and for Midwesterners,
Westerners, and Northeastern residents. The association between Democratic voting and
lynching policy views was greater for urban residents than for farmers and small-town
residents (.185 in urban areas, as compared to .08 for farm residents and .07 in small towns),
among men (.15 vs. .07 for females), and for the most educated. But even the least educated
group showed a strong relationship (.11 for those with no high school, as compared to .17 for
those with at least some college.).
Figure 8b displays the analogous results when banning the poll tax is the dependent
variable. As with the lynching items, the association between partisanship and support for civil
rights is stronger in urban areas – with the association in farm areas and small towns falling
short of statistical significance. Men once again also show a stronger tie than women, though
the association is significant for each group. The association is also stronger for the most
educated groups, though there appears to be evidence of at least some relationship between
62
An alternative approach, to be explored in future work, is to estimate hierarchical models that allow for
parameters of interest to vary across groups.
63
The model is estimated for all polls with lynching questions starting in December 1937, with a separate
intercept included for each poll.
35
support for the poll tax ban and presidential vote choice among the least educated group as
well.64
Finally, Figure 8c estimates the same set of models with the Gallup questions
concerning job discrimination policy as the dependent variable. Once again, the tie between
partisanship and civil rights liberalism is strongest in urban areas. For farm areas, the
relationship appears to emerge a bit later: if one restricts the analysis to polls conducted from
1948 on, there is a significant relationship between partisanship and support for the poll tax
ban, though again it is much smaller than for urban areas. The relationship is also stronger for
professionals than non-professionals. Once again, the strongest relationship is evident among
the most educated, but there is a significant relationship even for the least educated groups.
These results reinforce the message that the alignment between party and race was strongest in
urban areas and was somewhat stronger among those with higher socioeconomic status, but
that the connection was by no means restricted to a narrow slice of the electorate and reached
even those with lower socioeconomic status.65
These results cast doubt on the idea that the partisanship-ideology-civil rights linkage
emerged as a straightforward product of group-membership based politics. That is, union
members and non-members, professionals and non-professionals, low education and high
education respondents, all display broadly similar relationships. This is not simply a story of
CIO members following cues from their leadership and supporting the civil rights agenda
promoted by their leaders. It also does not appear to simply be a story of educated Democrats
responding to subtle cues emanating from party leaders.66
64
It is worth noting that the bivariate relationship between education and support for the poll tax ban is
relatively modest: college educated northern whites are only about seven points more supportive of the poll
tax ban than are northern whites with only a grade school education.
65
While the tie between civil rights views and partisanship is, on the whole, stronger for those with higher
education, this may have little to do with the particular features of civil rights as an issue area. When one
examines the relationship between party identification and economic issue views, one finds a similar
pattern: the most educated respondents typically show tighter constraint than the less educated (see
Converse 1964and Zaller 1992 more generally on this strong tendency).
66
Cues from intellectual leaders – highlighted by Noel (2012) – likely played a role in forging these
connections, but the extent to which the connections spanned across different groups (even reaching into
36
Nonetheless, changes in the group composition of the Democratic Party in the 1930s
may well have played a role in promoting the linkages among racial liberalism, Democratic
partisanship, and economic liberalism (Karol 2009; Feinstein and Schickler 2008). With the
1936 elections, African Americans, CIO unions, urban liberals, and Jews emerged as core
voters in the Democratic coalition outside the south. These same groups were prominent civil
rights supporters (Kesselman 1948). Thus, even as Roosevelt kept silent on civil rights, rankand-file voters may have taken the cue from these affiliated groups that the Democratic brand
includes – or ought to include – civil rights. These cues need not have been limited to group
members. That is, as the Democratic Party in the North increasingly became associated with
African Americans, Jews, and CIO unions, some voters may have been repelled by these new
groups, while other voters – though not belonging to any of these groups – may have viewed
them more favorably. It also may be that urban voters were particularly aware of the changing
composition of the party coalition, and thus were quicker to respond to it, with conservatives
increasingly repelled by the new associations and liberals more receptive.67
The CIO emerged as a particularly salient cue to many voters in the late 1930s, as it put
forward a bold policy agenda and used assertive organizing tactics – such as sit-down strikes –
that made it a lightning rod for conservatives and a hero to many on the left (see Schickler and
Caughey 2011). Before the CIO entered the scene, the labor movement was by no means
identified with civil rights; instead, AFL-dominated unions generally had a poor record on race
relations (and AFL unions continued to have at best a mixed record long after the CIO became
a competitor). One indicator of labor’s distance from civil rights in the pre-CIO era is that a
1934 petition circulated by the NAACP urging Roosevelt to support anti-lynching legislation
the south, when it comes to the economic liberalism-racial liberalism tie), suggests that broader forces were
at work as well.
