2012 and before* - university of akron › dotasset › 625316c8-2ba8-4168-9517-6a97… · 2012...
TRANSCRIPT
1
The Ground Game from the Voter’s Perspective:
2012 and Before*
by
Paul A. Beck
The Ohio State University
and
Erik Heidemann
(late of) Kent State University
Paper presented at the conference on
The State of the Parties: 2012 and Beyond
Bliss Institute of Applied Politics
The University of Akron
November 7-8, 2013
*We gratefully acknowledge the support for this project from Dick Gunther, co-PI for the 2004 and 2012
survey; Chip Eveland, Kelly Garrett, and Erik Nisbet, co-PIs of the 2012 survey; GfK/Knowledge
Networks for conducting the two surveys; and The Ohio State University for its help in financing the
surveys.
2
Direct contacts with voters, the so-called “ground game,” have been an important
focus of political campaigns since the beginning of democratic politics in the United
States. Historically, the ground game was the province of the local party organizations,
but their grass roots activity declined as those local organizations lost their “patronage
armies” of campaign workers. After a period where the ground game received little
attention, its tempo came to be driven by the presidential campaigns by the time of the
hotly contested election of 2000 and has continued at a higher pace thereafter. As the
competitive balance between the parties has tightened and the parties have polarized, it is
understandable that more attention has been devoted to mobilizing the party’s base and to
persuading uncommitted voters in order to gain a critical edge over the opposition,
especially in the Electoral College battleground states.1
Building upon their organizational success in 2008, the 2012 presidential campaign
of Barack Obama is credited with having achieved unprecedented (at least in modern
times and for presidential contests) effectiveness in its ground game. Both sides attribute
Obama’s victory in 2012, as they did in 2008, at least partially to its ground game
advantages over the Romney campaign (see, inter alia, Balz and Silverman 2013,
Rutenberg 2013, Sides and Vavreck 2013) despite Romney’s increased attention to party
contacts compared to the McCain campaign in 2008.2 By most accounts, the parties
were seen as focused on mobilizing their base in the hope that higher turnout of loyal
partisans would provide the critical margin in close races. The very fact that the Obama
vote was higher than many models, including those of the Republicans, were predicting is
seen as testimony to the Obama edge in the ground game.
This paper examines the ground game of the 2012 presidential campaigns from the
perspective of reports of party contacts by respondents in a national survey of the
American electorate.3 Respondents were asked: “Did representatives of any of the
political parties or presidential candidates contact you during the 2012 campaign?”
Those who reported a contact were then asked to specify which party or candidate they
were representing and how the contact was made – by mail or other printed literature, by
phone, in person, or by email or other electronic means. (See the Appendix for the full
text of the questions.) The paper begins by reporting levels of party contact in 2012,
comparing them by party, type of contact, and whether the state was a presidential
1 See (Beck and Heidemann forthcoming) for an extensive examination of survey reports of party
contacting in presidential campaigns from 1956 to 2012 using the American National Election Studies
(ANES) time series. 2 For careful analysis of the practice and challenges of the ground game, see Issenberg 2012. Also see
Popkin 2012. 3 The survey was conducted from November 7 to November 19, 2012, via the Internet by GfK Knowledge
Networks. Respondents were U.S. citizens drawn from a pre-existing probability-based web panel
designed to be representative of the adult citizen population of the U.S. The panel was created from a
representative sample of adults without regard to whether they used the Internet. Those who could not
respond to interviews via the Internet were equipped with a device that enabled them to participate via a
netbook computer that GfK provided. The resulting respondents for our survey were then weighted to
match key demographic characteristics of the adult population. For more information on the GfK methods,
see Dennis (2001), Chang and Krosnick (2009), and Knowledge Panel Design Summary (2012). For other
reports on GfK/KN methods see marketing.gfkamerica.com/knowledgenetworks/ganp/reviewer-info.html.
3
battleground. It then analyzes who was contacted in 2012 under each of these conditions.
To gain some historical perspective on the 2012 ground game, we then compare these
results with those from a similar survey in 2004.4
The 2004 and 2012 U.S. surveys that we employ were part of a cross-national effort
to measure party contacting through the auspices of the Comparative National Election
Project (CNEP). To date, with more surveys planned, CNEP contains comparable
measures of party contacts in national elections for over 20 elections around the
democratic world, most of which also measured different types of contacts.
Respondent survey reports may be the most reliable way to determine how party
contacts reached the eligible electorate. Party and candidate organizations often claim
great success in contacting potential voters, but it is difficult (for them and for us) to
disentangle the effort in the aggregate, which sometimes is accompanied by extravagant
claims for effectiveness, from the actual contacts on the ground. There also is a tendency
to attribute greater ground game effectiveness to the party or candidate who has won the
election and their claims.
Yet, survey reports contain their own frailties as estimates of party contacts. People
may have difficulty remembering party contacts over the course of a long campaign, and
especially differentiating among different types of contact. Our question asks for both
party and candidate organization contacts in the presidential campaign,5 but respondents
may have trouble identifying the source of the contact. Was it on behalf of the
presidential candidate or a candidate for some other office? Was it from a party, a
candidate, or some independent group?6 Appreciating that ordinary citizens might not be
aware of whether it is a candidate or party representative making the contact (and even
the canvassers may not know), we have combined the two into one question. Finally, how
can a national survey of fewer than 1300 people estimate party contacts that are guided
by micro-targeting strategies that segment the electorate into a multitude of groups?
Where the group is sizable (e.g., African-Americans or Hispanics), we can generate
reliable estimates; where it is a small group (e.g., young college students), by contrast, we
cannot. With these reservations in mind, we submit that the survey reports can shed
4 The 2004 survey was conducted via the Internet by the same firm, then known as Knowledge Networks,
using essentially the same methods and weights. The full text of the questions is in the Appendix. 5 The reason to ask for both is that campaign finance regulations force separate reporting of spending by the
party organization and the candidate organization and regulate how much money can be raised by each and,
for the party, whether money spent should be counted as a candidate contribution. This treatment of parties
and candidate organizations as independent entities is uniquely American and is designed to capture efforts
on behalf of the presidential nominees by both the party and the candidate’s organization, both of which are
engaged in the ground game. 6 Strictly speaking, under federal campaign financing laws, the party and the candidate organization operate
separately in their campaign fund-raising and spending. Some coordination under strict guidelines is
allowed between them, but campaign finance regulations necessitate careful segregation of fund-raising and
spending between the two. By contrast, independent groups are prohibited from coordinating their
campaigns efforts with those of either the party or the presidential candidates. Even the so-called candidate
“Super PACs” cannot coordinate with the candidate whom they favor. For more on these relationships
between the parties, the candidate campaign organizations and the independent groups on the eve of the
2012 election, see Malbin et al. 2011 and La Raja 2011.
