Social networks in political campaigns: Facebook and the congressional elections of 2006 and 2008

Profiles of early adopters: diffusion of innovation literature

Rogers (2003) describes diffusion as the process by which members of a social system come to know, persuade, decide, implement and confirm their adoption of an innovation. When a new technology is introduced, innovators are the first to embrace it, followed by early adopters, early majority, late majority, and finally laggards. The empirical literature has studied the characteristics of early adopters and also the timing and extent of the adoptions for both individuals and organizations (Fichman, 1992; Frambach and Schillewaert, 2002). Adoption decisions are thought to depend not only on the characteristics of the adopter, but also on characteristics of the innovation or technology and of the environment. For example, Kwon and Zmud (1987) identify five categories of these contextual factors: characteristics of the adopting organization, the user community, the innovation or technology, the task, and the environment. Robertson and Gatignon (1986) describe a large set of factors related to what they call the competitive supply-side and adopter environments that affect the diffusion of new technologies. The former include factors such as the degree of competition, standardization of the technology, research and development resources. The competitive adopter environment includes both structural attributes of the adopter industry such as its homogeneity and uncertainty of demand (customer needs) and communication attributes such as the professionalism and cosmopolitanism (external orientation) of the adopter industry.

Political science studies of the diffusion of campaign websites have examined fewer categories and a more limited number of contextual factors. Most draw upon the same finite set that is divided between constituency (the user community) and political (the environment) factors. Constituencies are described demographically by median income and percentage of urban, white, college-educated, and elderly (Foot and Schneider, 2006). The political environment is described by characteristics of the electoral contest and candidate: level of office, competitiveness of the race, party identification of the candidate, party status (major or minor party), status of the seat (incumbent, challenger, open seat), and amount of campaign funds raised.

The constituency attributes selected to explain campaign website adoption are those that have been shown to correlate with citizens’ access to and use of the Internet (Chadwick, 2006; Klotz, 2004; Mossberger et al., 2003). Higher levels of education make people more comfortable with and skilled in the use of technology, while higher levels of income make computers easier to afford. Although whites use the Internet at higher rates than blacks, racial differences have diminished over time and seem to be a reflection of disparities in education and income (Marriott, 2006). The age gap persists, however: Internet use declines with each advancing age group. Urban areas have greater Internet use than rural areas, but the difference has declined substantially. These constituency demographics in turn influence candidates’ Internet use (Herrnson et al., 2007).

The diffusion of innovation literature suggests an additional reason why constituency factors should lead candidates to adopt new technologies. Organizations are mindful of the degree to which an innovation is compatible and incompatible with expectations (existing norms and values), as well as the needs and capacities of its users or customers (Tornatzky and Klein, 1982; Ward and Gibson, 2009). Our interviews with staff and consultants from the 2008 campaigns seemed to support this view. For example, one campaign aide told us that ‘Our campaign did not utilize Facebook or other social networking sites as part of our campaign. We have an all-volunteer campaign and our outreach efforts focused largely on grassroots methods such as door-to-door campaigning and other personal interaction with voters.’ ¹ Another noted that ‘Facebook is more of a national election tool right now.’ ²

In the early days of web campaigning, incumbents were less likely than challengers to campaign on the web, but a competitive race increased its use by incumbents and challengers alike (Foot and Schneider, 2006; Herrnson et al., 2007; Kamarck, 2002; Xenos and Foot, 2005). Financially disadvantaged candidates also were less likely to have a campaign website in those early days (Gibson et al., 2003), but this has proved less of a barrier subsequently. Financial resources and major party status still differentiate which campaigns incorporate the latest technology and features, however (Foot and Schneider, 2006; Strandberg, 2009). Electoral attributes are less important today in differentiating which campaigns have a website, but remain important determinants of the degree to which they provide more sophisticated content and use their website to engage and mobilize supporters (Gulati and Williams, 2007).

