Polls and elections: party identification in the 2012 presidential election.
Winneg, Kenneth M. ; Jamieson, Kathleen Hall ; Hardy, Bruce W. 等
Background
Although party affiliation's role in individual political
behavior in the United States has been studied since the 1940s
(Berelson, Lazarsfeld, and McPhee 1954; Lazarsfeld, Berelson, and Gaudet
1944), it became a central focus in the Michigan studies of the 1950s
and 1960s (Belknap and Campbell 1951; Campbell, Converse, Miller, and
Stokes 1960). Indeed, as Nie, Verba, and Petrocik (1979, 47) noted, in
the 1950s and 1960s, party affiliation "was the central thread
running through interpretations of American politics" where it was
considered "a stable characteristic of the individual: it was
likely to be inherited, it was likely to remain steady throughout the
citizen's political life, and it was likely to grow in strength
during that lifetime."
Party identification is important in part because of its power to
predict the presidential vote choice (Barrels 2000). Analyzing data from
the American National Election Survey (ANES) from 1952 through 2004,
Bafumi and Shapiro (2009) showed that party identification achieved its
highest predictive value in 1996 and 2004. Consistent with that
analysis, the 2012 National Exit Poll revealed that more than nine in 10
self-identified Republicans (93%) reported casting their ballots for
their party nominee, and a nearly equal proportion (92%) of self-styled
Democrats sided with the Democratic incumbent president Barack Obama
(Cable News Network [CNN] 2012). Indeed strong party attachment predicts
straight ticket voting (Schaffner, Streb, and Wright 2001).
The drop in the number of citizens reporting a strong tie to party,
a decline that occurred between the mid-1960s and mid-1970s and leveled
off in the late-1970s and early-1980s, did not change the fact that most
continued to identify with a party (Miller and Shanks 1996). But changes
in partisan affiliation did focus scholars on its variability from one
election to another with some attributing the differences to simple
survey measurement error (Green and Palmquist,1990, 1994; Johnston
2006), and others arguing that party identification can be affected in
the short run by events (Franklin and Jackson 1983; Markus and Converse
1979).
Using the Annenberg Public Policy Center of the University of
Pennsylvania's (APPC) National Annenberg Election Survey's
(NAES) rolling cross-section data, we tracked the root party
identification (whether respondent identifies as a Democrat, a
Republican, or an Independent) in the 2000, 2004, and 2008 elections. In
2000, Johnston, Hagen, and Jamieson found,
as forces favored Al Gore, the intensity of identification
strengthened among Democrats and weakened among Republicans, and
the Democrats share among Independent "leaners" grew and Republican
share shrank. The opposite happened when forces favored Bush. At
the end of the campaign, identification with each party
intensified. (2004, 42-43n.9)
In subsequent years (Winneg and Jamieson 2005, 2010), we also noted
some variability during the course of the general election campaign and
greater stability in the weeks leading to the elections. In 2004, the
gap between self-identified party affiliation narrowed, but in 2008 the
tracking indicated a net Democratic advantage, a finding that also
showed up in the exit polls (CNN 2008). In this article, we analyze the
extent of variability in self-identified party affiliation in the 2012
U.S. presidential general election campaign. We then compare the
aggregate 2012 self-identified party affiliation with the prior three
U.S. presidential general elections--2008, 2004, and 2000.
Research Questions and Methodology
To track campaign knowledge and learning during the final two
months of the presidential campaign, the 2012 APPC Institutions of
Democracy (IOD) project conducted a six-wave telephone panel survey that
contained both discrete cross-section samples and multiple wave panels.
In each wave, party affiliation was collected via self-report. This
article utilizes the cross-sectional data from this study to compare the
first four waves of discrete cross-section samples to prior election
years' comparable dates. We use these data to examine the extent of
variability of self-identified party affiliation during the latter
stages of the campaign. Looking first at 2012, we address the following
question:
RQ1: Were there aggregate shifts in self-identified party
affiliation during the final two months of the campaign and immediately
following the 2012 election?
