Sage Journals: Discover world-class research

Abstract

This article analyzes the strategic allocation of presidential campaign visits in 2016. In particular, we test whether each campaign disproportionately targeted its presidential versus vice presidential candidates’ visits toward voters with whom they shared a salient demographic or political characteristic. Our purpose in doing so is to discern whether—and, if so, among which groups—the campaigns perceived the candidates as having a strategic advantage in appealing to affiliated voters. To this end, we analyze an original database of 2016 campaign visits that includes local population characteristics for each host site. Our results indicate that each ticket’s visits were highly coordinated across states, but frequently divergent within states. At the substate level, we find several systematic differences in the populations visited by presidential versus vice presidential candidates—in some cases aligning with a candidate’s personal characteristics. We discuss these findings’ implications with respect to campaign strategy and vice presidential selection.

Keywords

political science social science U.S. presidency U.S. vice presidency campaigns and elections political parties political behavior political geography campaign visits

Introduction

Strategic considerations figure prominently in public discussions of vice presidential selection. Certainly, this was evident in 2016. On the Democratic side, many observers speculated that vice presidential candidate Tim Kaine’s fluency in Spanish would help Hillary Clinton to win votes among Latinos (Felix & Shaik, 2016; Krauze, 2016). Likewise, Kaine’s reputation as a “centrist bridge builder” might increase the ticket’s appeal to political moderates (Krauze, 2016), and as Virginia’s senator and former governor surely he could “deliver” that battleground state’s electoral votes (Tribune News Services, 2016). On the Republican side, Donald Trump’s selection of Mike Pence reportedly “was made, in part, to help shore up conservative support throughout the Midwest and Rust Belt states” from which the Indiana governor hailed (Hillyard, 2016; also see McPherson, 2016; Villa, 2016).

Recent empirical analyses cast doubt upon a running mate’s ability to secure an electoral advantage among select demographic groups of voters and voters within his or her home state (see Devine & Kopko, 2016; Kopko & Devine, 2016). Nonetheless, the perception that running mates could produce such electoral advantages is widespread. But does that perception affect the actual conduct of a presidential campaign? Do the campaigns view the running mate as a unique strategic asset, to be deployed according to his or her (perceived) strengths, or as a means of reinforcement for the presidential candidate and his or her message?

It is possible, after all, that the strategic objective of choosing a running mate who appeals to voters alienated from, or less than enthusiastic about, the presidential candidate primarily is symbolic. In this case, the campaign’s goal simply might be to create the image of a well-rounded ticket and, in doing so, to improve perceptions of the presidential candidate among skeptical but persuadable voters. The vice presidential candidate’s role, then, would be to echo a shared message of what the ticket stands for and campaign in a parallel fashion. Alternatively, the presidential candidate might view his or her running mate as a more effective ambassador to voters who share a particular identity or affiliation, when those voters are unrepresented at the top of the ticket. In this case, the campaign would strategically deploy the vice presidential candidate to engage such voters with direct appeals that exploit their common bond in a way that the presidential cannot—or, at least, not as effectively. Likewise, the presidential candidate would skew his or her efforts toward appealing to voters with a shared identity or affiliation of their own. Essentially, the ticket would appeal to the same group of voters, as a whole, but divide its labors in such a way as to maximize receptivity by playing to each candidate’s (perceived) strengths.

Moreover, if the latter is true, then among which groups of voters—if any—can we reasonably discern that the campaigns perceived the presidential versus vice presidential candidate to have a strategic advantage? For that matter, is there an empirical basis for discerning which advantages the campaign might have believed to be real and electorally consequential when selecting the vice presidential candidate? Anecdotal evidence from the campaign trail might indicate such a perception; for instance, Tim Kaine made high-profile speeches entirely in Spanish to largely Latino audiences (Morin, 2016), and Mike Pence raised reporters’ eyebrows with “peculiar” visits to conservative outposts in rural America (Hillyard, 2016; also see Beaumont, 2016). But anecdotal evidence is not sufficient to establish patterns of activity and make reliable inferences; to do so, requires systematic analysis.

This study is the first to evaluate whether campaigns use the presidential versus vice presidential candidate disproportionately to appeal to voters with whom that individual shares a strategically advantageous affiliation (and, if so, which ones). To answer this research question, we analyze an original database of campaign visits in the 2016 presidential election, that includes population characteristics (i.e., demographic and political) associated with each locale that hosted a campaign visit. Our analysis provides direct insight into the strategic considerations that reasonably might be inferred to have influenced specific vice presidential selections—in contrast to previous studies analyzing the strategic considerations influencing selection processes, generally (Baumgartner, 2008; Sigelman & Wahlbeck, 1997).

To the extent that presidential tickets campaign in parallel fashion, echoing a common message targeted at the same voting constituency, we should find no systematic differences in the populations visited by either member of the ticket. Also, the candidates should visit the same locales roughly in equal proportions. Alternatively, if a campaign perceives that the presidential versus vice presidential candidate has a strategic advantage in appealing to voters who share a particular identity or affiliation, and campaign accordingly, the populations visited by each candidate should vary systematically and in accordance with the candidate’s characteristics. Also, the candidates should visit a given locale quite disproportionately, essentially dividing up the campaign’s “battleground” territory.

Why Campaign Visits?

Evaluating campaign strategy typically requires making inferences based upon patterns of resource allocation, as most scholars do not have direct access to the inner workings of a campaign (but see King & Morehouse, 2004; Shaw, 2006) and campaign personnel either will not describe their strategy openly or cannot be relied upon to describe it accurately. The most commonly cited empirical indicators of campaign strategy are candidate visits and advertising expenditures—often described by political scientists as the campaign equivalents of “time” and “money” (Bartels, 1985; Johnston, Hagen, & Jamieson, 2004; Shaw, 1999, 2006). The former, by many accounts, provides particular strategic insight; according to Althaus, Nardulli, and Shaw (2002), “Most campaign strategists will say that a candidate’s time is the campaign’s most valuable resource” (p. 50; also see Chen & Reeves, 2011; King & Morehouse, 2004).¹

Why is this the case? First, “Unlike financial resources, which depend on willing contributors, time is contributed by candidates and therefore controlled by them” (West, 1983, p. 517). Second, candidate appearances—the principal means by which campaigns commoditize the candidate’s time—serve a number of strategically valuable purposes, including providing a “controlled environment” (Wood, 2016, p. 111) in which candidates typically speak to “sympathetic, partisan audiences” (Holbrook & McClurg, 2005, p. 692) while attracting free, abundant, and largely positive local media coverage that is exceptionally likely to reach voters in the surrounding media market (Herr, 2002; Holbrook, 2002; King & Morehouse, 2004). In this way, campaign visits can serve to mobilize supporters through direct means, while also persuading undecided voters through indirect means (Jones, 1998).