67
An extension of this analysis might be to use a multi-level model with post-stratification to estimate the
party gap at the state-level over time (see, e.g., Lax and Phillips 2009). A key question would be whether
the party gap emerged at the same time across a wide range of states or was concentrated in a handful of
urban states. Figures 8a-c show that the gap was not confined to a single region (e.g. the Northeast, West,
or Midwest) but a more fine-grained, state-level analysis may provide additional insight into the timing of
changes in the civil-rights/party linkage.
37
included signatures from governors, mayors, attorneys, ministers, journalists, and college
presidents, but no labor union officials were identified as signatories.68 But just a few years
later, CIO officials would assume a prominent role in such pro-civil rights organizations as the
Southern Conference on Human Welfare (formed in 1938) and the National Committee to
Abolish the Poll Tax (formed in 1942).
From early on, the CIO stood out among white-led organizations in the extent to which
it publicized its support for civil rights. The CIO News, the official weekly that was distributed
to CIO households across the U.S., included many articles promoting civil rights initiatives
soon after it began publishing in December 1937. For example, the January 29, 1938 issue
included an article headlined, “CIO Attacks Filibuster on Lynching Bill.” When the CIObacked Labor Non-Partisan League issued a list of major bills it was monitoring in February
1938, the lynching bill was placed alongside legislation on strikers’ rights, wages and hours,
and farm tenants (CIO News, February 26, 1938, p. 3). Subsequent CIO News stories
promoted the poll tax ban and FEPC, and praised Court decisions on behalf of equal education
opportunity. More generally, the CIO News sought to link the issues of labor rights and
policies against racial discrimination under the common rubric of “civil rights.” Thus, many
stories referred to the right-to-strike as a “civil right,” just as other stories referred to African
Americans’ struggle in similar terms. In the CIO vocabulary, the enemies of African American
rights and the enemies of labor rights were the same: southern “Tories,” business interests that
sought to weaken labor by dividing workers on the basis of race, and even fascists from abroad
(see Figure 9 for an example of the latter appeal).69
68
A copy of the petition was obtained from the Edward Costigan papers at the University of Colorado.
Half of the governors were Republicans and half were Democrats; the former group included Alf Landon
of Kansas. Roosevelt did not respond to the petition’s plea for action.
69
A fuller account of the CIO’s role will be provided in the book. The rich literature on the CIO and race
relations makes clear that several CIO affiliates had a mixed record on civil rights (see, e.g., Zieger 1995;
Stein 1993; Stevenson 1993; Riker 1948; Mason 1945; Korstad 1993; Goldfield 1993; Gerstle 1993).
However, the union nonetheless stands out as one of the very few predominately white organizations in the
1930s-40s that vocally advocated for civil rights legislation. Accounts of civil rights politics written in the
late 1930s and 1940s almost invariably identify the CIO as among the handful of predominately white
organizations that joined African Americans in promoting civil rights initiatives.