Internet diffusion depends on characteristics of the technology, including its relative advantage, compatibility, versatility and visibility. A fifth characteristic, ‘trialability’, refers to the ease of experimenting with an innovation as it is adopted (see Rogers, 2003). Social networks in particular benefit from a large number of members or from serving a specific niche, and they require technical expertise and specialized knowledge to be designed and leveraged effectively. Campaign staff members recognized some clear advantages social networks hold over traditional campaign tools. They often referenced the importance of Facebook’s niche: ‘We are trying to reach younger voters, college students, and young professionals who have graduated in the last five years … Most of the people who use Facebook are younger students aged 18−22.’ ³ Other campaigns were drawn by low cost: ‘The biggest benefit of Facebook is that it is free. So in terms of value, it is probably one of the best things that we can use to connect to voters.’ ⁴

That social networks are experiencing trialability is also evident in staff members’ assessments. ‘Right now the campaign is not ready to jump into the Facebook phenomenon. We’re taking a conservative approach and watching closely what other campaigns are doing to make changes in the next election.’ ⁵

New features are going to always come up, new programs will be developed for it. It was six degrees, then Friendster, MySpace, now Facebook. I think we are going to see this trend continue to grow, whether or not it’s with Facebook or something else I can’t say, but it will definitely be around. I think with more projects and more media attention, campaigns will have no choice but to use this type of technology in their strategies. ⁶

Media attention has given social networks enormous visibility (see, e.g., Vascellaro, 2006). For example, early press accounts generated significant coverage of MySpace (e.g. Keen, 2006; Kelly, 2006; Lovley, 2006), YouTube (e.g. Fairbanks, 2006; Wasserman, 2006), and Facebook (e.g. Grynbaum, 2006; Yahr, 2006). And campaigns have taken note: ‘Nationwide this year social media networking has been huge, and that is becoming obvious … to ignore that is foolish quite frankly.’ ⁷

Research hypotheses

We hypothesize that the constituency and electoral variables predicting which candidates updated their Facebook page in 2008 will mirror those that predicted web presence in the early days of Internet campaigning. Hypothesis 1 reflects the diffusion of innovation expectation that early adoption will be more common among organizations whose competitive environment offers incentives to innovate, costs not to do so, or is uncertain. Hypothesis 1 more explicitly recognizes the influence that characteristics of the user community (i.e. potential voters) have on early technology adoption decisions. In contrast with party-centered, proportional representation systems (Anstead and Chadwick, 2009; Carey and Shugart, 1995; Swanson and Mancini, 1996), US campaigns operate under a candidate-centered (Wattenberg, 1991), winner-take-all system. As a result, the affinity voters feel for individual candidates and how they relate to those personal characteristics matter in their voting decisions, and thereby become a factor in election campaign strategies and outcomes (e.g. Manzano and Sanchez, 2010; Ragsdale, 1980).

Finally, the diffusion of innovation literature suggests four interrelated hypotheses based on the profiles of adopter (or adopting organization) categories. The first is that those who have already demonstrated a propensity to adopt other recent technologies will be among the early adopters of the latest ones. The second recognizes that information, professional networks (Howard, 2006) and other exposure to a new technology will expedite adoption. In addition to positive peer influence, adoption by competitors constitutes a potential threat that will create pressure on candidates to copy what their rivals are doing. For purposes of this analysis, we operationalize such proximity contagion effects using the state as the geographically relevant peer/competitor group. Finally, the decision to adopt and the extent of adoption or implementation differ, with the latter more dependent on internal, organization-specific factors (Fichman, 2004).

Online social networking, 2006, 2008

For the 2006 elections, Facebook carved out a special space – ‘US Politics’ − on its network for all US congressional and gubernatorial candidates. Facebook initiated these profiles by designing a standard template with only the candidate’s name, office being sought, and basic contact information. Passwords that allowed the candidates to assume responsibility for personalizing their profiles were forwarded to the offices of the Republican and Democratic national committees, who then distributed them to their candidates. ⁸ After assuming control of their profile, candidates could initiate a discussion topic, post comments on their wall, and post notes, event information and videos and photographs. Facebook users who ‘friended’ a candidate as a way to show their support on their own profile could also post materials and comments. A candidate did not need to activate the profile for users to register their support and post content.

Facebook made some modifications for the 2008 elections. Political candidates were given pages instead of profiles. These pages were similar to personal profiles but offered the candidates greater capability to post various kinds of campaign material (e.g. announcements, links to other pages, YouTube links, notes, photo albums, and event information) and allowed their supporters to post their own materials (see Figure 1). A second change was to eliminate the ‘US Politics’ section and place all the candidates’ pages within a ‘Politicians’ sub-section of fan pages. Thus politicians were clustered near celebrities and other public figures, sports teams, films, restaurants, bars and clubs, products, non-profits, and other organizations. In addition, current elected officials and candidates for all levels of office in any country were eligible for ‘Politicians’ pages as long as an official representative of the politician created the page (see Figure 2).