Finally, and most importantly for our purposes, candidate visits can be targeted toward particular groups of voters that a campaign seeks to mobilize or persuade. As West (1983) explains, “In a world of limited time, candidates must make choices. They must emphasize certain constituencies and deemphasize others. These choices are not random” (p. 525). Indeed, he finds that presidential candidates during the 1980 primary and general election campaigns tended to speak to audiences that reflected their intended electoral coalition. For instance, Republican candidates who sought to expand beyond the party’s base were more likely to address audiences of voters—including union members, Catholics, and racial minorities—who belonged to groups that traditionally aligned with the Democratic Party. Moreover, West (1983) finds that the candidates often made personalized appeals at these events, by “communicat[ing] their ‘identifications’ and sympathies with voters” (p. 517).

It is, therefore, reasonable that political scientists often use campaign visits—and, in many cases, advertising expenditures—to make inferences about presidential campaign strategy. For example, resource allocations have been used to discern whether campaigns seek to maximize their Electoral College versus popular vote totals (Brams & Davis, 1974), employ a “base strategy” of mobilization versus a “peripheral strategy” of persuasion (Chen & Reeves, 2011), and perceive the vice presidential candidate as providing an electoral advantage in his or her home state (Devine & Kopko, 2016). The logic of our analysis mirrors that of previous studies: We infer the campaigns’ perceptions of strategic advantage based upon the pattern of choices that they made in allocating a resource generally within their control and regarded as valuable—namely, candidate visits. In essence, we seek to determine whether the 2016 presidential campaigns acted as if they believed that their presidential versus vice presidential candidates more effectively could appeal to voters with whom they shared a salient identity or affiliation.

Data, Hypotheses, and Methodology

Data for this analysis come from an original database of campaign visits in 2016—specifically, including visits made by either member of the Democratic or Republican ticket, between the vice presidential candidate’s official introduction (July 16 for Republicans and July 22 for Democrats) and Election Day (November 7). We use these data to test a series of hypotheses, described in this section, pertaining to the strategic allocation of presidential versus vice presidential campaign visits.

Data

For the purposes of this analysis, we define campaign visits as any public appearance—counted discretely, to include multiple visits within a given day—made by the presidential and/or vice presidential candidate that was apparently organized or selected by the campaign, or the candidates themselves, for the purpose of appealing to a localized concentration of voters. Included in this analysis are unscheduled events, such as visits to local businesses, restaurants, and campaign offices, provided that they involved direct interaction with voters and attracted media coverage. Excluded from our analysis are events that do not meet this definition, such as press conferences, private fundraisers, national conferences or conventions, and attendance at other nationally oriented events (e.g., the 9/11 commemoration in New York City).²

To identify qualifying events, first we consulted the candidates’ public schedules—from their campaign websites and other media sources—and then we verified the event’s occurrence, or in many cases discovered unscheduled visits, by searching for information from media websites and the candidates’ social media accounts. If we were able to validate an event’s occurrence and location using two sources of reliable, direct documentation (e.g., coverage by a reputable news source, video footage, photos posted by the campaign),³ we counted it in our analysis. If an event was canceled or if we could not document it properly, we excluded it from our analysis. In total, we identified 515 campaign visits—including 88 by Hillary Clinton; 96 by Tim Kaine; 13 by Clinton/Kaine, jointly (counted once for each candidate); 139 by Donald Trump; 130 by Mike Pence; and 18 by Trump/Pence, jointly (counted once for each candidate).

For each campaign visit, we also collected data on the demographic and political characteristics of the county in which it occurred.⁴ Demographic data were obtained from the U.S. Census Bureau,⁵ and they included estimates of a county’s population density, median income level, percentage of college graduates, African Americans, Latinos, military veterans, and 2010-2015 population growth. In addition, we obtained county-level estimates for the number of evangelical Protestants, mainline Protestants, and Catholics per 1,000 residents, from the Association of Statisticians of American Religious Bodies’ 2010 U.S. Religion Census.⁶ In terms of political characteristics, we measure a county’s competitiveness as the squared difference in two-party vote share from the 2012 presidential election.⁷ Because party competition only roughly approximates ideology and there is no reliable county-level measure of the latter,⁸ we also estimate local ideology using the DW-NOMINATE score for the U.S. Representative in whose district the event took place.⁹ Finally, we include variables to indicate whether the visit took place in the presidential or vice presidential candidate’s home region, as defined by the U.S. Census Bureau.¹⁰

Hypotheses

There is ample evidence to suggest that presidential campaigns often select vice presidential candidates at least partly based upon their demographic (e.g., home state, age, sex, religion) or political (e.g., ideology) characteristics (Baumgartner, 2008; Devine & Kopko, 2016; Goldstein, 2016). Presumably, they do so in hopes of appealing to strategically important voters who share these characteristics—even if there is very limited evidence that running mates actually win votes among these targeted groups (Devine & Kopko, 2016; Kopko & Devine, 2016). To the extent that campaigns perceive targeted appeals based upon shared characteristics to be effective and seek to capitalize upon them during the campaign, we should expect to find a positive relationship between a candidate’s demographic and political affiliations and the population characteristics associated with his or her campaign visits. This is our overarching research hypothesis. From it, we can also derive more specific hypotheses for the Republican and Democratic presidential tickets in 2016, using the population variables described above.