38
It is difficult to know how much causal weight to put on the CIO’s push for civil rights
in shaping whites’ attitudes. The direct effects may well have been limited. When one models
views on lynching, the poll tax, and fair employment practices as a function of respondents’
demographic characteristics, CIO membership is generally a statistically significant predictor,
but the substantive size of the association is modest: CIO members are generally about .02-.04
more supportive of these initiatives (on a 0 to 1 scale) than non-members, controlling for other
demographics. But the CIO – along with allied groups, such as the Labor Non-Partisan
League, the American Civil Liberties Union, the Southern Conference on Human Welfare, and
the Union for Democratic Action (a forerunner of the Americans for Democratic Action) –
may have had an impact on even non-members by shaping the meaning of liberalism to include
civil rights.70 At the same time, African American activists in the NAACP and National Negro
Congress, among other groups, worked to keep the pressure on these urban liberals to
incorporate civil rights into their list of priorities.
Another, related possibility is that the meaning of New Deal liberalism itself became
sharpened amidst the Democrats’ travails of 1937-38. The fierce reaction against the sit-down
strikes and court-packing, along with the recession of 1937-38, may have led some members of
the broad Roosevelt coalition of 1936 to peel off from the party, while sharpening the
ideological division between FDR’s supporters and opponents. The emergence of a viable
conservative opposition to New Deal liberalism may have generated a clearer sense among
voters of what it means to be a “New Deal Democrat” and what it means to be anti-New Deal.
Similarly, the prominent role of the CIO in the Democratic coalition presumably was most
likely to alienate those voters who had a more modest view of the New Deal’s aspirations.
This sharpening of lines may have helped foster a closer alignment between New Deal
economic liberalism and racial liberalism. That is, the individuals who continued to support
70
News stories from the late 1930s credit the CIO with playing a major role in swinging African American
voters to the Democrats (see, e.g., Frank Kent, “The Great Game of Politics: A Solid Black Belt?” Wall
Street Journal, June 30, 1938; Kent’s column was widely syndicated).
39
FDR and the Democrats amidst the setbacks of 1937-38 may have also been more likely to be
the type of people who support the broad activist government required to safeguard civil
rights.71 Similarly, the types of people who drifted away from Roosevelt and the Democrats as
the meaning of New Deal economic liberalism sharpened in the late 1930s may have been the
same kinds of people reluctant to support broad social change, such as civil rights.
It seems safe to rule out the potential explanation that anti-civil rights attitudes
themselves caused white non-southerners to become economic conservatives or Republicans in
this era. The low salience of civil rights on the policy agenda – and the paucity of clear
national elite cues on the issue in the late 1930s – makes it implausible to believe that the
connection between racial and economic conservatism that was forged at the mass level in the
1930s reflects a causal effect of racial policy views on economic attitudes. A more plausible
alternative is that the connection between economic and racial conservatism is rooted in a more
general stance towards the role of government in addressing social and economic problems.
This stance could well be rooted in individual personality characteristics, as recent research has
provided powerful evidence that personality traits – such as openness to new experiences – can
shape a broad range of attitudes and ideological dispositions (Gerber et al 2010). It may be
that the New Deal economic policy “brand” resonated to the same types of people likely to
view the treatment of African Americans as a policy problem, even in the absence of explicit
elite cues.72 The observation that the tie between civil rights and economic views existed even
in the white south – where group-based cues supportive of civil rights were far weaker than the
north – provides additional reason to take seriously the idea that broad ideological dispositions
were at work.
71
While white Democrats clearly did not join the party in this era due to concerns about civil rights, it is
plausible that the kind of people attracted to the party were also the kind of people more likely to support
liberal civil rights policies.
72
In other words, as New Deal economic liberalism came to signify a strong governmental role in promoting
economic opportunities for the lower classes and empowerment of such previously subordinated groups as labor
unions, this programmatic vision resonated with the same kinds of people who were likely to see government
activism to help racial minorities as appropriate.