4
considerable light, and raise interesting questions, about the ground game in 2012 – and,
in comparison with 2004, before.
Ground Game Performance in the 2012 Presidential Campaign
Two claims about the presidential ground games in 2012 have dominated coverage
of the campaigns and accounts of campaign strategies by campaign managers (Balz and
Silverman 2013; Sides and Vavreck 2013). First, the Obama campaign, through its own
organization and the Democratic party, is credited with being more effective at the grass
roots than the Romney campaign. They had many more field offices, especially in the
battleground states. Both sides were devoted to micro-targeting, but the Obama forces
seemed to have an advantage there as well (Rutenberg 2013). When the dust had settled
after election day, many observers credited the Obama campaign’s ground game as
having carried the day for its candidate. Second, the campaigns’ foci in the 2012 ground
game was on mobilizing their support base. His ability to mobilize young voters and
minorities had been seen as a key to Obama’s victory in 2008. Declines in the turnout of
these key Democratic groups were seen as underlying the Republican successes of 2010.
The key question for 2012 was whether the Democrats could replicate its 2008 turnout
levels among its base in face of waning enthusiasm for now-President Obama and
extensive ground game efforts on the other side. Our 2012 survey data can address both
of these claims.
First, as Figure 1 shows, the Obama campaign did enjoy an edge over the Romney
campaign in reports of party contacts, but that edge was slight, just beyond the +/-2.5%
conventional bounds for sampling error. The Obama edge was built on more extensive
personal and electronic (email, twitter, etc.) contacts, with the Romney campaign having
an edge in contacts by mail or through literature. Even though the overall near parity is
somewhat surprising given the “conventional wisdom,” the patterns by type of contact
probably are not. Needing to mobilize a base of young people and disadvantaged
minorities who are commonly less motivated to vote, the Obama efforts understandably
concentrated on face-to-face contacts, shown in carefully-controlled field experiments
(Green and Gerber 2008) to be the most effective among all of these types in mobilizing
support.
The other noteworthy result from Figure 1 is the low percentage of contacts in
person and electronically in comparison by mail/literature and telephone. It is far easier
to distribute literature or to phone potential voters than it is to talk with them directly at
their door or in some more public place. It is undoubtedly easier to contact voters via
email or other electronic means as well. However, this assumes that the recipients of the
contact are Internet, twitter, or smart phone users and the campaign knows their
addresses, which can severely limit their reach. Personal face-to-face contacts are labor
intensive and challenging to canvassers, so it is little wonder that they are relatively rare –
even if they may be more effective. Although electronic and in-person contacts reach
only a small percentage of the electorate, it is worth remembering that, in an electorate of
222 million eligible voters (McDonald 2013) and 130 million presidential voters, even
5
small percentages involve millions of citizens: over 20 million eligible voters contacted
by each party through email/electronic means and 10-15 million in person.
Figure 1
Party Contacting in 2012
US CNEP Survey
Modern presidential campaigns are not really national campaigns, but instead have
devoted their scarce their resources to an increasingly smaller set of “battleground”
states. Figure 2 replicates Figure 1, but this time focusing only on the eleven states
(Colorado, Florida, Iowa, Michigan, Nevada, New Hampshire, North Carolina, Ohio,
Pennsylvania, Virginia, and Wisconsin) that emerged early on as the battlegrounds of the
2012 campaign – the states in which both campaigns invested considerable time, staff,
and money. It clearly shows much more party contacting overall in the battlegrounds
compared with all fifty states. Both parties contacted about 60% of adult citizen
respondents in the battlegrounds, almost double the figure for the non-battleground states.
With one interesting and important exception, each type of contact reached many
more voters in the battlegrounds. The exception is contacts by email or other electronic
messaging, where attention to the battleground states is only slightly (and not
significantly) higher. The reason for this exception, we surmise, is that electronic
messages go out repeatedly to a pre-existing list of party and candidate supporters,
especially those who already have contributed money to the campaigns. Much of this
messaging probably is directed towards fund-raising, which does not play state
“favorites” and indeed probably draws heavily upon partisans from non-competitive party
enclaves where rich veins of potential contributions are to be found (e.g., New York,
California, Texas), as much as mobilization of the base. As we shall see later, there are
other indications that this is a special type of party contact, distinctly different from the
other three.
6
Figure 2
Party Contacting in Battleground States in 2012
US CNEP Survey
These data challenge the claim of an Obama advantage in the ground game,
especially one substantial enough to be credited with a victory built upon a virtual sweep
of the battleground states. Instead, and especially in the battlegrounds, the Romney
campaign and Republican party seemed to dual the Obama campaign and Democratic
party to a draw. The lone exception lies with “in person” contacts, where Obama enjoyed
about a two to one edge percentage wise, most importantly in the battleground states
where the election was to be won or lost. While the edge was based on contacts with
only a small slice of the electorate, it probably was the most consequential of all the
contacts. Green and Gerber’s (2008, especially pp. 43-45 and 139) experimental
evidence shows that, while door-to-door canvassing is the most difficult of campaign
contacts, it is much more cost-effective in mobilizing voters than the other types of
contacts we have measured. Using McDonald’s (2013) figures for the voting eligible
population and our estimates of contacts, we project that the Obama campaign personally
contacted about 7 million more voters than the Romney campaign in all states and about
3.6 million more in the battleground states, which Obama won by a total of 1.6 million
votes.
Personal contacts are especially important for Democratic candidates. The
Democratic base contains numerous potential voters who because of their educational,
income, age, and mobility disadvantages do not habitually participate in elections, even
presidential elections.7 More effort is required, on average, to mobilize them. So, the
Democrats need the advantage in personal contacts that they achieved in 2012. Even if
the Obama forces were successful in personally contacting only 11% of the eligible
electorate in the battleground states, it probably was a vital 11%, producing what might
have been as many as 167,000 extra votes using the Green and Gerber estimates of
7 Age, education, income, and residential stability are key individual-level predictors of turnout in elections
because they affect both the motivation to vote and the cost of voting. Because Americans with pro-
Democratic preferences are on average younger, less educated, lower in income, and more mobile (due to
their age in part), they may be assumed to require more mobilization effort to get them to the polls.