Based upon their demographic and political characteristics or affiliations, we hypothesize that Mike Pence was more likely than Donald Trump to visit counties that were Midwestern (home region), more rural (population density), more conservative (ideology), and more populated by traditional, middle- to upper-class members of the Republican establishment (income, college education) and evangelical Protestants (religion). Likewise, we hypothesize that Tim Kaine was more likely than Hillary Clinton to visit Southern (home region), moderate (ideology), battleground (competitiveness) counties, with larger populations of Catholics (religion), and Spanish speakers (Latinos). Also, as both vice presidential candidates had sons serving in the military at the time of the campaign, we hypothesize that they were more likely than their presidential counterparts to visit counties with higher percentages of military personnel (veterans).

For the presidential candidates, we hypothesize that Trump and Clinton both were more likely than their running mates to visit northeastern (home region) and more urban (population density) counties, as well as ones with more mainline Protestants (religion). Also, based upon his campaign rhetoric, we expect that Trump was more likely than Pence to visit White (race), working-class (income, college education) counties in economically distressed areas (population growth). Given Clinton’s strong support among African Americans during the 2016 primaries (Kurtzleben, 2016), we hypothesize that she was more likely than Kaine to visit counties with a higher African American population (race). Finally, as the first female major party nominee, we also hypothesize that she was more likely than Kaine to visit counties marked by indicators of socially progressive views on gender roles—including those that were more educated (college graduates), more wealthy (income), economically and socially dynamic (population growth), and less traditional (evangelical Protestants).

Methodology

To identify systematic differences in the strategic allocation of campaign visits, and test the hypotheses described above, we conduct a logistic regression analysis—separately for each party ticket—predicting whether a visit to a given county was made by the presidential (1) versus vice presidential (0) candidate. The independent variables in our models include each of the demographic and political characteristics previously described. Also, we add a series of variables representing the number of times that each candidate, on either ticket, had visited the same county prior to the visit in question.

If it is the case that presidential tickets campaign as a united front, targeting the same constituency of voters, we should find that the demographic and political characteristics of a given population have no bearing on whether it was targeted for a visit by the presidential versus vice presidential candidate. Also, with respect to previous campaign visits, we should find that each ticket essentially trades off visits within a given county; that is, if the presidential candidate has visited the county, the vice presidential will take the next turn. And the two will be equally likely to respond to a visit by their counterpart on the other ticket, rather than assigning the (vice) presidential candidate to counteract a visit by the opposing (vice) presidential candidate.

If, on the contrary, presidential campaigns differentially allocate campaign visits based upon the perceived electoral strengths of the presidential versus vice presidential candidate, we should find the opposite pattern. That is, the demographic and political characteristics of the local population should predict whether it receives a visit from one member of the party ticket or the other, and in a direction consistent with that candidate’s identity or affiliations. Furthermore, we should find that the campaigns largely divide up, rather than trade off, duties by county. In other words, a visit by the presidential candidate to a given county should be positively related to the number of times that he or she previously has visited that county, and negatively related to the number of times that the vice presidential candidate previously has visited that county.

Results

To provide context—as well as some initial evidence—for the analysis to follow, in Table 1 we present the total number of presidential and vice presidential campaign visits in 2016, by state. Then, we compare visit allocations at the state versus county level, to identify potential differences in campaign strategy—specifically, with respect to the activity of presidential versus vice presidential candidates—across versus within states.

Table 1.

Presidential and Vice Presidential Candidate Visits by State, 2016.

State	Republican ticket			Democratic ticket
State	Trump	Pence	Total	Clinton	Kaine	Total
AL	0	0	0	0	1	1
AZ	3	3	6	1	2	3
CO	10	7	17	3	2	5
CT	1	0	1	0	0	0
DC	2	0	2	0	0	0
FL	25	10	35	23	26	49
GA	0	3	3	0	0	0
IA	8	11	19	5	5	10
IL	1	0	1	1	0	1
IN	0	3	3	0	0	0
LA	2	2	4	0	0	0
ME	3	0	3	0	0	0
MI	9	8	17	5	5	10
MN	1	1	2	0	0	0
MO	0	2	2	0	0	0
MS	1	0	1	0	0	0
NC	19	20	39	17	15	32
NE	0	1	1	1	0	1
NH	9	9	18	3	7	10
NM	1	4	5	0	0	0
NV	5	5	10	9	4	13
NY	3	2	5	0	0	0
OH	19	24	43	15	10	25
PA	19	15	34	18	17	35
TX	1	0	1	0	4	4
UT	0	1	1	0	0	0
VA	9	13	22	0	5	5
WA	1	0	1	0	0	0
WI	5	4	9	0	6	6
Total	157	148	305	101	109	210

Total Visits by State

It is clear from Table 1 that both candidates on a given party ticket generally campaigned in the same states. The Republican candidates visited 28 states in total, with both appearing in 16 of those states. While Trump visited six states that Pence did not (Connecticut, Illinois, Maine, Mississippi, Texas, Washington), plus Washington, D.C., and Pence visited five states that Trump did not (Georgia, Indiana, Missouri, Nebraska, Utah), very few of the ticket’s appearances occurred in these states, individually or in total. In fact, 286 of the Republicans’ 305 campaign visits (93.8%) took place in states that both candidates visited. Trump and Pence also visited states with similar frequency. For instance, the three states that Pence visited most often—Ohio (24), North Carolina (20), Pennsylvania (15)—also were tied for the second-most visits by Trump, at 19 apiece. Florida, the state that Trump visited most often (25), ranked just sixth on Pence’s list—the only major discrepancy in the data. Overall, there is a very high correlation (.859) between the number of visits that Trump versus Pence made to the 28 states on their campaign itinerary in 2016. Including the states that both candidates did not visit, the correlation is even higher, at .892.

The Democratic candidates visited 16 states in total, with both appearing in 10 of those states. Clinton visited just two states that Kaine did not (Illinois and Nebraska), while Kaine visited four states that Clinton did not (Alabama, Texas, Virginia, Wisconsin). Again, most Democratic campaign visits—192 of 210 (91.4%)—took place in states to which both candidates traveled, and with similar frequency. In fact, Clinton and Kaine most frequently visited the same four states, and in the same rank order: Florida (Clinton 23, Kaine 26), Pennsylvania (18, 17), North Carolina (17, 15), Ohio (15, 10). Clinton’s and Kaine’s visits are even more highly correlated than those of their Republican counterparts—.906 for the 16 states in which one or both candidates appeared and .938 when including all states.