40
Discussion
In their classic study of the civil rights realignment, Carmines and Stimson (1989)
correctly emphasize the unsettled nature of civil rights politics within the national parties as of
the late 1950s and early 1960s. Many national Democratic elites – such as John F. Kennedy
and Adlai Stevenson – sought to straddle the issue, avoiding a clear stand that would alienate
their party’s southern wing. At the same time, many northern Republican elites – such as
Richard Nixon – also avoided taking a clear stand but were identified as at least somewhat
supportive of the civil rights cause. The national leaders of both parties saw the explosive
potential of the civil rights issue – not just to realign the electorate but to empower new actors
within each party. While some leaders – Humphrey and Goldwater come to mind – were eager
to embrace that potential, others were not. The open-endedness of the situation highlighted by
Carmines and Stimson captures this essential dilemma facing national party elites.
Below the surface, however, a series of developments dating back to the 1930s and
1940s had remade both parties so that the intra-party pressures in favor of an embrace of civil
rights liberalism were much stronger on the Democratic side, while Republicans increasingly
were pushed towards racial conservatism and a concomitant alliance with southern anti-civil
rights forces. These developments included the mass-level alignment of racial liberalism with
economic liberalism and Democratic partisanship among northern whites; the increased role of
African Americans in the Democratic coalition in key northern states (and the reluctance of
economically-liberal African Americans to vote Republican);73 the gradual reorientation of
northern Democratic state parties in the 1940s to become more of a home for racial liberals; the
efforts of northern Democratic members of Congress to force civil rights measures to the floor
even as that challenged a key pillar of the national party coalition; the early alignment between
economic liberalism and less strident racial conservatism in the south, and, starting in the late
73
The book project will devote considerable attention to the African American realignment. An important
point, not lost on political observers in the 1930s-50s, was that African Americans’ strong economic
liberalism left them significantly closer to the Democrats on issues other than civil rights, making it more
difficult for Republicans to envision winning back a substantial share of African American voters (see
Schickler and Caughey 2011).
41
1950s, the capture of revived southern state Republican organizations by activists who
articulated an extreme brand of both racial and economic conservatism.74
The patterns of mass opinion described here are not best viewed as the cause of the
realignment. But they meant that the wind would be at the back of liberal Democrats – such as
Humphrey and other Democrats elected in 1948, such as Paul Douglas (D-IL) – who backed
civil rights either out of personal conviction or as a way to appeal to African American voters
or other Democratic coalition partners, such as CIO unions.75 Similarly, rank-and-file
Republican voters’ greater skepticism towards government policies promoting civil rights
provided a permissive backdrop for entrepreneurial conservative politicians as they sought to
build a coalition with southern whites.
Translating this permissive mass opinion context into an actual change in national party
alignments required the confluence of two forces. First, the relatively independent electoral
bases of state parties and rank-and-file members of Congress provided an institutional foothold
for those seeking to redefine the parties’ stance towards civil rights. Second, the AfricanAmerican-led civil rights movement raised the salience of the civil rights issue, forcing
politicians to choose sides rather than simply to straddle the issue. These two sets of forces
reinforced one another: civil rights activists pushed Democratic politicians to put legislation
onto the agenda at the state and national levels; these legislative efforts then further raised the
salience of the issue and helped sharpen the divisions between pro- and anti-civil rights forces.
74
A full account of changes in the South would also need to grapple with the effort to organize a coalition
of urban liberals, CIO unions, and African Americans in the region starting in 1938 with the formation of
the Southern Conference on Human Welfare (see Sullivan 1996; Krueger 1967). The SCHW fell apart in
the mid-to-late 1940s due to the communism issue, as well as racial divisions, but it nonetheless took a bold
stand in favor of civil rights from its inception and sought to define a broad liberal agenda for the region
that linked economic and racial liberalism. The CIO’s efforts to organize integrated unions in the South in
the late 1930s and 1940s may have also played a role in forging a linkage between economic and racial
liberalism in the region.