7
effectiveness for the incumbent president, over 10% of his victory margin across these
states.8 Moreover, it is plausible that party contacts of various types were more important
in mobilizing the Democratic base than they were in getting out the vote for Republicans,
especially because Republicans were even more motivated to vote in 2012. Thus, even
the parity in overall contacting that we have reported advantages the Democrats more
than the Republicans.
Ground Game Performance in the 2004 Presidential Campaign
How does ground game activity in 2012 compare with earlier campaigns?
Responses to a question similar to ours in the American National Election Study (ANES)
show that the highest level of party contacting from 1956 to 2012 was reported in 2004.
In that year about 45% of all ANES respondents and well over 50% in the battleground
states said that they had been contacted by representatives from one of the major parties
or its candidates. The highest level of contacting by a single party in that 1956-2012
series came for the Democrats in 2008, in line with the conventional wisdom about the
Obama campaign’s ground game effectiveness that year.9
Unfortunately, the ANES data do not allow us to “drill down” to determine type of
contact, but we can compare our 2012 results with the similar questions asked in our
2004 U.S. CNEP survey. These results, presented in Figure 3 below, show somewhat
more reported contacting in 2004 by both parties than in 2012 with the exception of
more email and electronic contacts.10
In other important respects, though, they echo what
we found for 2012: There was parity between the parties in contacts by mail, by phone,
and (within sampling error) overall. In person contacts also were relatively rare in both
years, but show a Democratic edge. That more contacts were made by email or other
electronic means in 2012 is hardly surprising, but that the increase from 2004 to 2012
was so small comes as a surprise. It appears that the much-touted attention to the
Internet, twitter, and other electronic messaging as a feature of the new campaigning may
be exaggerated. That these electronic means of party contacting have become more
frequent cannot be doubted. But, they appear to reach only a thin slice of the electorate,
barely more than 10% in 2012. It may be a case of much more email/electronic traffic for
those who draw upon it, but only slightly more people using it than before.
8 Green and Gerber (2008, p. 139) estimate that door-to-door canvassing yields one vote per 14 contacts.
The battleground state electorates in 2012 gave 21.4 million votes to Obama. If 2.36 million of them were
contacted (11% x Obama voters) and 1 in 14 of these are added to his vote due to the contact, the result is
167,000 additional Obama votes, or about 0.8% of his vote in the battleground states. Interestingly, Seth
Masket estimates that the Obama gained a 0.8% boost in a county based on his edge in field offices there
(Matthews 2012). 9 For an extensive analysis of the ANES series on party contacting, see Beck and Heidemann
(forthcoming). 10
As the Appendix shows, there were slight differences in the survey questions between 2004 and 2012. In
2012, we broadened the questions about types to include “other printed material” as well as “mail” and
“other electronic messaging” beyond “email.” We doubt that changes in question wording had much effect
on response frequencies, but (if anything) they should have inflated the number of mail/literature and
email/electronic contacts, thereby making the 2004 edge in contacting even more substantial.
8
Figure 3
Party Contacting in 2004
US CNEP Survey
Comparisons between all states and the battleground states in 2004, presented in
Figure 4 below, show considerable parallels with 2012. More respondents reported
contacts by each party in the battlegrounds:11
Roughly 10% more of the electorate was
contacted overall and by mail and telephone in the battlegrounds. Although again
relatively rare, in-person contacts also were more frequent in the 2004 battlegrounds.
Interestingly, they were slightly more frequent in 2004 than they were eight years later in
2012, with growth coming for both Democratic and Republican contacts. As with the
2012 comparisons, reported contacts via email were not significantly different between
battleground and non-battleground states in 2004, again probably because such contacts
are made more in pursuit of campaign donations than votes with already-dependable
party voters.
11
Fifteen states were counted as battleground states in 2004. Nine were the same as in 2012: Colorado,
Florida, Iowa, Michigan, Nevada, New Hampshire, Ohio, Pennsylvania, and Wisconsin. To them were
added North Carolina and Virginia in 2008, then again in 2012. Six states were battlegrounds in 2004, but
not in 2012: Maine, Minnesota, New Mexico, Oregon, and Washington had moved comfortably into the
Democratic “blue state” column. By contrast, West Virginia became a dependable Republican state in
2008 and again in 2012.
9
Figure 4
Party Contacting in Battleground States in 2004
US CNEP Survey
Who Was Contacted via the Ground Game in 2012
It is well known that parties and candidates target specific groups of voters in their
ground game activities. Previous studies of party contacting document who is contacted
from respondent reports in the ANES series (see, inter alia, Rosenstone and Hansen
1993, Wielhouwer 2003, and Gershtenson 2003). They find that contacts are
significantly higher with the party’s own identifiers; habitual voters; older, better
educated, and higher income individuals; union households; and people who are more
socially connected (such as home owners and those who regularly attend church). More
competitive elections, such as living in a battleground state (Gimpel et al. 2007), also
promote contacting.
In identifying variables to employ in determining the targets of party contacting in
2012 (and, below, in 2004), we build upon these studies and add targets that the
campaigns seemed to see as particularly important in the particular years. On the one
hand, parties and candidates seek to mobilize their base – to make sure that they are
maximizing turnout from potential voters who are their loyal supporters. In recent years,
this leads us to expect differences between the Democratic and Republican campaigns in
whom they are contacting. The drop off in turnout among minorities and young people
between 2008 and 2010, for example, is credited to a considerable degree with the
widespread Republican victories across the states in the 2010 contests. The Democratic
ground game that was so noticeably successful in 2008 fell short just two years later, then
seems to have returned to form in 2012. On the other hand, ground game efforts should
be expected to concentrate regularly on contacting the most easily identifiable likely
voters, perhaps with micro-targeting within these groups to make sure that they are
mobilizing their own supporters.
10
We start with a result that does not square with the conventional wisdom that the
2012 ground games were focused primarily on mobilizing the partisan base. It is
reasonable to expect partisans to be contacted much more, maybe almost exclusively
under the mobilization of base strategy, by the campaigns of their parties than by the
opposition. There is even more reason for focus on loyal partisans in the 2000s, when the
percentages of partisans voting for their party’s candidates for president topped 90%.