Based upon this evidence, it appears that each presidential campaign used its presidential and vice presidential candidates to target mostly the same subset of battleground states, and with almost the same frequency. However, this does not mean that the campaigns used each candidate to target the same population of voters. After all, states are diverse geographical units, and for various reasons—including those relating to the candidates’ identities and affiliations, as described above—campaigns might send the presidential versus vice presidential candidate to different parts of a state (e.g., a city or county) where they are expected to have greater appeal. If so, the similarities found in Table 1, with regard to the presidential tickets’ state-level campaign visits, might mask stark differences in the strategic deployment of candidates within the same states.¹¹

Total Visits by County

Indeed, a parallel analysis of county-level visits shows much greater disparities between the candidates. First, most of the counties visited by a presidential or vice presidential candidate in 2016—180 of 258 counties (69.8%), overall—were visited by one member of the party ticket but not the other. Among Republicans, Donald Trump visited 108 counties, 57 of which Mike Pence did not (52.8%); Pence visited 111 counties, 60 of which Trump did not (54.1%). Among Democrats, Hillary Clinton visited 48 counties, 21 of which Tim Kaine did not (43.8%); Kaine visited 69 counties, 42 of which Clinton did not (60.9%).

In terms of total campaign visits, a sizable proportion took place within counties visited by one member of a ticket and not the other—for Republicans, 126 of 305 visits (41.3%), and for Democrats, 84 of 210 visits (40.0%). These figures are virtually identical for Trump (40.8%) and Pence (41.9%), but more disparate for Clinton (30.7%) and Kaine (48.6%). Most striking of all is the correlation between the number of visits per county by the presidential versus vice presidential candidate (including only the counties that were visited by at least one member of the ticket)—just .300 for Trump–Pence and .462 for Clinton–Kaine.

This analysis suggests an important conclusion about presidential campaign strategy in 2016: The presidential and vice presidential candidates on a given ticket did, in fact, tend to campaign in the same battleground states and with similar frequency, but at the same time they tended to campaign in different parts of those states. Were those differences random, or systematically related to characteristics of the local population? And, if systematic, did the characteristics of the local population match those of the candidate in such a way as to suggest the perception of a strategic advantage based upon shared identity or affiliation? To answer these questions, next we analyze the local demographic and political characteristics associated with each campaign visit.

Bivariate Analysis

In this section, we estimate the empirical relationship between our dependent variable (DV)—coded to indicate a visit by the presidential (1) versus vice presidential (0) candidate on a given ticket—and a series of independent variables representing each of the demographic and political population characteristics previously described.¹² Our first step in doing so is to estimate the bivariate relationship between the DV and each population characteristic, using difference of means testing. Essentially, this analysis tells us whether the locales visited by the presidential candidate on a given ticket significantly differed from those visited by the vice presidential candidate, on average. We count each campaign visit discretely when calculating a candidate’s mean population characteristics. As a result, the population characteristics of a given locale are weighted according to the number of times that it was visited by the candidate in question. For instance, Hillary Clinton made five visits to Miami-Dade County, Florida, and one visit to Palm Beach County, Florida; therefore, when calculating the mean population characteristics of a Clinton visit, Miami-Dade County has 5 times the weight of Palm Beach County.

Our bivariate analyses indicate some significant differences in the populations associated with a presidential versus vice presidential candidate’s visit, on both tickets. First, with respect to the Democratic ticket, Hillary Clinton visited counties with significantly higher percentages of African Americans, on average, than those visited by Tim Kaine. Also, the average population density associated with a Clinton visit is higher, albeit at marginal levels of statistical significance. Kaine, in comparison with Clinton, visited counties that were significantly more competitive in the 2012 presidential election and had higher percentages of evangelical Protestants. Also, Kaine’s visits took place in significantly more conservative—or, in context, more moderate¹³—congressional districts than Clinton’s. There is additional evidence, although marginally significant, that Kaine visited counties with higher percentages of military veterans and college graduates. Some of these differences clearly align with the respective candidates’ characteristics. Clinton, as noted above, had strong ties to the African American community. Also, she was a senator from one of the nation’s most densely populated states, New York, and her campaign headquarters were located in Brooklyn. Kaine’s visits align with his profile as a relatively moderate senator from a battleground state, whose son was currently serving in the military. It is less clear why he would campaign in counties with more college graduates, and especially ones with more evangelical Protestants, as he is a Catholic.

Our analysis of the Republican ticket yields fewer significant differences. Donald Trump visited counties with higher percentages of African Americans in comparison with Mike Pence, but this difference is only marginally significant. Pence visited counties with significantly higher percentages of mainline Protestants, in comparison with Trump, and more ideologically conservative congressional districts. Also, Pence was marginally more likely to visit states in his home region of the Midwest. The latter differences align with Pence’s profile as a conservative, Midwestern governor, whose role in the campaign, many believed, was to reach out to voters who shared these affiliations. However, it is less clear why he would campaign in areas with more mainline Protestants, as Trump—not Pence, an evangelical Protestant—shares this identity. Trump’s visits to more heavily African American counties also seem surprising, given that his primary and general election campaigns were widely perceived as appeals to White, working-class voters. However, Trump announced his attention to appeal to African Americans during the campaign, and his pattern of campaign visits reflects just such an effort.

Of course, bivariate analyses are limited, methodologically, because they do not account for intervening variables and therefore might misidentify causal relationships. For example, it might be the case that Kaine did not visit counties with higher percentages of evangelical Protestants because he was trying to appeal to evangelical Protestants; rather, he might have been trying to appeal to ideological moderates, who just happen to live in counties with more evangelical Protestants as compared with the relatively liberal areas that Clinton tended to visit. To better disentangle the causal relationship between population characteristics and candidate visits, next we present results from a multivariate analysis conducted separately for each presidential ticket.