Since nearly all the media and dollars campaigns spend are aimed at
the battleground states, these audiences receive more targeted appeals,
voter contact, and local news coverage. Those living outside those
states do not receive the same level of campaign information either from
the campaigns or through the local news media. Therefore, in this
research we ask the following:
RQ2: Were shifts in self-identified party affiliation similar or
different in battleground and nonbattleground states in 2012?
In 2008, 2004, and 2000 our research tracked change over time using
the rolling-cross-sectional data across the entire campaign. Given the
constraints of the 2012 data set, and the unanswered question of the
extent of shifting in self-identified party affiliation during the final
months of the campaign, we ask the following:
RQ3: How different or similar are the aggregate shifts in
self-identified party identification during comparable periods leading
up to the election in 2000, 2004, 2008, and 2012?
While in the past, we restricted our analysis to the root question,
in this article we extend our analysis to the "leaners" in
order to better understand the direction in which they are heading,
especially in light of the increase in their numbers noted in prior
research (Pew 2009).
RQ4: In what direction do "leaners," those
"Independents" who say they lean toward one party or the
other, trend over time?
Our research builds upon the work we did in analyzing party
identification in the presidential election from 2000 through 2008 but
goes beyond the extant research by examining shifts in the final weeks
of the campaign. The data are from the APPC's IOD 2012 Political
Knowledge Survey, a six-wave national cross-sectional telephone survey
of U.S. adults, 18 years or older, and comparable numbers of U.S. adults
drawn from cross-sections from similar dates drawn in the 2000, 2004,
and 2008 NAES. In this article, we include only the first five waves of
the 2012 study and the first four waves when comparing to prior
years' elections. The five-wave sample includes a postelection
cross-section, and the four wave sample includes interviewing which
ended about a week before Election Day. Wave 6 was conducted much later,
in December.
In 2012, under contract to APPC, Social Science Research Solutions
(SSRS) completed 6,556 cross-sectional interviews for waves one through
five (5,340 for waves one through four). Each cross-section consisted of
randomly selected adults contacted through random-digit dialing (RDD) of
landlines and cell phones. (1) In each wave, we oversampled households
in battleground states (2), since battleground states were exposed to
most of the presidential campaign attention. In waves 2 through 5, we
included a mix of fresh sample randomly drawn using RDD and recontacts
with prior wave completes, (3) but as stated, we focus here only on the
fresh sample from each wave. For 2000 through 2008, the NAES utilized
the rolling cross-sectional design (Romer et al. 2004, 2006). The
surveys are conducted among adults in the United States. (4) Table 1
shows the breakdown of randomly selected completes by wave, date and
election year.
Dependent Variables
Self-Identified Party Affiliation, In each survey, we asked
respondents their party affiliation: Generally speaking, do you usually
think of yourself as a Republican, a Democrat, an Independent, or
something else? Interviewers recorded verbatim responses from those who
said "something else." The question wording is the same as
used in the ANES (National Election Studies; Center for Political
Studies; University of Michigan 2003). Interviewers asked the party
identification question prior to the vote intention question, with
several questions in between. However, research suggests that question
order has no significant impact on party identification (McAllister and
Wattenberg 1995). In our analysis for this article we focus on party
identification among adults and report aggregate shifts across the
campaign.
Self-Identified "Leaners." Also in each survey, we asked
respondents who indicated in the root question that they were
"Independents" to identify which party they leaned toward: Do
you think of yourself as closer to (Read) the Republican or Democratic
Party? The question wording is also the same as that used in the ANES
studies.
Battleground State Variables. In the 2012 survey, we oversampled
battleground states--those states determined to be the most competitive
by the campaigns and the news media. In each wave, we collected roughly
50% of our completed surveys from battleground state respondents. When
we analyzed national party shifts in the aggregate and battleground
versus nonbattleground, we weighted the battleground to its correct
proportion according to the U.S. Census (roughly 40%), along with key
demographic variables, gender, race, age, and education. In 2000, 2004,
and 2008, battleground states were not oversampled in NAES, so when we
analyze the battleground and nonbattleground states, there was no need
to adjust their weights up or down to match the national census
proportions. However, battleground versus nonbattleground state data in
those years are weighted as part of the national sample.