OPEN IN VIEWER

Figure 1.

An example of Facebook’s fan pages for politicians.

OPEN IN VIEWER

Figure 2.

Facebook’s ‘Politicians’ section.

While the new format stood to increase the visibility of the candidates on Facebook, only 8 percent of Americans reported that they had used social networking sites (although not Facebook specifically) to learn about the campaigns during the 2008 cycle (Smith and Rainie, 2008). Our analysis in this paper examines how extensively congressional candidates in 2006 and 2008 used Facebook and offers insights into how this new tool was integrated into their overall campaign strategy, given their limited, albeit ever-increasing, reach with the American public.

Data and methods

Analysis and findings

In 2006, only 16 percent of major party candidates running for the House personalized their Facebook profile in some way. By 2008, a large majority of both Democratic and Republican candidates had a presence on Facebook. Among major party House candidates, 72 percent (591 out of 816) had a ‘Politicians’ page, and about half of those candidates updated it (406 of 816, or 49.8 percent), or a threefold increase from the previous election cycle.

Turning to Hypotheses 1 and 2, Table 1 presents the results of the multivariate logistic regression analyses of profile activation in 2006 and page creation by House candidates in 2008 (see Table 1, Model 1, ‘Activation 2006’ and Model 2, ‘Page creation 2008’). These results show that Democrats, challengers, better-financed candidates, and candidates running in competitive races were the most likely to update their Facebook profile. There is also evidence that candidates for open seat contests were more likely than incumbents to activate their profile in 2006 and create a page in 2008 (the coefficients in Models 1 and 2 have a positive sign), although the coefficients for Model 1 are not significant.

OPEN IN VIEWER

Table 1.

Multivariate logit analysis of Facebook presence in 2006 and 2008 House races.

Independent variables	Model 1: Activation 2006	Model 2: Page creation 2008	Model 3: Page update 2008
Party (Republicans = reference category)
Democrats	0.208	0.436 ***	0.629 **
	0.213	0.166	0.245
Incumbency status (incumbents = reference category)
Challengers	0.56 **	0.778 ***	4.752 ***
	0.261	0.254	0.784
Open seat candidates	0.266	0.599 *	3.803 ***
	0.368	0.322	1.108
Contributions received (natural log)	0.133 **	0.315 ***	0.507 ***
	0.057	0.077	0.177
Competitive seat	0.677 **	0.757 ***	0.948 ***
	0.289	0.231	0.334
Constituency variables
Percentage rural	0.000	0.003	0.004
	0.008	0.006	0.008
Percentage white	0.009	0.018 ***	0.018 **
	0.008	0.006	0.008
Percentage with college degrees	0.018	0.034 ***	0.036 **
	0.014	0.012	0.017
Percentage age 65 and older	−0.012	−0.052 *	−0.034
	0.034	0.028	0.039
State average presence	1.713 **	0.399	0.697
	0.678	0.538	0.76
Campaign website presence	0.753 *	0.616 **	0.608 *
	0.449	0.293	0.336
Intercept	−5.776 ***	−6.229	−9.013 ***
	0.985	1.165	2.502
n	802	815	591
Percentage correctly predicted	84.2	69.5	79.2
Mode	84.0	50.1	68.7
Pseudo R2	0.127	0.245	0.476

Note: Bold entries are unstandardized logit coefficients; standard errors are in italics. * p < .10, ** p < .05, *** p < .01; two-sided tests of significance.

The data presented in the third column of Table 1 include only the 2008 candidates who created a Facebook page and then updated it in some way (n = 591). Among those candidates who had a profile present in Facebook’s ‘Politicians’ section, Democrats, challengers, better-financed candidates and candidates running in competitive races were the most likely to update their Facebook profile. The effects for all of these variables are significant.

The Democrats’ greater enthusiasm for this social network seems to reflect differences in mobilization strategies that find Democrats more eager than Republicans to use the Internet as a way to communicate with their supporters, and is consistent with their larger number of supporters in the youth demographic. Although Republicans were early pioneers in using recent new technologies to identify Republican voters, Republican strategists and activists typically have worked within a top-down organizational structure and, as the party in power, tended to rely on communication and mobilization strategies that they pursued and successfully implemented in the past (see, e.g., Rasiej and Sifry, 2008; Sifry, 2006; Stirland, 2007; Thompson, 2008).