Multivariate Analysis

Table 2 presents results from two logistic regression models, analyzing campaign visits made by the Republican ticket (column 1) and the Democratic ticket (column 2) in 2016. The DV in each model is coded 1 for a visit by the presidential candidate on a given ticket, and 0 for a visit by the vice presidential candidate. The independent variables included in each model represent the demographic and political characteristics of the geographic area in which a given campaign visit took place. In addition, we include four variables measuring the number of times that each candidate previously had visited the county in question, to better understand how presidential tickets “share” (or do not share) campaign territory and respond (or do not respond) to visits by the opposing ticket.

Table 2.

Logit Models Predicting Presidential Versus Vice Presidential Candidate Visits, 2016.

Independent variables	Republican ticket	Democratic ticket
Population growth, 2010-2015	0.058 (0.043)	0.206** (0.071)
Population density	0.000 (0.000)	0.000 (0.000)
Median income	0.000 (0.000)	−0.000 (0.000)
College graduate %	−0.071* (0.028)	−0.096* (0.043)
African American %	0.020 (0.017)	0.025 (0.032)
Latino %	−0.033^† (0.018)	−0.044^† (0.024)
Military veteran %	−0.112 (0.084)	−0.167 (0.169)
Evangelical protestants/1,000	0.002 (0.002)	−0.018 (0.006)
Mainline protestants/1,000	−0.009** (0.004)	0.007 (0.005)
Catholics/1,000	0.002 (0.002)	−0.004 (0.003)
Home region: Presidential candidate	0.106 (0.711)	−0.545 (0.696)
Home region: Vice presidential candidate	−0.588 (0.531)	0.343 (0.727)
County competitiveness, 2012	−0.000 (0.000)	−0.000 (0.000)
State competitiveness, 2012	−0.001 (0.001)	0.004 (0.003)
Congressional District Ideology	−0.701* (0.290)	−0.857* (0.405)
No. of previous visits: Clinton	−0.167 (0.216)	−0.462* (0.200)
No. of previous visits: Kaine	0.020 (0.177)	0.552* (0.221)
No. of previous visits: Trump	−0.789*** (0.230)	0.249 (0.222)
No. of previous visits: Pence	0.953*** (0.270)	0.402^† (0.241)
Constant	2.627^† (1.561)	5.815^† (3.000)
N	266	184
% Reduction in error	36.43%	48.86%
% correctly predicted
DV = 0 (Vice presidential campaign visit)	67%	74%
DV = 1 (Presidential campaign visit)	71%	77%

Note. The DV in each model is coded to represent a campaign visit by the presidential (1) versus vice presidential (0) candidate on a given party ticket. The independent variables represent county-level population characteristics, unless otherwise noted. Entries are logistic regression coefficients. Standard errors are in parentheses. DV = dependent variable.

Statistical significance: †p < .10. *p < .05. **p < .01. ***p < .001.

The results in column 1 indicate that Mike Pence visited counties with a higher percentage of college graduates and mainline Protestants, as well as congressional districts that were more ideologically conservative, than Donald Trump. Each of these differences is statistically significant at p < .05. Also, at marginal significance levels (p < .10), we find that Pence visited counties with a higher percentage of Latinos. Some of this evidence is consistent with the notion that Pence could more effectively appeal to traditional Republicans, including conservatives and individuals with a higher socioeconomic status, than Trump. Also, to the extent that Pence campaigned in more heavily Latino areas, this may reflect concerns that Trump was personally alienating to such voters given his harsh rhetoric on immigration and toward Mexicans or Mexican Americans. That Pence also campaigned in areas with more mainline Protestants is mystifying, given that Trump shared this religious affiliation and Pence did not. In fact, it may be even more interesting to note the population characteristics that did not predict a visit by Trump versus Pence. Contrary to popular perceptions of Pence’s electoral appeal, he was no more likely than Trump to campaign in rural areas, the Midwest, battleground states or counties, and among large populations of evangelical Protestants. As a whole, these results are decidedly mixed: On one hand, Trump and Pence campaigned in distinguishable locales that, in some cases, corresponded to salient aspects of their personal or political identity; on the other hand, the populations they visited were indistinguishable in most respects, some of which were commonly associated with the candidates and even viewed as likely reasons for Pence’s selection as a vice presidential candidate.

Further complicating our evaluation of the Republican candidates’ strategic partnership is the performance of the campaign visits variables. A visit by Trump versus Pence to a given county can be predicted, at conventional significance levels, by the number of times that either candidate previously had visited the county—but not in the direction that one would expect if the candidates were dividing territory between them, rather than sharing it. Indeed, Pence became more likely to visit a county as Trump’s previous visits to the same increased; likewise, Trump became more likely to visit a county as Pence’s previous visits to the same increased.¹⁴ It is also noteworthy that the number of campaign visits to a given county by the opposing candidates does not predict a visit by Trump versus Pence. This is not same as saying that the Republican ticket decided against responding to visits by their opponents, as our DV is constructed to predict a visit by one versus the other Republican candidate, not the ticket as a whole. However, one might suspect that such “response” visits, to the extent that they occur, are made in kind, with the (vice) presidential candidate being commissioned to directly counteract the efforts of the opposing (vice) presidential candidate. Judging by these results, that is not the case.

We find a similar pattern of results for the Democratic ticket, in column 2. Indeed, the populations visited by Clinton versus Kaine are distinguishable in several respects. Clinton was more likely to visit counties with higher levels of recent population growth, whereas Kaine was more likely to visit counties with a higher percentage of college graduates and evangelical Protestants. Also, Kaine was more likely to visit ideologically moderate, or less liberal, congressional districts, than Clinton. Each of these differences is statistically significant at p < .05. Interestingly, but only at p < .10, we also find that Kaine was more likely to visit areas with a higher percentage of Latinos. The latter finding validates, albeit with limited confidence, the notion that the Democratic campaign perceived Kaine as having a strategic advantage in appealing to Latinos, due to his fluency in Spanish, and that this might have contributed to his selection as a vice presidential candidate. The same can be said, with greater confidence, about his perceived appeal to moderate voters that might have been alienated by Clinton’s liberal reputation. Once more, though, it is unclear why Kaine would have campaigned among more college-educated and evangelical populations.