Yet, as the first two panels of Figure 5 below show (we will consider panels 3 and 4
later), more potential voters were contacted by both parties than by a single party in 2012.
Across all states, 30% of respondents reported contacts from both Democratic and
Republican campaigns versus only 19% from a single party. Many of those contacted,
surprisingly, were partisans of the opposing party: despite the unlikelihood that many of
them would defect to the opposition candidate, 34% of Democrats reported having been
contacted by Republican campaigns, and 36% of Republicans reported being contacted
by Democratic campaigns. These figures rise in the battleground states to a slight
majority (50.1%) reporting contacts from both parties, with 53-54% of partisans having
been contacted by the other party.12
The claim that the 2012 ground game was designed to mobilize the party base is
questionable in view of such figures, especially for the partisans who under even the most
precise micro-targeting should not have been a target of the opposing party. Party
identifications of course are not readily available to ground game canvassers, at least not
before they have collected background information from canvasses of potential voters.
The parties more often must work from official registration and voting rolls, which lack
up-to-date partisan identifications for many registered voters (even in states that record
party registration either through registrant’s selections or party primary participation), so
it is understandable that partisans may not be easy to identify. For whatever reason,
though, these results belie the claim that the 2012 ground game concentrated heavily on
mobilizing the parties’ loyal partisans, thus raising the risk that the parties may have
expended considerable effort in contacting voters unlikely to support them – or, even
worse, in motivating voters to oppose their candidate.
12
Even with the lower levels of overall party contacting reported in the 2012 ANES survey, which asked a
similar question to ours about party contacts, a surprisingly high number of respondents (18%) said that
they had been contacted by both parties compared to only one party (23%). In the battleground states
alone, the percentage contacted by both parties rose to 28% -- the same percentage who reported being
contacted by a single party. The 2012 ANES survey contained both face-to-face and Internet samples.
Higher levels of party contacts were recorded for the Internet sample – 44% vs. 37% in all states and 61%
vs. 48% in the battlegrounds – bringing them close to the 2012 figures from our own U.S. 2012 Internet
sample. Why the Internet surveys recorded higher levels of party contacts, even when weighted to
represent the same demographic mix, is not clear. One hypothesis worth considering is that Internet
respondents (who respond at their own pace rather than in response to an interviewer) may be able to give
more thought to whether they had been contacted by a party, producing more accurate responses. Nor is it
clear why our Internet survey showed greater contacts by both parties than did the ANES Internet survey.
11
Figure 5
Contacts by Party in All States and Battleground States, 2012 and 2004
US CNEP Surveys
Our U.S. CNEP surveys allow us to address the question of who was contacted more
broadly by determining the simple relationships between reported contacts, overall and
by type of contact, and the voter groups that seem most likely to be the targets of the
ground game. For all states, Table 1 below presents the Pearson product-moment
correlations (r’s) that reach significance at the .05 level between the various contact
measures and a series of variables that represent important groups within the electorate,
mostly measured as dummy variables (the presence or absence of a particular
characteristic). Empty cells signify correlations that fell below the standard .05 level of
significance.
In overall reported contacts, presented in columns one and two of the tables that
follow, there are clear similarities between the parties, some of which are unexpected.
Age is positively correlated with contacting for both: the older the voter the more likely
he or she is to report party contacts. Reported contacts by both parties also were more
likely in the battleground states; with those who voted in previous elections; and with
campaign activists. These results are consistent with previous studies, including Beck
and Heidemann (forthcoming) for 2012 using the ANES data. The biggest surprise in
these relationships is that union members appear to have been targeted by the Republican
as well as the Democratic campaigns. The relationships are not strong, but they are
significant, suggesting that both parties see votes to harvested among those who belong to
unions, a fact that union leaders (who typically support Democratic candidates) probably
understand. Another unexpected result in 2012 is that both parties targeted Protestants,
though which particular Protestants surely varies by race. By contrast, contacts by both
parties are less likely for people with no religious affiliation, Hispanics, people claiming
a race other than white, black, or Hispanic, those with little interest in the campaign, and
nonpartisans.
What is most unexpected in these paired negative relationships is that Hispanics
report less party contacting by both parties than do (their dummy-variable opposite) non-
Hispanics. It is understandable that Republican efforts are not directed towards
mobilizing Hispanics, who have heavily favored Democratic candidates in recent
12
elections. But, the negative relationship between reported Democratic contacts and
Hispanics is puzzling in light of the importance of Hispanic voters to the Democratic
coalition and the get-out-the-minority-vote claims that party made in 2012. What can
account for this anomalous result?13
A majority of Hispanics reside in one-party states
like California and Texas, where the motivation to turn out must be generally low. Yet,
when we examine party contacts with Hispanics in the battleground states (see Table 2
below), the surprising negative correlation with Democratic contacting remains. We can
think of several reasons why Hispanics are an especially challenging group for the
campaigns to contact, despite their attractiveness as potential Democratic voters. They
are a younger on average than the general population, and younger people are less
habitual voters and less likely to turn up on official voting rolls. Young people also are
harder to reach because they are much more residentially mobile, which prevents them
from being easily identifiable and contactable. Moreover, the Democratic party and
Obama campaign infrastructures were less deeply rooted in the Hispanic community,
which meant that they had fewer veteran organizers and volunteers to draw upon in their
canvassing efforts than they might want. Unlike African-Americans who were naturally
drawn to vote in record numbers for a Democratic ticket with the first black president at
its top, Hispanics also lacked the lure of a group member to draw them into the active
electorate.
That nonpartisans report less contact than partisans, while not unexpected, warrants
some comment. On the one hand, it is entirely reasonable that the respective parties
would focus more of their canvassing efforts on their own partisans than on nonpartisans.
Nonpartisans lack the predictability of partisans in gauging how they would vote if
mobilized, and campaigns do not want to encourage voters to go to the polls if they might
vote against them. They also turn out at much lower levels than partisans, only in part
because they may be less encouraged by party contacts. On the other hand, in a more or
less partisan-balanced electorate, the vote of nonpartisans often can spell the difference
between winning and losing. If the parties and candidates can identify the nonpartisans
who might be most favorable to them, it is well worth their while to target them. But,
how can they collect that information? The most readily available sources, official voting
and registration records, are of little help in identifying the most receptive targets. Most
nonpartisans do not register as partisans in states where party registration is permitted.