Analytical Methods
To answer the research questions related to aggregate level
changes, we conducted a descriptive analysis of the individual
cross-sections within 2012. In order to analyze election to election
changes, we combined each year's collection periods to create an
aggregate party affiliation (data collection periods 1-4 for 2000, 2004,
2008, and 2012, respectively, were combined into a single point for each
year). In order to test differences from election to election, we used a
z-score calculation comparing proportions from independent samples,
taking into account the weights and the design effects resulting from
weighting. Additionally, within the 2012 election year, we employed
backward difference regression coding. Using this technique, we are able
to compare the mean (party proportion) of one level (wave) of party
identification to the mean of party identification for the prior level
(wave).
Results
The results we present regarding party identification in our first
three research questions do not include "leaners":
Independents who are asked if they lean more to the Democratic or
Republican parties. We present only the root question in our findings
because the inclusion of "leaners" tends to contribute to
greater variability in party identification (Converse 1966; Miller and
Shanks 1996). However, in our final research question, we do include
"leaners."
Self-Identified Party Affiliation: Final Two Months of the 2012
Campaign
In 2012, from wave to wave, descriptive analysis shows the
cross-sections remained generally stable in the final two months of the
2012 campaign, with Independents tracking along with Democrats across
the waves. Democrats retained their advantage over Republicans, and by
the end of October, there was a slight, but not significant decline in
self-identified Republicans from wave 3 to 4 (See Table 2). It is
interesting to note that, according to the network exit polls, Romney
defeated Obama 50% to 45% among self-identified Independents (CNN 2012).
The proportions are generally, but not exactly, in line with
comparable findings from Pew's surveys conducted close to the same
time periods. In the Pew data, the proportion of self-identified
Democrats remained relatively flat while self-identified Independents
jumped slightly with an accompanying dip among self-identified
Republicans during a short three-day field period in mid-October (See
Figures 1 and 2).
2012 Party Identification Shifts and the Battleground States
Within the battleground states, which were exposed to maximum
campaign media and news coverage, there was a significant shift in
self-identified Democratic affiliation from the first to the second wave
in 2012. Self-identified Democrats increased in the aggregate from 27.4%
to 35.6% (z = -2.61, p < .01). Other shifts were not significant (See
Figure 3).
In the states not considered to be in the battleground, where
campaign-related media and communication were limited in the main to
national cable and network news coverage and the presidential debates,
results showed no significant variation in self-identified party
affiliation in the generally unaffected areas. The variation that
occurred was not statistically significant.
2012 Aggregate Shifts in Self-identified Party Compared to 2000,
2004, and 2008
Since 2000, the United States has elected a Republican to two terms
(2000 and 2004), and a Democrat to two terms (2008 and 2012). Where the
outcome of the two earlier elections was either contested (2000) or
close in the national popular vote, the last two were not. Comparing
cross-sections collected during comparable time periods in 2008, 2004,
and 2000, we note a significant pattern of variability in
self-identified party affiliation. Each data point in Figure 4
represents the combined waves for each election year of self-identified
party affiliation. Rather than a fixed level of party affiliation across
campaigns, our data show a significant increase among self-reported
Independents along with a significant decline among self-identified
Republicans. If one observes trend line from 2000 to 2012 in Figure 4,
the data indicate volatility in self-identified Independents. The
proportion calling themselves Independents in 2000 matches the
proportion self-identifying as such in 2008 (28.4% compared with 28.9%).
In 2004, the proportion of self-identified Independents was below that
of Republicans, but surpassed those affiliated with the GOP by 2008
continuing on to equal the self-identifying Democrats by 2012 during the
aggregate time periods in which we were interviewing. The Republican
Party suffers most from the increase in self-identified Independents.
The proportion of self-identifying Democrats remains above 30%, but
lower than its 2008 level, and the proportion of Republicans declines
from nearly 30% in 2004 to the low-twenties by the end of the 2012
campaign. The gaps between the Republicans and the Democrats and
Republicans and the Independents are both about 9%.
Combined wave data from battleground and nonbattleground states
follow a pattern similar to that of the broader national data. Figures 5
and 6 indicate a steady increase in self-identified Independents in both
designations. Drops in Republican self-identification in the
battleground states are significant between 2008 and 2012. Outside of
the battleground the drop in self-identified Republican affiliation is
nonsignificant between the two elections, but the dip in Democrats is
significant (See Figures 5 and 6).