75
Accounts of Humphrey’s rise are striking in this regard. Minnesota had a small African American
population, but Humphrey’s early embrace of civil rights liberalism helped demonstrate that his version of
anti-communist liberalism was not hopelessly compromised, as his critics from the left wing of the
Democratic-Farmer Labor Party charged (see Delton 2002; Thurber 1999).
42
A multi-layered historical perspective that incorporates both institutional and masslevel dynamics thus offers a way to move beyond the debate concerning whether the civil
rights realignment was the product of elite choice at a critical juncture or instead was
determined by deep structural forces. National political elites – such as Kennedy, Nixon,
Johnson, and Goldwater – faced genuine decisions on how to position themselves and where to
try to lead their parties on civil rights. The trade-offs involved meant that there was nothing
automatic or predetermined about the decisions that resulted. However, national political
leaders were making these decisions in a context in which economic liberalism and racial
liberalism had long been linked together at the mass level and in which key groups within the
Democratic coalition were pushing hard for a liberal stance on civil rights. These group and
mass-level dynamics had already been working their way through the state parties and rankand-file congressional membership when national elites were finally forced to choose sides.76
When Barry Goldwater spoke of hunting “where the ducks are” in the 1960s, he was
capitalizing upon a longstanding mass-level connection between economic and racial
conservatism in the south – rather than creating this linkage. The willingness of congressional
Republicans to help southern Democrats stall civil rights legislation in the late 1940s and
1950s was a harbinger of the future partisan alignment. Goldwater’s nomination in 1964 thus
represented the culmination of a long process of change within GOP, just as Johnson’s
personal transformation on civil rights reflected deep currents in the party he sought to lead.
76
These intra-party dynamics may help understand a paradox regarding the Democratic presidents of the 1940s60s. Truman, Kennedy, and Johnson had evidenced little interest in civil rights prior to assuming office; none of
them were part of the Hubert Humphrey / ADA wing of strong civil rights supporters. Yet each of these
presidents ended up eventually sponsoring major civil rights initiatives that played an important role in the
developing drama. National political incentives – the imperative of keeping southern Democrats on board in
November – led national Democratic elites to seek out nominees and platforms that were broadly acceptable. But
once in office these Democratic presidents were forced to make decisions that meant taking sides and the balance
of pressures within their party coalitions pushed more in the direction of supporting civil rights, regardless of their
personal predispositions.
43
References
Baum, Matthew A. and Samuel Kernell. 2001. “Economic Class and Popular Support for
Franklin Roosevelt in War and Peace.” Public Opinion Quarterly 65: 198-229.
Berinsky, Adam J. 2006. “American Public Opinion in the 1930s and 1940s: The Analysis
of Quota-Controlled Sample Survey Data.” Public Opinion Quarterly 70: 499-529.
Berinsky, Adam J., Eleanor Powell, Eric Schickler, and Ian Yohai. 2011. “Revisiting
Public Opinion in the 1930s and 1940s.” PS: Political Science and Politics. 44(3):515–
520.
Berinsky, Adam J., and Eric Schickler. 2006. “The American Mass Public in the 1930s and
1940s.” National Science Foundation, Collaborative Research Grant.
http://web.mit.edu/berinsky/www/nsf.pdf
Berinsky, Adam J. and Eric Schickler. 2011. “The American Mass Public in the 1930s and
1940s” [Computer file]. Individual surveys conducted by the Gallup Organization, Roper
Organization, NORC, and The Office of Public Opinion Research [producers], 1936-1945.
Bowen, Michael. 2011. “The First Southern Strategy: The Taft and the
Dewey/Eisenhower Factions in the GOP.” In Painting Dixie Red: When, Where, Why,
and How the South Became Republican, Glenn Feldman, ed. Gainesville: University
Press of Florida.
Brennan, Mary. 1996. Turning Right in the Sixties: The Conservative Capture of the GOP.