But perhaps more striking, and in keeping with our findings for Republicans, are the ways in which Clinton’s and Kaine’s campaign visits did not differ. Clinton was no more likely than Kaine to campaign in counties with higher population density or more African Americans, and Kaine was no more likely than Clinton to campaign in counties with more Catholics or veterans, in battleground states or counties, or in the region encompassing his battleground home state of Virginia.¹⁵ Our conclusion echoes that stated above with respect to the Republican ticket: These are decidedly mixed results, some of which point to strategic differentiation based upon candidate characteristics and many more that do not.

Our results further parallel the Republican model when analyzing the campaign visits variables, with one minor qualification. Again, we find that the frequency with which Democratic candidates previously had visited a given county significantly predicts a present visit by Clinton versus Kaine—but in the direction of sharing, rather than dividing, territory. That is, Kaine became more likely to visit a county as Clinton’s previous visits to the same increased; likewise, Clinton became more likely to visit a county as Kaine’s previous visits to the same increased. In terms of responding to the opposing ticket, Trump’s previous visits to a county do not significantly predict a present visit by Clinton versus Kaine. On the contrary, Pence’s previous visits predict a present visit by Clinton—but at the marginal significance level of p = .095. Overall, Clinton and Kaine—like Trump and Pence—appear more responsive to their running mate’s activity within a particular county, than that of their counterpart on the opposing ticket.

Finally, our models’ overall performance indicates that the allocation of presidential versus vice presidential campaign visits in 2016 was not random but, in fact, fairly systematic when accounting for local population characteristics and other strategic considerations. For the Democratic ticket, our model accurately predicts which candidate visited a given county approximately three quarters of the time—77% for Hillary Clinton and 74% for Tim Kaine. And, in comparison with predicting which candidate visited a county based on random chance, alone, the model reduces prediction error by an impressive 48.86%. The Republican ticket proved less predictable, but still far from random in its visit allocations. Here, our model accurately predicts which candidate visited a given county at least two thirds of the time—71% for Donald Trump and 67% for Mike Pence. And, in comparison with a random prediction model, it yields a proportional reduction in error of 36.43%.

If each presidential ticket had campaigned as a unit, or if a particular candidate’s deployment on the campaign trail reflected no strategy other than that which guided the ticket as a whole, then predicting visit allocation patterns based upon the variables included in our model should have provided little, if any, empirical insight. Clearly, that is not the case.

Discussion

Do presidential campaigns believe that the vice presidential candidate provides a strategic advantage in appealing to voters with whom he or she shares a salient identity or affiliation? And do those perceptions of an advantage influence the actual conduct of a campaign, such that presidential and vice presidential candidates disproportionately target their appeals toward groups of voters who share their personal characteristics? Or is the running mate’s role simply to reinforce the campaign’s message, by multiplying—rather than diversifying—the presidential candidate’s efforts to communicate that message to a shared constituency of voters?

To answer these questions, we analyze an original database of presidential and vice presidential candidate visits in 2016. In particular, we compare the geographic areas to which either candidate on a given ticket traveled, to identify whether the demographic and political characteristics of the local population differed systematically between them, and whether those differences align with each candidate’s personal characteristics in such a way as to suggest the campaign’s perception of a strategic advantage.

Our analysis indicates that the 2016 presidential and vice presidential candidates’ campaign visits differed significantly, in many respects. First, although the candidates on each party ticket campaigned almost entirely in the same battleground states, and with similar frequency, their travels within states often diverged. Approximately 40% of each ticket’s campaign visits occurred in counties that the presidential candidate visited but the vice presidential candidate did not, or vice versa. Moreover, the total number of visits per county by each set of candidates is modestly correlated, at .300 for Republicans and .462 for Democrats. By way of comparison, less than 10% of each ticket’s campaign visits occurred in states visited by one candidate and not the other, while the candidates’ total number of visits by state is highly correlated, at .859 for Republicans and .906 for Democrats. The campaigns’ evident coordination of state-level presidential and vice presidential candidate visits suggests a coherent strategy and the will, as well as the ability, to implement it; to assume that the much lesser degree of coordination within states somehow was not strategically motivated, then, seems unreasonable.

Our logistic regression models also indicate systematic differences in the allocation of presidential versus vice presidential campaign visits. For the Democratic ticket, we find that Hillary Clinton was more likely than Tim Kaine to visit counties that had recently experienced higher levels of population growth. Also, at conventional significance levels, we find that Kaine was more likely than Clinton to visit counties with a higher percentage of college graduates and evangelical Protestants, as well as more moderate congressional districts; at marginal significance levels, Kaine was more likely to visit counties with higher percentages of Latinos. For the Republican ticket, we find that Mike Pence was more likely than Donald Trump to visit counties with a higher percentage of college graduates and mainline Protestants, as well as more conservative congressional districts. Moreover, at marginal significance levels, we find that Pence was more likely to visit counties with a higher percentage of Latinos.

This evidence indicates that while presidential and vice presidential candidates run together on the same ticket, and usually in the same states, their campaign activities are not merely parallel or duplicative. Vice presidential candidates, at least in 2016, frequently campaigned in different counties than the presidential candidate, and among distinguishable populations. But does this prove that the observed differences were strategically motivated? For that matter, can we reasonably infer that the vice presidential candidates were selected, at least in part, for the purpose of appealing to the populations that they visited disproportionately on behalf of their ticket?