Direction of "Leaners" 2000-12
Since the trend from 2000 to 2012 shows a significant increase
among Independents, we analyzed the direction of "leaners." In
2000, 2004, and 2008, "leaners" were significantly more likely
to say they leaned toward the Democratic rather than the Republican
Party. But in 2012, the trend seemed to be shifting. In 2000,
"leaned" Democrats held a nearly four-point advantage over
leaned Republicans, one that widened in 2004 to just under nine points.
In 2008, "leaned" Democrats held that nine-point edge, but in
2012, the advantage dropped to an insignificant 1.5 points (See Figure
7).
Conclusion
In 2012, the Democratic edge over Republicans in self-identified
party affiliation remained strong. However, the proportion of adults
identifying themselves as Independents unaffiliated with any party has
grown and is perhaps ready to surpass that of both the major parties in
affiliation, or lack thereof. The data across these four elections
suggest that the Republican Party has suffered the most because of the
increase in unaffiliated Independents. Though it is not in as strong a
position as it was in 2008, the Democratic Party has held its ground.
However, "leaners" in 2012 are just as likely to align with
Republicans as they are with Democrats, a closing of the
"leaner" gap, suggesting a shift in the trend.
The original voting research, especially the Elmira study of the
1948 election, recognized that party affiliation varied across campaigns
(Berelson, Lazarsfeld, and McPhee 1954). Subsequent research found that
party affiliation was a stable force through the 1950s and 1960
presidential election. A major party realignment occurred in the 1960s,
mostly in the South, followed by a dealignment in the 1970s, especially
among post-New Deal voters (Miller and Shanks 1996). The 1980s brought
another realignment, again generationally, as New Deal Democrats began
shifting to the Republican Party.
Miller and Shanks identify a "funnel of causality" with a
"host of historic causes converging, in the passage of time, to
produce the vote at the small end of the funnel" (1996, 9). Factors
explaining partisan shifts and vote choice include communication effects
as well as other factors. Party affiliation, since 2000, was not an
"unmoved mover" (Johnston 2006) but shifted over time away
from the two major parties to an increasing growth in identification
with Independents, despite a still strong advantage for the Democratic
Party over the Republican Party. The increase in the proportion of the
electorate made up of Independents and the decrease and the closing of
the leaner gap in 2012 points to the need for further research into the
causes of these shifts.
Appendix
Methodology: 2012 Annenberg Institutions of Democracy (IOD) Survey
The Annenberg Public Policy Center of the University of
Pennsylvania (APPC) contracted with Social Science Research Solutions
(SSRS) to field a sequence of six surveys leading up to, and following,
the 2012 general election campaign. The surveys were designed to measure
Americans' attitudes and media behaviors surrounding the election,
and to track changes in attitude and opinion through repeated interviews
with the same respondents over the course of the campaign. To that end,
the surveys included both interviews with new respondents
("fresh" sample) each wave of interviewing, as well as
callbacks (recontacts") to respondents who participated in previous
waves.
SSRS employed a dual-frame, landline and cell phone sampling design
to account for the growing number of households in the United States
that are cell phone only. In total, 13,978 interviews were completed
across all six waves, with 7,788 unique respondents completing an
interview in at least one wave. SSRS employed a dual-frame, landline and
cell phone sampling design. Of these, 9,482 interviews were completed by
dialing landline phones and 4,496 with respondents on their cell phones.
For purposes of this article, we analyzed the first five fresh waves
(6,556 interviews).
The response rates for each wave of unique cross-section sample
were calculated using the standard AAPOR RR3 formula: wave 1 = 12%, wave
2 = 14%, wave 3 = 13%, wave 4 = 11%, and wave 5 = 17%.
Methodology: National Annenberg Election Survey
The National Annenberg Election Survey (NAES) was conducted in
2000, 2004, and 2008 presidential election seasons by the Annenberg
Public Policy Center of the University of Pennsylvania. The 2008
National Annenberg Election Survey is based on a telephone and online
survey. Telephone interviews began on December 17, 2007, and concluded
on November 3, 2008. For this article we focused on the telephone
portion only.