Nor are they likely to participate in party primaries or to turn up on the lists of party
activists or donors to a presidential campaign. Moreover, nonpartisans are
disproportionately young, hence hard to locate or to identify as dependable votes if
mobilized.
13
This result from our CNEP U.S. 2012 survey is echoed by the 2012 election survey of Hispanics
conducted by Latino Decisions, which reported that only 31% were contacted by a party or candidate
during the campaign (Sanchez 2013). Even in the battleground state of Florida, only 37% of Hispanics
surveyed reported having been contacted. Further corroboration of the under-mobilization of Latinos
comes from the national exit polls in 2012: Even with the increasing number of eligible Hispanic voters
between 2008 and 2012, their percentage of the national exit poll “electorate” grew from 9% to only 10%.
13
Table 1
Correlations between Party Contact Measures and Voter Characteristics,
All States, in 2012, US CNEP Surveys
14
Finally, there are voter groups who are differentially contacted by the two parties.
Republicans were much more likely to contact Republicans, not Democrats; Democrats
to contact Democrats, not Republicans. Each party contacted its activists, albeit perhaps
not as consistently as might have been expected. Democrats canvassed blacks, while
Republicans concentrated their attention more on whites. Republican contacts were
focused more on high income voters and home owners, whereas reported Democratic
contacts did not vary by income or ownership. None of these differences are surprising,
as the parties are working on mobilizing their bases.
Table 2 presents the correlations for the same variables in the eleven battleground
states of the 2012 campaign. Most of them (27 of 36) are more substantial (i.e., more
positive or more negative where expected) in the states where the parties and presidential
candidates devoted their resources and efforts. Where income is concerned, restricting
the analysis to the battlegrounds elevates the correlations of Democratic contacts with
lowest income (negative) and highest income (positive) respondents into the range of
significance, making their contacting pattern similar to that of the Republicans. That the
Democrats now contacted the lowest income voters less and the highest income voters
more does not necessarily square with their traditional base, but it probably reflects
attempts to harvest those most likely to vote – and, through micro-targeting, they might
be able to identify the more supportive of the higher income voters. Interestingly, while
most of the correlations are higher in the battlegrounds, the enhancement is not very
substantial (though it would grow if the battleground states were removed from Table 1).
Columns 3-10 of Tables 1 and 2 contain the correlations between the voter
characteristics and the four types of party contacting. A few results here are worth
singling out for special attention. First, mail/print contacts and telephone contacts are
associated with more voter characteristics than personal and email/electronic contacts,
especially in the battleground states, which means that targeting is better defined there.
Second, contacts through mail/literature are more associated with the voter characteristics
in the battleground states than are phone, personal, or (for the Democrats)
email/electronic contacts, which are about equally correlated in battlegrounds and non-
battlegrounds. Third, Democratic personal contacts with both whites and blacks are
much more sharply divergent in the battleground states, producing some of the highest
correlations in the two tables. Fourth, union members are considerably more likely to be
contacted by both parties through mail/literature and phone in the battleground states,
showing that they are important targets for both Democrats and Republicans through
these methods.
15
Table 2
Correlations between Party Contact Measures and Voter Characteristics,
Battleground States in 2012, US CNEP Surveys
16
Including the overall contacts and the four types, 670 correlations were calculated for
Tables 1 and 2. Of this total, only 306 are significant, surprisingly more in non-
battleground than in battleground states and surprisingly fewer overall than one might
expect from voter characteristics that should figure into party contacting strategies.
Moreover, only 6 of 670 reach .30 – all but one of them in battleground states.
Republican activists in battleground states by Republican email/electronic contacts (r =
.38) are the most targeted of all, suggesting that the GOP was especially assiduous in
reaching out to its activist base through these new means of contacting, probably as much
for fund-raising than for mobilization (as suggested by a correlation of .31 for all states).
Personal contacts of blacks by Democrats (r = .37) are a close second, signaling the
extraordinary effort the Obama campaign and Democratic party made to mobilize these
most loyal members of its base. Both parties used telephones to contact people
increasingly by age (r = .30 and .33), doubling down on those already more likely to vote.
Similarly, Republicans distributed more mail/literature to past voters than past non-
voters, substantially exceeding Democratic efforts of that type.
That these relationships are not stronger has implications for what one makes of the
ground war in 2012. First, most of the groups identified in Tables 1-4 are fairly large
groups, more heterogeneous in their likely partisan preferences than parties ideally would
want in targeting them as a whole. Consequently, the campaigns are likely to micro-
target within these groups. Rather than contact all union members, for example, it is
likely that Republicans focus on those who exhibit Republican tendencies, while the
Democrats focus on those more likely to be Democrats. (It is also likely that the
mobilization of union members is left to the unions themselves.) Or, while both parties
target regular voters, they usually do so with more information about how they might
have voted in the past, such as in which primaries they voted, to guide them in contacting
the most responsive people within this important group. Blacks are a counter example –
the most homogeneously Democratic group of all the groups in the tables. Democrats
run little risk of mobilizing opponents by contacting them without more precise targeting.
With this one exception, the limitation of a national sample is that it cannot disaggregate
these groups more than we have to be able to adequately test for micro-targeting.
Second, while we undoubtedly underestimate the precision of party contacting with
our survey data, our results nonetheless challenge many of the claims made about ground
game efforts and effects in the 2012 presidential campaign. While there is ample
evidence that the parties do tend to try to mobilize their base, they also reach out to
potential voters who are not easily identifiable as part of their base. In many instances,
these seems to be contacts of convenience – regular voters, older people more likely to be
at home and at the same place for many years, land-line rather than cell phone
subscribers, partisans who already populate campaign mailing and emailing lists.
Moreover, for all the talk of prodigious and precisely targeted ground war efforts, the
reports of party contacts in our survey suggest that there is probably a substantial
disconnect between plans and execution, just as there is in the “delivery” of so many
other campaign messages. Unlike the placement of a campaign advertisement with a
television station, ground war activities are inherently difficult to focus, requiring
17
substantial planning in their targeting and assiduous follow through by volunteers in the
field. This disconnect may be greatest where the most personal effort is required, in face-
to-face contacts between campaigners and voters, even though these types of contact are
well known to be the most valuable.