Obviously, we cannot know the answers to these questions with any certainty. However, we do find evidence of an alignment between some of the population characteristics that predict a vice presidential campaign visit and the personal characteristics of that candidate, which at least suggests that perceptions of a strategic advantage might have influenced vice presidential selection. For instance, both running mates were significantly more likely than the presidential candidate to visit congressional districts that aligned with their ideological profile—in Kaine’s case, more moderate districts, and in Pence’s case, more conservative districts. On the contrary, the demographic predictors of a vice presidential campaign visit are less clearly related to the candidate’s personal characteristics, except for the marginal finding that Kaine was more likely to campaign in areas with higher Latino populations, and perhaps the finding that Pence was more likely to campaign in areas with a higher percentage of college graduates (i.e., higher socioeconomic status). Indeed, many of the personal characteristics cited in public discussions of Kaine’s and Pence’s electoral appeal, prior and subsequent to vice presidential selection, had no evident effect on where they campaigned—including religious affiliation, population density, military service, or home region. This evidence undercuts any suggestion that the campaigns believed their running mates presented a strategic advantage in appealing to voters on these bases, or even selected them with such advantages in mind. For that matter, our analysis indicates that campaigns are more likely to value a vice presidential candidate’s ability to appeal to voters on the basis of ideological, rather than demographic, affiliation.

Indeed, we must not go too far in characterizing the strategic significance of vice presidential candidates, and the degree to which presidential tickets diverge in the targeting of their campaign appeals. While we have identified systematic differences in the campaign activities of presidential and vice presidential candidates, in some respects those activities overlap. Perhaps most notably, our logistic regression models indicate that the candidates did not simply divide up the campaign battleground, repeatedly visiting a designated sphere of influence to the exclusion of the other candidate. Rather, for both tickets, we find that the presidential candidate becomes more likely to visit a county as his or her vice presidential candidate’s previous visits to the same increase, and vice versa. This does not mean that the candidates trade off appearances, in tat-for-tat fashion, so as to visit each locale in equal proportions; the evidence already presented demonstrates that this is not the case. A more reasonable interpretation is that the campaigns require both candidates’ presence in counties deemed sufficiently important to merit multiple, or even many, visits. Furthermore, it is worth noting that while a sizable proportion of campaign visits—approximately 40%, for both parties—take place in counties not visited by the other member of the ticket, conversely it is the case that the majority take place in counties that both candidates have visited.

For these reasons, it would be a misinterpretation of our empirical findings to conclude that presidential and vice presidential candidates conduct divergent campaigns, targeting different populations of voters. It would be more accurate to conclude that, at least in 2016, the presidential campaigns coordinated their candidates’ activities so as to appeal to mostly the same voters while in some instances engaging in disproportionate allocations based upon the perceived electoral strengths of either candidate.

Conclusion

This analysis makes an important contribution to existing scholarship on vice presidential candidates, and campaign strategy in general. First, whereas previous studies examine the patterns and electoral effects of vice presidential (as well as presidential) campaign visits (e.g., Althaus et al., 2002; Hill, Rodriguez, & Wooden, 2010; Shaw, 2006), ours is the first to test for systematic differences in the strategic allocation of campaign visits within a party ticket, based on demographic and political population characteristics. In doing so, we also provide direct insight into the strategic considerations that may have influenced specific vice presidential selections—in contrast to previous studies analyzing the strategic considerations influencing selection processes, generally (Baumgartner, 2008; Sigelman & Wahlbeck, 1997). Finally, this research builds upon a limited number of previous studies that use local population characteristics to discern campaign strategy with respect to the allocation of campaign visits within, rather than across, states (Althaus et al., 2002; Chen & Reeves, 2011). Indeed, our analysis adds relevant variables not included in those previous studies, such as religious affiliation and ideology, that we find to be predictive of visit allocations.

This analysis also suggests several valuable opportunities for future research. The most obvious such opportunity is to extend our methodology to past elections, so as to generalize or contextualize our conclusions about the strategic role of vice presidential candidates, beyond 2016. This is a practical goal, as there are identifiable strategic advantages associated with each vice presidential candidacy and they can be measured using the same variables, from the same data sources (e.g., U.S. Census Bureau, DW-NOMINATE, election results), employed here. Also, the empirical insights gained from this research can be used to better understand the effect of vice presidential, as well as presidential, campaign visits on voting behavior and state-level election outcomes. Specifically, one could investigate the potential moderating influence of strategically relevant local population characteristics on the electoral consequences of campaign visits to determine whether those visits are more effective at increasing voter turnout or party vote share when targeted toward populations that align with the characteristics of a particular candidate. In essence, this would extend our analysis by evaluating whether the campaign’s perception that a candidate has particular strength in appealing to an affiliated group of voters actually has the intended effect.

Such research might be useful to presidential campaigns, as well as scholars. As King and Morehouse (2004) note, in reference to the latter’s experience working as a strategic advisor and trip director for the Gore-Lieberman campaign in 2000, that campaign’s leadership “did not base [its campaign travel expenditures] on any quantifiable data. They, like other campaigns before and after, simply know that candidate visits to targeted areas have a positive effect on voters” (p. 305). If, in fact, candidates’ visits are disproportionately allocated to appeal to groups of affiliated voters, but the anticipated strategic advantages prove to be ephemeral, then what the campaigns do not know actually might hurt them. And, if the campaign’s anticipation of such advantages influences the selection of a vice presidential candidate who is less qualified to assume office if elected, then what they do not know might hurt all of us.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

ORCID iD

Kyle C. Kopko

Author Biographies

Christopher J. Devine is an assistant professor of Political Science at the University of Dayton in Dayton, Ohio, USA. He earned his PhD in political science from The Ohio State University in 2011. His research focuses on the U.S. presidency, the U.S. vice presidency, voting behavior, political psychology, partisanship, and ideology.

Kyle C. Kopko is an associate dean of Institutional Effectiveness, Research, and Planning and an associate professor of Political Science at Elizabethtown College in Elizabethtown, Pennsylvania, USA. He earned his PhD in political science from The Ohio State University in 2010. His research focuses on the U.S. presidency, the U.S. vice presidency, judicial politics, voting behavior, political psychology, and partisanship.

References

Althaus

S. L.

Nardulli

P. F.

Shaw

D. R.

(2002). Candidate appearances in presidential elections, 1972-2000. Political Communication, 19, 49-72.

Bartels

L. M.

(1985). Resource allocation in a presidential campaign. The Journal of Politics, 47, 928-936.

Baumgartner

J. C.

(2008). The veepstakes: Forecasting vice presidential selection in 2008. PS: Political Science & Politics, 41, 765-772.