The 2004 National Annenberg Election Survey is based on telephone
interviews which began October 7, 2003, and concluded on November 16,
2004. The 2000 National Annenberg Election Survey is based on telephone
interviews which began on December 14, 1999, and concluded on January
19, 2001.
In each of the NAES studies, the sample of telephone exchanges
called was randomly selected by a computer from a complete list of
thousands of active residential exchanges across the country. Within
each exchange, random digits were added to form a complete telephone
number, thus permitting access to both listed and unlisted numbers.
Within each household, one adult was designated by a random procedure to
be the respondent for the survey. The interviewing was conducted by
Schulman, Ronca, Bucuvalas, Inc. The results have been weighted to take
account of household size and number of telephone lines into the
residence and to adjust for variation in the sample relating to
geographic region, sex, race, age and education.
The 2008 NAES telephone survey interviewing reached 55,590 adults.
The 2004 NAES was conducted from October 7, 2003, through November 16,
2004, reaching 77,993 adults in the U.S. The 2000 NAES was conducted
from December 14, 1999 to January 19, 2001, among 58,373 adults. The
response rate for the 2008 telephone survey was 20%, 21% for the 2004
NAES, and 25% for the 2000 NAES. All response rates were calculated
based on guidelines set forth by the American Association for Public
Opinion research.
In theory, in 19 out of 20 cases, results for all adults included
in the NAES portion of this study will differ by just over one
percentage point up or down (+/-1.11 for 2000, +/-1.03 for 2004, +/-1.25
for 2008), from what would have been obtained by interviewing all
American adults. For smaller subgroups the margin of sampling error
would be higher.
In addition to sampling error, the practical difficulties of
conducting any survey of public opinion may introduce other sources of
error into the poll. Variations in the wording and order of questions,
for example, may lead to somewhat different results.
References
Bafumi, Joseph, and Robert Y. Shapiro. 2009. "A New Partisan
Voter." The Journal of Politics 71 (1): 1-24.
Bartels, Larry. 2000. "Partisanship and Voting Behavior.
1952-1996." American Journal of Political Science 44 (1): 35-50.
Belknap, George, and Angus Campbell. 1951. "Political Party
Identification and Attitudes Toward Foreign Policy." Public Opinion
Quarterly 15 (4): 601-23.
Berelson, Bernard R., Paul E Lazarsfeld, and William N. McPhee.
1954. Voting. Chicago: University of Chicago Press.
Campbell, Angus, Philip E. Converse, Warren E. Miller, and Donald
Stokes. 1960. The American Voter. New York: Wiley.
Cable News Network. 2008. Election Exit Poll Results.
http://www.cnn.com/ELECTION/2008/ results/polls/#USP00pl (accessed
September 10, 2012).
--. 2012. Election Exit Poll Results.
http://www.cnn.com/election/2012/results/race/president (accessed
September 10, 2012).
Converse, Philip E. 1966. "The Concept of a Normal Vote."
In Elections and The Political Order eds. Campbell, Angus, Philip E.
Converse, Warren E. Miller, and Donald Stokes. New York: Wiley, 9-39.
Franklin, Charles H., and John E. Jackson. 1983. "The Dynamics
of Party Identification." American Political Science Review 77 (4):
957-73.
Green, Donald P., and Bradley Palmquist. 1990. "Of Artifacts
and Partisan Instability." American Journal of Political Science 34
(3): 872-902.
Green, Donald P., and Bradley Palmquist. 1994. "How Stable is
Party Identification." Political Behavior 16 (4): 437-66.
Johnston, Richard. 2006. "Party Identification: Unmoved Mover
or Sum of Preferences." Annual Review of Political Science
9:329-51.
Johnston, Richard, Michael Hagen, and Kathleen Hall Jamieson. 2004.
The 2000 Election and the Foundations of Party Politics. Cambridge, UK:
Cambridge University Press.
Lazarsfeld, Paul E, Bernard Berelson, and Hazel Gaudet. 1944. The
People's Choice: How the Voter Makes Up His Mind in a Presidential
Campaign. New York: Columbia University Press.
Markus, Gregory B., and Philip E. Converse. 1979. "A Dynamic
Simultaneous Equation Model of Electoral Choice." American
Political Science Review 73 (4): 1055-70.