Chapel Hill and London: University of North Carolina Press.
Brinkley, Alan. 1995. The End of Reform. New York: Knopf.
Cantril, Hadley and Mildred Strunk. 1951. Public Opinion, 1935-1946. Princeton:
Princeton University Press.
Caldeira, Gregory A. 1987. “Public Opinion and The U.S. Supreme Court: FDR’s CourtPacking Plan.” The American Political Science Review, 81: 1139-1153.
Carmines, Edward G., and James A. Stimson. 1989. Issue Evolution: Race and the
Transformation of American Politics. Princeton, NJ: Princeton University Press.
Chen, Anthony. 2006. “‘The Hitlerian Rule of Quotas’: Racial Conservatism and the
Politics of Fair Employment Legislation in New York State, 1941–1945,” Journal of
American History 92: 1238.
Chen, Anthony. 2007. “The Party of Lincoln and the Politics of State Fair Employment
Practices Legislation in the North, 1945-64,” American Journal of Sociology 112: 1713–
74.
44
Chen, Anthony. 2009. The Fifth Freedom: Jobs, Politics, and Civil Rights in the United
States, 1941-1972. Princeton: Princeton University Press.
Chen, Anthony, Robert Mickey and Robert Van Houweling. 2008. “Explaining the
Contemporary Alignment of Race and Party: Evidence from California’s 1946 Ballot
Initiative on Fair Employment,” Studies in American Political Development 22: 204-228.
Converse, Jean. 1987. Survey Research in the United States: Roots and Emergence.
Berkeley: University of California Press.
Converse, Philip. 1964. “The Nature of Belief Systems in Mass Publics.” In Ideology
and Discontent, ed. David E. Apter. New York: Free Press.
Eisinger, Robert M. and Jeremy Brown. 1998. “Polling as a Means Toward Presidential
Autonomy: Emil Hurja, Hadley Cantril and the Roosevelt Administration.” International
Journal of Public Opinion Research 10 (3): 237-256.
Farhang, Sean, and Ira Katznelson. 2005. “The Southern Imposition: Congress and Labor
in the New Deal and Fair Deal.” Studies in American Political Development, 19 (April),
1-30.
Feinstein, Brian D. and Eric Schickler. 2008. “Platforms and Partners: The Civil Rights
Realignment Reconsidered.” Studies in American Political Development.
Gerber, Alan S., Gregory A. Huber, Conor M. Dowling, and Shang Ha. 2010.
“Personality and Political Attitudes: Relationships Across Issue Domains and Political
Contexts.” American Political Science Review, 104 (February): 111-33.
Igo, Sarah. 2007. The Averaged American: Surveys, Citizens, and the Making of a Mass
Public. Cambridge: Harvard University Press.
Jenkins, Jeffrey, Justin Peck, and Vesla Weaver. 2010. “Between Reconstructions:
Congressional Action on Civil Rights, 1890-1940.” Studies in American Political
Development 24: 57-89.
Karol, David. “Realignment without Replacement: Issue Evolution and Ideological
Change among Members of Congress” (paper presented at the Annual Meeting of the
Midwest Political Science Association, Chicago, IL, April 15–18, 1999).
Karol, David. 2009. Coalition Management: Explaining Party Position Change in
AmericanPolitics. Cambridge University Press.
Katznelson, Ira. 2012. Fear Itself: New Deal Democracy in a Southern Cage. Norton
(forthcoming).
45
Katznelson, Ira, Kim Geiger, and Daniel Kryder. 1993. “Limiting Liberalism: The
Southern Veto in Congress, 1933-1950.” Political Science Quarterly 108:283-306.
Katznelson, Ira, and John Lapinski. 2006. “Congress and American Political Development:
Missed Chances, Rich Possibilities.” Perspectives on Politics 4(2): 243-260.