Who Was Contacted via the Ground Game in 2004
The availability of our 2004 U.S. CNEP survey provides us with an opportunity to
determine how party and candidate contacting patterns may have changed between that
year’s presidential contest and 2012. Tables 3 for all states and Table 4 for the
battleground states replicate for 2004 the previous Tables 1 and 2, using the same
measures for the most part in the questions asking about the types of contact.14
Several
of these patterns are especially interesting in capturing the differences in the ground game
between these two elections.
In examining the contacting patterns in detail, we first turn as we did for 2012, to the
question of how much the ground game could be said to have concentrated on the party
bases. Primed by memories of the razor-thin margin in the 2000 presidential election and
an electorate that by 2004 seemed almost equally balanced between Democrats and
Republicans, it is hardly surprising that mobilizing the base was a dominant goal in 2004.
The third and fourth panels of Figure 5 above suggest that the 2004 campaign may
have been even more base-oriented than 2012. Contacts from a single party were higher
in 2004, and contacts by both lower: By a margin of 35% to 28%, more people across the
country reported contacts from a single party than from both parties. This edge was
slightly reversed in the battleground states where about 35% were contacted by one party
versus 39% by both, but it still was smaller than in 2012 (when a bare majority was
contacted by both parties). As expected, Republicans were much more likely to receive
only Republican than only Democratic contacts (39% to 6%), and Democrats to receive
only Democratic than only Republican contacts (34% to 5%), margins that were not
significantly different in the battleground states. As before, however, a surprising
number of partisans reported contacts from both parties – a third of Republicans (32%)
and Democrats (32%), and an even higher 45% for Republicans and 43% for Democrats
in the battleground states. Even though the percentage contacted by both parties is still
unexpectedly high in 2004, it does not approach the slight majority in the battlegrounds
who received both Democratic and Republican contacts eight years later. Even if
mobilizing the base appears to have been more pronounced in 2004 than in 2012, the
figures still suggest that this was either far from a singular strategy or a goal that was
very difficult to achieve on the ground, given the imprecision of data identifying voter
potential to support the contacting party.
14
To a focus on mail contacts in 2004 was added “or other printed material” in 2012. To a focus on email
in 2004 was added “or other electronic messaging” in 2012.
18
Table 3
Correlations between Party Contact Measures and Voter Characteristics in 2004,
US CNEP Survey
19
Table 4
Correlations between Party Contact Measures and Voter Characteristics,
Battleground States in 2004, US CNEP Survey+
20
There is good reason to expect some changes in who was contacted between the two
years as a result of changes in the candidates and the strategic environment. In particular,
the presence of Barack Obama at the top of the ticket in 2012 might have led the
Democratic campaign to make greater efforts to mobilize minorities more recently,
especially blacks. While blacks have been the most dependable group in the Democratic
base for years, extraordinary mobilization efforts in the black community were not
featured in discussions of the 2004 election. The correlations support this impression.
Overall Democratic contacts were more correlated with black racial identifications in
2012. And, when we turn to personal contacts, which recorded the highest black-contact
correlations in 2012, the difference between the two years is substantial. Democratic
mail/literature, telephone, and personal contacts with blacks also increased in 2012, rising
from insignificance in 2004 to significance in the battleground states and for all but
mail/literature in all states. Correspondingly, whites were considerably more likely to
have reported Republican contacts in 2012. These results are indicative of ground games
that were more targeted along racial lines in 2012 than they had been in 2004.
We already have seen that the consistently negative correlations suggest that neither
party successfully targeted Hispanics in 2012. This result was especially surprising for
the Democrats given the credit they received for their mobilization efforts in the Hispanic
community. Even more surprising is that Hispanics reported even less party contacts in
all states from the Democrats, and Republicans too, in 2012 than eight years before – and
that there was little difference in contacting of Hispanics between the two years in the
battleground states. This result challenges the conventional wisdom about 2012. Given
the general support for Democrats and antipathy towards Republicans among Hispanics,
the Democrats’ meager harvest of Hispanic votes in 2012 is both surprising, and for them
surely disappointing. The Hispanic vote in 2012 and polls of Hispanics since then all
indicate that the Democrats face golden opportunities to draw Hispanic voters
overwhelmingly into their base, thereby reducing Republican electoral prospects, if only
they could mobilize more of them into the active electorate.
Alternatively, 2004 was seen as year in which the Republicans targeted religious
voters, especially fundamentalist Christians in the battleground states. Several
battleground states had gay marriage issues on the ballot that year, purportedly placed
there to draw social conservatives to the polls. Our data show circumstantial evidence of
success in this effort in the greater contacting of regular churchgoers and Protestants by
Republicans in all and battleground states, as well as in the negative correlations between
GOP contacts and those with no religion or who never attends church. By contrast,
weekly churchgoers and Protestants were less distinctive as recipients of Republican
contacts in 2012.
Two other differences between 2004 and 2012 are of particular note because they
seem to defy expectations. First, in 2004, union members surprisingly were not contacted
more than non-union members by the Democrats, although as might be expected they
were less likely to be contacted by the Republicans. By 2012, union members were more
likely than non-members to be contacted by both parties (in all states and in the
battlegrounds), as the group apparently was seen by both as containing good targets (most
21
likely differentiable micro-targets) for mobilization. Second, despite all of the talk about
a Republican “war on women” and a sizable gender gap in voting, women did not report
more contact than men from the Democrats in 2012, nor did men conversely emerge as
much of a target for the GOP. This was in sharp contrast to 2004 in which women were
more likely than men to report having been contacted by the Democrats. The 2004 result
is understandable, but on the surface the 2012 is not, except that it too may be obscuring
more precise micro-targeting.
There also is good reason to expect consistency between the two years in some of
these relationships. The types of voters who generally are logical targets for party
contacting do not vary much from year to year, so we should expect them to be correlated
with reports of party contacts in both 2004 and 2012. These expectations are generally
supported. In both years, the Democratic and Republican campaigns were more likely to
focus their efforts on older people, the highest income quartile, past voters, and (the
presumably more residentially-identifiable) home owners. Democrats and Democratic
activists were more likely to receive contacts from the Democratic campaigns,
Republicans and their activists more likely from the Republican campaigns. These
consistencies between years generally were accentuated in the battleground states. While
some of the patterns represent a ground game that is directed at the party’s base, others
show that the parties’ efforts go beyond their base to pluck the “low hanging fruit” of
habitual voters, many of whom do not need a push to register and vote.