Beaumont

(2016, October 5). Some see Pence, post-debate, as top-of-ticket material. Associated Press. Retrieved from http://bigstory.ap.org/article/ff938630c20341c98605a7cdfa8afac8/some-see-pence-post-debate-top-ticket-material

Brams

S. J.

Davis

M. D.

(1974). The 3/2’s rule in presidential campaigning. American Political Science Review, 68, 113-134.

CBS News. (2016, November 6). Hillary Clinton speaks at Philly church. CBS News. Retrieved from http://philadelphia.cbslocal.com/2016/11/06/hillary-clinton-speaks-at-philly-church

Chen

L. J.

Reeves

(2011). Turning out the base or appealing to the periphery? An analysis of county-level candidate appearances in the 2008 presidential campaign. American Politics Research, 39, 534-556.

Devine

C. J.

Kopko

K. C.

(2016). The VP advantage: How running mates influence home state voting in presidential elections. Manchester, UK: Manchester University Press.

Felix

Shaik

(2016, July 28). Why it matters that Tim Kaine speaks Spanish. Univision News. Retrieved from http://www.univision.com/univision-news/why-it-matters-that-tim-kaine-speaks-spanish

10.

Goldstein

J. K.

(2016). The White House vice presidency: The path to significance, Mondale to Biden. Lawrence, KS: University Press of Kansas.

11.

Herr

J. P.

(2002). The impact of campaign appearances in the 1996 election. The Journal of Politics, 64, 904-913.

12.

Hill

J. S.

Rodriguez

Wooden

A. E.

(2010). Stump speeches and road trips: The impact of state campaign appearances in presidential elections. PS: Political Science & Politics, 43, 243-254.

13.

Hillyard

(2016, August 27). Mike Pence carves familiar path across swing states for Donald Trump. NBCNews.com. Retrieved from http://www.nbcnews.com/politics/2016-election/mike-pence-carves-familiar-path-across-swing-states-donald-trump-n638756

14.

Holbrook

T. M.

(2002). Did the whistle-stop campaign matter? PS: Political Science & Politics, 35, 59-66.

15.

Holbrook

T. M.

McClurg

S. D.

(2005). The mobilization of core supporters: Campaigns, turnout, and electoral composition in United States presidential elections. American Journal of Political Science, 49, 689-703.

16.

Johnston

Hagen

M. G.

Jamieson

K. H.

(2004). The 2000 presidential election and the foundations of party politics. Cambridge, UK: Cambridge University Press.

17.

Jones

J. M.

(1998). Does bringing out the candidate bring out the votes? The effects of nominee campaigning in presidential elections. American Politics Research, 26, 395-419.

18.

King

D. C.

Morehouse

(2004, January). Moving voters in the 2000 presidential campaign: Local visits, local media . In Schultz

D. A.

(Ed.), Lights, camera, campaign!: Media, politics, and political advertising. New York, NY: Peter Lang.

19.

Kopko

K. C.

Devine

C. J.

(2016, July 22). Will Tim Kaine deliver Virginia (and Catholics)? Don’t count on it. The Washington Post. Retrieved from https://www.washingtonpost.com/news/monkey-cage/wp/2016/07/22/will-tim-kaine-deliver-virginia-and-catholics-dont-count-on-it/?utm_term=.a6c5254c40d2

20.

Krauze

(2016, July 24). Tim Kaine, En Español. The New Yorker. Retrieved from http://www.newyorker.com/news/news-desk/tim-kaine-en-espanol

21.

Kurtzleben

(2016, March 1). Understanding the Clintons’ popularity with Black voters. NPR. Retrieved from http://www.npr.org/2016/03/01/468185698/understanding-the-clintons-popularity-with-black-voters

22.

McPherson

(2016, July 14). Pence could build bridges to social conservatives, GOP lawmakers say. Roll Call. Retrieved from http://www.rollcall.com/news/politics/gop-lawmakers-say-pence-build-bridges-social-conservatives

23.

Morin

(2016, October 16). Kaine delivers entire speech in Spanish at Miami church. Politico. Retrieved from http://www.politico.com/story/2016/10/tim-kaine-delivers-entire-speech-in-spanish-at-miami-church-229866

24.

Pence

(2016, November 5). Twitter.com. Retrieved from https://twitter.com/mike_pence/status/795021732474195968

25.

Shaw

D. R.

(1999). The effect of TV ads and candidate appearances on statewide presidential votes, 1988-96. American Political Science Review, 93, 345-361.

26.

Shaw

D. R.

(2006). The race to 270: The electoral college and the campaign strategies of 2000 and 2004. Chicago, IL: The University of Chicago Press.

27.

Sigelman

Wahlbeck

P. J.

(1997). The “veepstakes”: Strategic choice in presidential running mate selection. American Political Science Review, 91, 855-864.

28.

Tribune News Services. (2016, July 31). In Tim Kaine, democrats might have the weapon to keep Virginia blue. Chicago Tribune. Retrieved from http://www.chicagotribune.com/news/nationworld/politics/ct-tim-kaine-virginia-20160731-story.html

29.

Villa

(2016, July 14). Newt Gingrich: Mike Pence would have “Midwestern appeal.” Time. Retrieved from http://time.com/4406836/donald-trump-running-mate-mike-pence-newt-gingrich

30.

West

D. M.

(1983). Constituencies and travel allocations in the 1980 presidential campaign. American Journal of Political Science, 27, 515-529.

31.

Wood

(2016). What the heck are we doing in Ottumwa, anyway? Presidential candidate visits and their political consequence. The Annals of the American Academy of Political and Social Science, 667, 110-125.

Split Tickets? On the Strategic Allocation of Presidential Versus Vice Presidential Campaign Visits in 2016

Abstract

Keywords

Introduction

Why Campaign Visits?

Data, Hypotheses, and Methodology

Data

Hypotheses

Methodology

Results

Total Visits by State

Total Visits by County

Bivariate Analysis

Multivariate Analysis

Discussion

Conclusion

Footnotes

Declaration of Conflicting Interests

Funding

Notes

ORCID iD

Author Biographies

References