Kesselman, Louis. 1948. The Social Politics of FEPC: A Study in Reform Pressure
Movements. Chapel Hill: University of North Carolina Press.
Krueger, Thomas. 1967. And Promises to Keep: The Southern Conference for Human
Welfare, 1938-1948. Nashville, TN: Vanderbilt University Press.
Lax, Jeffrey R., and Justin H. Phillips. 2009. Gay Rights in the States: Public Opinion
and Policy Responsiveness. American Political Science Review 103 (August) 367-86.
Lee, Taeku. 2002. Mobilizing Public Opinion. Chicago: University of Chicago Press.
Levendusky, Matthew. 2009. The Partisan Sort: How Liberals Became Democrats and
Conservatives Became Republicans. Chicago: University of Chicago Press.
Lowndes, Joseph. 2009. From the New Deal to the New Right: Race and the Social
Origins of Modern Conservatism. New Haven: Yale University Press.
Mayhew, David R. 2000. America’s Congress: Actions in the Public Sphere, James
Madison Through Newt Gingrich. New Haven: Yale University Press.
McAdam, Doug. 1999. Political Process and the Development of Black Insurgency,
1930-1970. Chicago: University of Chicago Press.
McMahon, Kevin J. 2004. Reconsidering Roosevelt on Race. Chicago: University of
Chicago Press.
Noel, Hans. 2012. “The Coalition Merchants: The Ideological Roots of the Civil Rights
Realignment.” Journal of Politics.
Norpoth, Helmut, Victor Lange, and Michael Morzenti. 2009. “Voting in Wartime: The
1944 Election.” Paper presented at the Midwest Political Science Association Annual
Meeting, April 2009.
Orren, Karen, and Stephen Skowronek. 1998-99. “Regimes and Regime Building in
American Government: A Review of Literature on the 1940s.” Political Science
Quarterly (Winter 1998-99), v. 113, pp. 689-702.
Orren, Karen, and Stephen Skowronek. 2004. The Search for American Political
Development. Cambridge: Cambridge University Press.
46
Ruggles, Steven, Matthew Sobek, Trent Alexander, Catherine A. Fitch, Ronald Goeken,
Patricia Kelley Hall, Miriam King, and Chad Ronnander. 2004. “Integrated Public Use
Microdata Series: Version 3.0”. Minneapolis, MN: Minnesota Population Center.
Rusk, Jerrold. 2001. A Statistical History of the American Electorate. Washington, DC:
CQ Press.
Schickler, Eric, and Devin Caughey. 2011. “Public Opinion, Organized Labor, and the
Limits of New Deal Liberalism, 1936-1945.” Studies in American Political Development
25: 162-89.
Schickler, Eric, and Kathryn Pearson. 2009. “Agenda Control, Majority Party
Power, and the House Committee on Rules, 1937-1952.” Legislative Studies
Quarterly, Forthcoming.
Schickler, Eric, Kathryn Pearson, and Brian Feinstein. 2010. “Shifting Partisan
Coalitions: Support for Civil Rights in Congress from 1933-1972.” Journal of Politics.
72(3): 672-89.
Schlozman, Kay Lehman, and Sidney Verba. 1979. Injury to Insult: Unemployment,
Class, and Political Response. Cambridge: Harvard University Press.
Sniderman, Paul M., and Edward G. Carmines. 1997. Reaching Beyond Race.
Cambridge, MA: Harvard University Press.
Sniderman, Paul M., and Edward H. Stiglitz. 2009. The Rational Partisan. Book
manuscript, Stanford University.
Sugrue, Thomas. 2008. Sweet Land of Liberty: The Forgotten Struggle for Civil Rights
in the North. New York: Random House.
Sullivan, Patricia. 1996. Days of Hope: Race and Democracy in the New Deal Era.
Chapel Hill and London: University of North Carolina Press.
Verba, Sidney and Kay Lehman Schlozman. 1977. “Unemployment, Class
Consciousness, and Radical Politics: What Didn’t Happen in the Thirties.” The Journal of
Politics 39: 291-323.