McAllister, Ian, and Martin P. Wattenberg. 1995. "Measuring
Levels of Party Identification: Does Question Order Matter?" Public
Opinion Quarterly 59 (2): 259-68.
Miller, Warren E., and J Merrill Shanks. 1996. The New American
Voter. Cambridge, MA: Harvard University Press.
National Election Studies; Center for Political Studies; University
of Michigan. 2003. Party identification 7-point scale 1952-2002.
http://www.umich.edu/-nes/nesguide/toptable/tab2a_l.htm (accessed April
5, 2005).
Nie, Norman H., Sidney Verba, and John R. Petrocik. 1979. The
Changing American Voter. Cambridge, MA: Harvard University Press.
Pew Research Center for the People and the Press. 2009.
"Independents Take Center Stage in Obama Era: Trends in Political
Values And Core Attitudes: 1987-2009."
http://people-press.org/report/ 517/political-values-and-core-attitudes
(accessed May 21, 2009).
Romer, Daniel, Kate Kenski, Paul Waldman, Chris Adasiewicz, and
Kathleen Hall Jamieson. 2004. Capturing Campaign Dynamics: The National
Annenberg Election Survey. Oxford, UK: Oxford University Press.
Romer, Daniel, Kate Kenski, Kenneth Winneg, Chris Adasiewicz, and
Kathleen Hall Jamieson. 2006. Capturing Campaign Dynamics 2000 &
2004: The National Annenberg Election Survey. Philadelphia, PA: Penn
Press.
Schaffner, Brian, Matthew J. Streb, and Gerald Wright. 2001.
"Teams without Uniforms: The Nonpartisan Ballot in State and Local
Elections." Political Research Quarterly 54 (1): 7-30.
Winneg, Kenneth, and Kathleen Hall Jamieson. 2005. "Elections:
Party Identification in the 2004 Election." Presidential Studies
Quarterly 35 (3): 576-89.
--. 2010. "Party Identification in the 2008 Presidential
Election." Presidential Studies Quarterly 40 (2): 247-63.
KENNETH M. WINNEG
KATHLEEN HALL JAMIESON
BRUCE W. HARDY
University of Pennsylvania
(1.) In each wave of new sample the approximate proportion of
landlines to cell phone completes was roughly 65% to 35%.
(2.) 2012 Battleground states were defined as Colorado, Florida,
Iowa, Michigan, Nevada, New Hampshire, North Carolina, Ohio,
Pennsylvania, Virginia, and Wisconsin. 2008 Battleground states were
defined as Colorado, Florida, Indiana, Iowa, Michigan, Minnesota,
Missouri, Montana, Nevada, New Hampshire, New Mexico, North Carolina,
Ohio, Pennsylvania, Virginia, Wisconsin. 2004 Battleground states were
defined as Arizona, Colorado, Florida, Iowa, Maine, Michigan, Minnesota,
Missouri, Nevada, New Hampshire, New Mexico, Ohio, Oregon, Pennsylvania,
Tennessee, Washington, West Virginia, and Wisconsin. 2000 Battleground
states were defined as, Arkansas, Arizona, Florida, Iowa, Louisiana,
Maine, Michigan, Minnesota, Missouri, New Hampshire, New Mexico, Nevada,
Ohio, Oregon, Pennsylvania, Tennessee, Vermont, Washington, West
Virginia, and Wisconsin.
(3.) All interviews were conducted by live interviewers using
computer assisted telephone interviewing (CATI) software. In households
reached by landline, an adult was selected at random based on age (or
most recent birthday). The questionnaire was delivered in Spanish for
those who selected that option. Nonresponsive phone numbers were
contacted up to six times, and in cases where initial attempts were met
with soft refusal (i.e., abrupt hang-ups), refusal-conversion attempts
were made. Accounting for design effects produced by survey weights, the
margin of error is +/-3.2% for wave 1, +/-3.5% for wave 2, +/-3.7% for
wave 3, +/-3.6% for wave 4, and +/-3.7% for wave 5. See Appendix for
detailed methodology.