Other campaigns that are the most likely to embrace social networking sites are those that view this new communication medium as an additional tool for winning votes within specific blocs or the general electorate that constitute their user community. Challengers must find a way to overcome the advantages of incumbency, which allows members of Congress to draw upon an established network of supporters and contacts. Our interviews revealed that challengers in 2008 were looking to capitalize on every resource that could give their campaign an advantage or net them additional votes. But since over three-fourths of incumbents already have to maintain both an office and a campaign website, there is even less of a need to dedicate staff time to a third presence online.

When the race is more competitive, however, all candidates, regardless of incumbency status, may try to exploit every available technological resource to help them expand their electoral base and maximize turnout among their supporters. Since the cost of updating profiles and posting content is extremely low, we believe that better-financed candidates were more likely to activate a profile (2006) or create a ‘Politicians’ page (2008) on Facebook because campaigns with the most money also tend to have the most sophisticated and professional organizations, i.e. where professionalism and cosmopolitanism allow them to leverage the competitive environment to their advantage. In this setting, these are the campaigns that first see the Internet as a fundamental component of any communication and mobilization strategy (cf. Howard, 2006).

The data presented in the lower half of the first two columns of Table 1 are mixed and therefore only partially support Hypothesis 1(e−h). These results show that candidates running in districts with a higher percentage of college graduates were also the most likely to have used Facebook in both 2006 and 2008, even though the effect in 2006 was not statistically significant. Since Facebook’s membership was originally open only to those associated with colleges and universities, and continues to have a strong foothold in that demographic, it is no surprise that candidates running in districts with a high proportion of college graduates would want to experiment with Facebook as a way to communicate their message.

We see a similar result with respect to ethnic diversity, which also increased the likelihood of creating or updating a page in 2008, but had no statistically significant effect in 2006. It is not clear whether candidates running in districts with a younger population were more likely to update their profiles or not. The coefficients for age are statistically significant in only one of the models. In 2008, the presence of more senior citizens decreases Facebook page creation (see Model 2) but had no effect on updating (see Model 3). There was also no relationship between age and adoption in 2006. ¹⁶ The level of urbanization in the district had no effect in any of the three models. While there is evidence linking Internet use and broadband adoption to place of residence (Horrigan, 2009), candidates from rural areas may not have been deterred from activating a page, since simply viewing basic information on a page does not require the user to have a high-speed connection.

The last two variables in Table 1, Models 1 and 2, present the tests of two new categories of predictive variables suggested by the diffusion of innovation literature that political scientists who study campaign technology adoption have hitherto ignored. In 2006 and 2008, past experience with technology has large, positive and significant coefficients, and thus provides strong support for Hypotheses 2(a) and 2(b). The coefficients for geographic network are positive and statistically significant only in 2006, however. This suggests that while early adopters are making decisions more on peer influence and probably personal qualities and biases we are unable to measure, later adopters become more strategic and take other factors into account such as the expectations of voters in their geographic constituency.

Which campaigns use Facebook most extensively?

Turning to Hypotheses 2(c) and 2(d), Model 3 presents the first, lower-effort measure of the extent to which campaigns are implementing and using Facebook by analyzing those who have taken the step of posting some form of additional content to their pages after initially creating the page (n =591). Table 2 presents the results of our second high-effort measure of usage, Facebook activity, which counts the combined usage of four candidate- and supporter-generated features. We collected these data for only a subsample of House candidates who created pages, which accounts for the much smaller n of 152 after accounting for missing data.

OPEN IN VIEWER

Table 2.

OLS regression analysis of Facebook activity in the 2008 House races.

Independent variables	Model 4
Party (Republicans = reference category)
Democrats	0.142 *
	0.086
Incumbency status (incumbents = reference category)
Challengers	−0.086
	0.115
Open seat candidates	−0.293 **
	0.136
Contributions received (natural log)	0.108 ***
	0.042
Competitive seat	0.152
	0.113
Constituency variables
Percentage rural	0.006 *
	0.003
Percentage white	−0.003
	0.003
Percentage with college degrees	0.024 ***
	0.007
Percentage age 65 and over	0.014
	0.011
State average presence	−0.273
	0.258
Campaign website presence	−0.066
	0.178
Intercept	−0.793
	0.618
n	152
R2	0.192
Standard error	0.533
Standard deviation of dependent variable	0.606

Note: Bold entries are unstandardized OLS regression coefficients; standard errors are in italics. * p < .10, ** p < .05, *** p < .01; two-sided tests of significance.