Finally, our data contain evidence that the battleground states were singled out for
more ground game targeting in 2012 than they had been just eight years before. Earlier,
Figures 1 and 2 showed that in 2012 both Democratic and Republican party contacts were
more frequent in the battleground states than in the comparison group of all states (by
about 20% overall and for the most common types of contact).15
Party contacts were
reported more frequently in the battlegrounds in both years of course, but the differences
in contacts between them and all states were about half the size in 2004 than they reached
in 2012.
Our data suggest that the targeting of specific groups also seems to have been more
precise in 2012. One convenient way to demonstrate this is to compare summary
measures of the predictive powers of the voter characteristic variables in Logistic
regression analyses of the party contact.16
Table 5 shows that for eighteen of the 20
15
It is important to reiterate that the differences between battleground and non-battleground states are even
greater than they appear because battleground state respondents are included in the all-state figures. 16
Logistic regression is designed for cases where the dependent variable is dichotomous, as it is for each of
our (yes or no) party contact variables. The independent variables in the analysis are the voter
characteristics that appear as the row variables in Tables 1 to 4 – except that where categorical variables
have been transformed into dummy variables to represent each category, one of the categories is excluded
for estimation purposes, and the resulting coefficients compare each of the specific included categories
against the excluded (base) category. For example, whites were excluded for the race variable so it became
the base against which blacks, etc., are compared. The choice of excluded or base variable affects the
coefficients, but it has no effect on the overall fit of the model as measured, in our case, by the pseudo R2
statistics.
22
comparisons of two pseudo R2 measures (Cox-Snell and Nagelkerke
17) between all states
and battlegrounds across the ten different party contacting measures, our model predicted
better in 2012 than in 2004. The average R2 values for all states versus battlegrounds
summarize this tendency well: .10 in 2012 vs. .04 in 2004 for the Cox-Snell measure and
.14 vs. .06 for the Nagelkerke measure. Not only has the ground game received more
attention in the recent decade or so (see Beck and Heidemann forthcoming), but it
appears that its targeting may have become more efficient as well.
Table 5
Comparison of 2004 and 2012 Pseudo R2’s
from Logit Regression Analysis of Party Contacts on Voter Characteristics,
US CNEP Surveys
Conclusion
Reports of party contact by respondents in two national surveys have provided a
valuable window through which to view the reach of the presidential ground games in
2012 and, by comparison, 2004. These data record overall contact by the presidential
campaigns and enable us to differentiate among four types of contact – mail/literature,
telephone, personal, and email/electronic. While the samples we rely on are not always
large enough for us to drill down to the voter groups a campaign wants to micro-target,
17
These two pseudo R2’s are the most commonly used in Logit regressions.
23
they do support a number of inferences about the ground games in 2012 and eight years
before.
The results of our analysis both support and challenge the conventional wisdom
about the presidential ground games in 2012 and 2004. Overall, the Obama campaign
enjoyed a significant edge in ground game contacts in 2012 only in personal contacts.
They held this edge in 2004 as well, with slight but somewhat greater advantages for
other kinds of contacts. Both parties targeted their bases in these election campaigns –
albeit more in 2004 than in 2012 – but with considerable inefficiency in both years.
Perhaps the biggest surprise in our results, especially in view of the recent emphasis on
campaigns designed to mobilize the party’s base, is that many respondents reported being
contacted by both parties, especially in 2012. There is evidence too that the 2012
campaigns were more concentrated on the battleground states than they had been in 2004.
We found numerous similarities between parties and between the two elections,
sometimes where we did not expect them. In both years, the campaigns directed their
canvassing efforts to the most likely voters – older, more affluent, more politically
involved, habitual voters. They concentrated them disproportionately on the battleground
states. No surprises here. Both campaigns also depended more on impersonal types of
contact than personal contacts or emails and other electronic messages. While easier to
accomplish, distributing literature and making phone calls are much less likely to be
effective in mobilizing voters than face-to-face approaches, which remain conspicuously
rare. Paradoxically, for all the talk of the new attention to email and other electronic
forms of messaging by 2012, few respondents reported receiving these messages, and
their number was only slightly higher in 2012 than in 2004. For all the attention paid to
Democratic successes with minority voters, moreover, they were less likely to have
contacted Hispanics in either year. Conversely, union members received inordinate
attention from both parties in 2012, but were neglected by the Republicans in 2004, when
it was home owners who were targeted by both in all states and in the battlegrounds.
Our analysis also identified party differences in their ground game efforts, albeit
again in some unexpected ways. Both campaigns were more likely to reach out to their
own partisans much more than their partisan opponents and nonpartisans, with the caveat
that surprisingly large numbers of partisans were contacted by both parties. The
Democratic campaign effectively targeted African Americans – in 2012 more than in
2004. The Republican campaign paid significantly more attention to whites. By contrast,
only in 2004 did regular church-going Protestants stand out as receiving more Republican
contacts. While both parties seemed to target high income voters, especially in the
battleground states, it was the Republicans showed consistency across the years in this
effort.
In subjecting the ground game to scrutiny through the window of reported party
contacts, it is not our intention to question the party and candidate organizations’
expertise and efforts, or their pride in their ground game successes. Both campaigns
poured enormous resources into their ground games in 2012, as they had in 2004. Party
contacts are now a staple of presidential campaigns! Both contacted millions of voters, in
24
many cases multiple times and through multiple means. Both made use of sophisticated
modeling and state-of-the-art data mining and targeting techniques to single out viable
targets to approach. Without these contacts, many eligible voters might not have cast a
ballot or supported the party’s candidate. The ground game probably was executed more
skillfully in 2012 by both campaigns than ever before.
While we do not doubt the conventional wisdom that the Obama campaign enjoyed
an edge in the ground game in 2012, our analysis suggests skepticism about how large –
and how consequential – that edge may have been. We also are skeptical about how
much each party’s ground game has improved its reach compared with just eight years
before. The campaigns’ increasingly intensive use of technology surely has improved
their productivity, but it cannot substitute for the labor-intensive face-to-face contacts that
seem so effective with voters.