(4.) In 2008 National Annenberg Election Survey (NAES) interviewed
55,590 adults in the United States by telephone from December 17, 2007,
to November 3, 2008. The 2004 NAES survey was conducted by telephone
from October 7, 2003, through November 16, 2004, among 77,993 U.S.
adults. In 2000, NAES interviewed 58,373 adults from December 14, 1999,
and January 19, 2001. The margins of error are reported in Table 2, and
response rates are reported in the Appendix.
Kenneth M. Winneg is the Managing Director of Survey Research at
the Annenberg Public Policy Center of the University of Pennsylvania.
His work focuses on political campaigns, political advertising,
knowledge acquisition, survey research, and political participation.
Kathleen Hall Jamieson is the Elizabeth Ware Packard Professor of
Communication at the Annenberg School for Communication and Walter and
Leonore Annenberg Director of the Annenberg Public Policy Center at the
University of Pennsylvania.
Bruce W. Hardy is a senior research analyst at the Annenberg Public
Policy Center of the University of Pennsylvania. He is the coauthor of
The Obama Victory: How Media, Money, and Messages Shaped the 2008
Election.
TABLE 1
2000-12 Distribution of Interviews by Waves *
2012 2008 2004 2000
(MoE) ** (MoE) (MoE) (MoE)
Wave 1: 9/13-18 1,522 1,422 1,903 1,855
(+/-3.2%) (+/-3.5) (+/-2.70) (+/-2.74)
Wave 2: 9/27-10/3 1,337 1,742 2,436 2,219
(+/-3.5%) (+/-2.96) (+/-2.28) (+/-2.47)
Wave 3: 10/17-22 1,233 1,491 2,067 1,818
(+/-3.7%) (+/-3.4) (+/-2.54) (+/-2.78)
Wave 4: 10/24-29 1,248 1,498 2,587 1,927
(+/-3.6%) (+/-3.4) (+/-2.28) (+/-2.70)
Wave 5: 11/7-18 1,216 NA *** NA *** NA ***
W-3.7%)
TOTAL 6,556 * 6,173 8,993 7,819
(+/-1.6) (+/-1.68) (+/-1.14) (+/-1.23)
* Waves 1-4 in 2012 yielded a sample size of 5,340
and a margin of error of +/-1.7.
** Margin of Error noted.
*** No equivalent cross-section interviews were conducted during
this time period in 2008. There was a cross-section conducted in
the weeks following 2004 and the disputed 2000 election but we
have chosen not to include in our analysis to maintain consistency.
TABLE 2
National Self-Reported Party Affiliation--2012 Institutions
of Democracy Knowledge Survey
Q: Generally speaking, do you usually think of yourself as
a Republican, a Democrat, an Independent or something else?
Wave 1 Wave 2 Wave 3
9/13/18 9/27-10/3 10117-22
(n = 1,522) (n = 1,337) (n = 1,233)
PARTY Identification % % %
Republican 23.6 23.0 25.0
Democrat 31.2 32.6 32.7
Independent 31.1 33.2 33.1
Other 9.2 7.7 4.9
Don't know/NA 4.8 * 2.9
Wave 4 Wave s
10/24/29 11/7-118
(n = 1,248) (n = 1,216)
PARTY Identification % %
Republican 21.6 22.1
Democrat 33.1 33.2
Independent 33.5 30.9
Other 9.1 11.2
Don't know/NA 2.7 2.6
Note: Independents are respondents who indicate they
are "Independents." This number excludes those who give
a response of "something else," "don't know," or refused to answer.
* Less than half of 1%.
FIGURE 1. Self-Identified Party Affiliation Annenberg Institutions of
Democracy Survey during Final Weeks of 2012 Election.
Q: Generally speaking, do you usually think of yourself as a
Republican, a Democrat, an Independent or something else?
Republican Democrat Independent
9/13-18 23.7% 31.1% 31.2%
(n = 1,522)
9/27-10/3 23.0% 32.5% 33.1%
(n = 1,337)
10/17-22 25.1% 32.7% 33.1%
(n = 1,233)
10/24-29 21.7% 33.1% 33.7%
(n = 1,248)
Note: Table made from line graph.
FIGURE 2. Pew Research * Self-Identified Party Affiliation during
Final Weeks of 2012 Election.