Weatherford, M. Stephen, and Boris Sergeyev. 2000. “Thinking about Economic
Interests: Class and Recession in the New Deal.” Political Behavior, 22: 311-339.
Weir, Margaret. 2005. “States, Race, and the Decline of New Deal Liberalism.” Studies
in APD, vol 19 (October), pp. 157-72.
Zaller, John R. 1992. The Nature and Origins of Mass Opinion. Cambridge: Cambridge
University Press.
47
Figure 1: Difference in Expected Values between Northern Democrats’ &
Northern Republicans’ Likelihood of Signing Civil Rights Discharge
Petitions
48
.1
-.1
0
Coefficient
.2
Figure 2a: Relationship between Democratic Vote Choice and Support for Federal
Intervention Against Lynching, 1937-1950. Northern whites. (Dependent and
independent variables are scored 0,1; point estimate and 95% confidence interval
presented for each survey).
1936
1938
1940
1942
1944
Year
1946
1948
1950
1952
.15
.1
0
.05
Coefficient
.2
.25
Figure 2b: Relationship between Economic Liberalism and Support for Federal
Intervention Against Lynching, 1937-1950. Northern whites. (Dependent variable is
scored 0,1; independent variable ranges from 0 to 1).
1936
1938
1940
1942
1944
Year
1946
1948
1950
1952
49
-.4
-.2
0
Coefficient
.2
.4
Figure 2c: Relationship between Economic Liberalism and Support for Federal
Intervention Against Lynching, 1937-1950. Southern whites. (Dependent variable is
scored 0,1; independent variable ranges from 0 to 1).
1936
1938
1940
1942
1944
Year
1946
1948
1950
1952
50
Note: Excluded category consists of Economically-conservative Republicans.
51
-.05
0
.05
Coefficient
.1
.15
Figure 4a: Relationship between Democratic Vote Choice and Support for Poll Tax Ban,
1940-1953. Northern whites. (Dependent and independent variables are scored 0,1).
1940
1942
1944
1946
1948
1950
1952
1954
Year
.1
0
.05
Coefficient
.15
.2
Figure 4b: Relationship between Economic Liberalism and Support for Poll Tax Ban,
1940-53. Northern whites. (Dependent variable is scored 0,1; independent variable
ranges from 0 to 1).
1940
1942
1944
1946
1948
Year
1950
1952
1954
52
.2
0
-.2
Coefficient
.4
.6
Figure 4c: Relationship between Economic Liberalism and Support for Poll Tax Ban,
1940-53. Southern whites. (Dependent variable is scored 0,1; independent variable
ranges from 0 to 1).
1940
1942
1944
1946
1948
Year
1950
1952
1954
53
Figure 5a. Relationship between Democratic Vote Choice and Support for Fair Employment Policies. Northern whites. (Dependent
and indep. variables range from 0 to 1). Note: Gallup=blue lines; Roper=green lines; NES: red lines. NES 1956-60 items refer to
“jobs and housing.”
54
Figure 5b: Relationship between Economic liberalism and Support for Fair Employment Policies. Northern whites. (Dependent and
independent variables range from 0 to 1). Note: Gallup=blue lines; Roper=green lines; NES: red lines.
55
Figure 5c: Relationship between Economic liberalism and Support for Fair Employment Policies. Southern whites. (Dependent and
independent variables range from 0 to 1). Note: Gallup=blue lines; Roper=green lines; NES: red lines.
56
Note: Excluded category consists of Economically-conservative Republicans.
57
Figure 7: Comparison of Northern White Republicans and Democrats on Whether Administration Pushing Immigration Too Fast vs.
Too Slow, 1962-68. (Positive score indicates more likely to see administration as moving too fast than too slow).
58
59
60
61
Figure 9: CIO News Cartoon, November 28, 1938.