Rather, ours is a cautionary tale. We recognize the difficulty of maintaining an
effective ground game across multiple states amidst a complicated and long presidential
campaign and in a voting-eligible electorate of over 220 million Americans. Even if the
efforts are focused on a dwindling number of battleground states, they have to be
prodigious. However assiduously the campaigns may build detailed voter profiles, it is
inevitable that they will fall far short of perfection in their micro-targeting. However
conscientiously their skilled staffs may plan and coordinate the efforts or their armies of
eager campaign volunteers may work the telephones, approach voters on their doorsteps,
or stuff mailers, what we know about campaigns in politics and other walks of life is that
much of this effort fails to reach, much less move, many possible recipients.
In writing about local party organizations fifty years ago, Eldersveld (1964, p. 526)
concluded that “… the party is no ‘master institution’ but a minimal-efficiency structure.”
That observation applies as well to modern presidential campaigns. Even the best of
campaigns are necessarily far from perfect in their ground game execution, and always
will be. Of course, what matters in an election campaign is relative effort and success.
The more relevant question is: Are they more effective than their opponents? To answer
this question, we need to determine how well the campaigns have reached out potential
voters. The window reported party contacts have provided in our analysis takes a
valuable first step in this direction.
25
References
Balz, Dan, and James Silberman. 2013. Collision 2012: Obama vs. Romney and the
Future of Elections in America. New York: Viking.
Beck, Paul A., and Erik Heidemann. Forthcoming. Changing Strategies in Grassroots
Campaigning: 1956 to 2012. Party Politics.
Chang, Linchiat, and Jon A. Krosnick. 2009. National Surveys Via RDD Telephone
Interviewing Versus the Internet: Comparing Sample Representativeness and
Response Quality. Public Opinion Quarterly 73: 641-678.
Dennis, J. Michael. 2001. Are Internet Panels Creating Professional Respondents? The
Benefits of Online Panels Far Outweigh the Potential for Panel Effects. Marketing
Research: Summer 34-38.
Eldersveld, Samuel J. 1964. Political Parties: A Behavioral Analysis. Chicago: Rand
McNally.
Gershtenson, Joseph. 2003. Mobilization Strategies of the Democrats and Republicans,
1956-2000. Political Research Quarterly 56: 293-308.
Gimpel, James G., Karen M. Kaufmann, and Shanna Pearson-Merkowitz. 2007.
Battleground States versus Blackout States: The Behavioral Implications of Modern
Presidential Campaigns. The Journal of Politics 69: 786-97.
Green, Donald P., and Alan S. Gerber. 2008. Get Out the Vote: How to Increase Voter
Turnout. Washington, D.C.: The Brookings Institution.
Issenberg, Sasha. 2012. The Victory Lab: The Secret Science of Winning Campaigns.
New York: Crown.
Knowledge Panel Design Summary. 2012. Posted on marketing.gfkamerica.com/
knowledgenetworks/ganp/reviwer-info.html.
La Raja, Raymond J. 2011. Back to the Future? Campaign-Finance Reform and the
Declining Importance of the National Party Organization. In John C. Green and
Daniel J. Coffey, eds., The State of the Parties, 205-222.
Malbin, Michael J., Aaron Dusso, Gregory Fortelny, and Brendan Glavin. 2011. The
Need for an Integrated Vision of Party and Candidates: National Political Party
Finances, 1999-2008. In John C. Green and Daniel J. Coffey, eds., The State of the
Parties, 185-204.
26
Matthews, Dylan. 2012. What’s a Good Ground Game Worth? About One Point.
Wonkblog of The Washington Post, posted on October 26, 2012 and downloaded on
May 6, 2013.
McDonald, Michael. 2013. 2012 General Election Turnout Rates. Posted on website at
elections.gmu.edu/Turnout_2012G, 7/22/13 version.
Popkin, Samuel L. 2012. The Candidate: What It Takes to Win – and Hold – the White
House. New York: Oxford University Press.
Rosenstone, Steven J., and John Mark Hansen. 1993. Mobilization, Participation, and
Democracy in America. New York: Macmillan.
Rutenberg, Jim. 2013. Data You Can Believe in: How the Precision Targeting of
‘Persuadable” Voters That Put Obama Over the Top in 2012 Could Revolutionize
the Advertising Industry. The New York Times Magazine, June 23, pp. 22-29 and 36.
Sanchez, Gabriel. 2013. The Untapped Potential of the Latino Electorate. Posted on the
Latino Decisions website on January 15, 2013
Sides, John, and Lynn Vavreck. 2013. The Gamble: Choice and Chance in the 2012
Presidential Election. Princeton, NJ: Princeton University Press.
Wielhouwer, Peter. 2003. In Search of Lincoln’s Perfect List: Targeting in Grassroots
Campaigns. The Journal of Politics 31: 632-69.
27
Appendix: Party Contact Questions
2012 U.S. CNEP Internet survey (conducted by GfK/Knowledge Networks)
Q1. “Did representatives of any of the political parties or presidential candidates contact
you during the 2012 campaign?” Check which ones among Democrat, Republican, and
another party (specify) options.
Q2. (IF Democrat checked) “Concerning the Democrats, was that contact with you
…
Q2a. … by mail or other printed material?”
Q2b. … on the telephone?”
Q2c. … in person?”
Q2d. … through email or other electronic messaging?”
Q3. (IF Republican checked) “Concerning the Republicans, was that contact with
you …
Q3a. … by mail?”
Q3b. … on the telephone?”
Q3c. … in person?”
Q3d. … through email or other electronic messaging?”
2004 U.S. CNEP Internet survey (conducted by Knowledge Networks)
Q1. “Please tell me whether any of the political parties or presidential candidates or their
representatives contacted you during the recent election campaign.” Check which ones
among Democrat, Republican, or another party (specify) options)
Q2. (IF Democrat checked) “Concerning the Democrats, was that contact with you
…
Q2a. … by mail?”
Q2b. … by telephone?”
Q2c. … in person?”
Q2d. … by email?”
Q3. (IF Republican checked) “Concerning the Republicans, was that contact with
you …
Q3a. … by mail?”
Q3b. … by telephone?”
Q3c. … in person?”
Q3d. … by email?”
Q4. (IF another party specified) “Concerning the (other party), was that contact
with you …
Q4a. … by mail?”
Q4b. … by telephone?”
Q4c. … in person?”
Q4d. … by email?”