Republican Democrat Independent
9/12-16 24.0% 35.0% 36.0%
(n = 1,806)
10/4-7 27.0% 31.0% 36.0%
(n = 1,511)
10/12-14 21.0% 33.0% 40.0%
(n = 1,006)
10/24-28 28.0% 33.0% 33.0%
(n = 2,008)
10/31-11/3 26.0% 34.0% 34.0%
(n = 2,008)
* All surveys conducted by telephone for the Pew Research Center for
the People and the Press (2009).
Note: Table made from line graph.
FIGURE 3. 2012 IOD Battleground States Wave by Wave Self-Identified
Party Affiliation.
Q. Generally speaking, do you usually think of yourself as a
Republican, a Democrat, an Independent or something else?
Republican Democrat Independent
9/13-18 25.8% 27.4% 33.2%
(n = 801)
9/27-10/3 23.7% 35.6% * 28.2%
(n = 676)
10/17-10/22 22.9% 35.4% 32.9%
(n = 585)
10/24-10/29 21.2% 32.3% 34.9%
(n = 635)
* p < .05 change in reported party affiliation from 9/13-9/18
to 9/27-10/3.
Note: Table made from line graph.
FIGURE 4. Self-Identified Party Affiliation from 2000 to 2012. (Each
Year Represents Party Affiliation from the 4 Waves Combined.) (^)
Republican Democrat Independent
2000 25.6% 31.9% 28.4%
(n = 7,819)
2004 28.7% 35.0% *** 25.1% ***
(n = 8,993)
2008 26.2% *** 35.9% 28.9% ***
(n = 6,173)
2012 23.3% *** 32.3% *** 32.7% ***
(n = 5,340)
*** p <.001 Significant differences compared to the prior election
year within party.
(^) The election years represent combined results of the data
collected during the four time periods in those years: 9/13-9/18;
9/27-10/3; 10/17-10/22; 10/24-29.
Note: Table made from line graph.
FIGURE 5: Self-Identified Party Affiliation from 2000 to 2012
Battleground States. (Each Year Represents Party Affiliation from the
4 Waves Combined.) (^)
Republican Democrat Independent
2000 24.1% 32.8% 29.5%
(n = 3,025)
2004 29.8% ** 34.3% 25.9% *
(n = 8,993)
2008 28.6% 34.8% 29.4% *
(n = 6,173)
2012 23.6% ** 32.9% 31.7%
(n = 5,340)
** p < .01 * p < .05 Significant differences compared to the prior
election year within party.
(^) The election years represent combined results of the data
collected during the four time periods in those years: 9/13-9/18;
9/27-10/3; 10/17-10/22; 10/24-29.
Note: Table made from line graph.
FIGURE 6. Self-Identified Party Affiliation from 2000 to 2012 States
Outside of Battleground. (Each Year Represents Party Affiliation from
the 4 waves combined.) (^)
Republican Democrat Independent
2000 26.6% 31.8% 27.8%
(n = 4,794)
2004 28.1% 35.4% *** 24.6% **
(n = 8,993)
2008 24.6% ** 36.7% 28.6% **
(n = 6,173)
2012 23.3% 32.1% ** 33.1% **
(n = 5,340)
*** p < .001 ** p < .01 * p < .05 Significant differences compared to
the prior election year within party.
(^) The election years represent combined results of the data
collected during the four time periods in those years: 9/13-9/18;
9/27-10/3; 10/17-10/22; 10/24-29.
FIGURE 7. Party Leaners: 2000-12. (^) (Asked Among "Independents"
In Root Party Question.)
Q: Do you think of yourself as closer to (Read) the Republican
or Democratic party?
Republican Democrat Independent
2000 33.7% 37.4% 23.6%
(n=2,223)
2004 32.5% 41.2% 21.3%
(n=2,258)
2008 31.2% 40.2% 23.8%
(n=1,783)
2012 36.8% 38.3% 21.4%
(n=2,220)
* p<.05 Difference between Leaned Democrat and Leaned Republican
within election year.
(^) The election years represent combined results of the data
collected during the four time periods in those years: 9/13-9/18;
9/27-10/3; 10/17-10/22; 10/24-29.
Note: Table made from bar graph.