The first finding of note is that the results for Model 3, page update mirror those for page creation in 2008. This indicates that merely undertaking to add one (or more) type of content to the page is indeed a very modest effort and low-bar criterion for extent of usage. If a campaign takes the more important step of creating a Facebook presence, they almost certainly return at some point to extend its content. Given that this third model achieves the highest explained variance, .48, it suggests that actually updating the page is likely a better indicator of what the concept ‘adopting’ an innovation signifies. We also believe it confirms that our summated index is a more stringent test of extent of usage and conclude that it is probably also a more appropriate one for capturing this diffusion of innovation concept in what we measure.

As a comparison of Table 1 with Table 2 shows, while there are some similarities across the sets of results, the differences between them strike us as the more important finding. The amount of contributions received is the only variable that is statistically significant across all four models. For the three 2008 models, only the coefficients for Democrats and the percentage of the college graduates living in the district are significant and have a positive sign.

Neither the geographic proximity network nor a propensity to adopt other technology innovations influences the degree to which campaigns implement and use its features. Consistent with the way the literature differentiates these and Hypothesis 2(c), our data substantiate that the two are clearly different decisions and subject to different processes. In retrospect, it suggests an explanation for the low explained variance reported in 2006 by Foot and Schneider (between .09 and .20) and also here in Model 4. Like ours, their dependent variables are features connoting extent of usage (in campaign websites through informing, connecting, involving and mobilizing). Political science research has not captured the constructs most related to extent of usage. We suspect that is because researchers have not hitherto correctly differentiated these two decisions in a way that would lead us away from simply appropriating the same conventionally employed independent variables for both research questions.

The second striking difference across models is the direction of influence of incumbency status. In the previous three models, which we now reconceptualize as the technology innovation adoption decision, challengers and open seat candidates were the early adopters. In Model 4, the technology innovation implementation decisions about how extensively it will be used, those coefficients now show incumbents to be the most active Facebook users in 2008. It is a finding that is consistent with a recent study of YouTube usage in 2008 (Gulati and Williams, 2010) and another study of campaign websites (Druckman et al., 2009). The diffusion of innovation literature and empirical studies suggest that organization contingent factors such as compatibility are what matter here. Incumbents have more capacity (organizational and resource wherewithal) to generate the content that constitutes (campaign postings) or creates (supporter postings) usage. As earlier comments of staff members illustrated, among House candidates, non-incumbents’ campaigns are heavily dependent on volunteers, have trouble keeping up, or do not have much happening from day to day about which to post.

The predictive power of the percentage of college-educated residents within House districts and the percentage of residents living in rural areas are two more interesting findings. It is the one instance where our findings are not wholly consistent with Hypothesis 2(d). Education could be viewed as corroborating the justifications and interpretations applied to it in the campaign website studies, which tie it to Internet usage behavior. Yet age and race/ethnicity continue to demarcate an important, albeit diminishing, digital divide in social networks more generally, and much more so in Facebook specifically. The diffusion of innovation literature suggests a new way of conceptualizing this result with respect to the question of usage. College education is an attribute of this technology in the sense that it was among those with a college education that Facebook originated and who comprise its most significant user base. Choosing between a diminishing digital divide (a maturation or developmental explanation) and this technology attribute explanation would require testing the same model with another differently constituted social network such as MySpace and across another type of technology innovation such as YouTube or Twitter.

Discussion and future directions

The diffusion of innovation literature served to inform our research hypotheses, model specification, analysis and interpretation of our data on who adopts new campaign technologies and to what extent. Our findings demonstrate the importance of differentiating between the decision to become an early adopter and the extensive implementation and use of the technology once it has been adopted. We anticipate that examination of the next step − assessing how well the technology performs, i.e. the quality and/or degree of its success − will likewise benefit from attending to these constructs in refining the measures, and for the identification of additional or different variables in specifying our models. We should not be using the same conventional variables for all three research questions.

In addition to the point that the selection of independent variables should be made in relation to the dependent variable under investigation, diffusion of innovation literature and our findings underscore that the interrelationships between these two will change depending on whether we are examining adoption, usage or performance. What competition ‘means’ and why it is important or not in the decision to be an early adopter or extensive implementer may change because our measures are only surrogates or imperfect operationalizations of their underlying constructs. In other words, and taking the example of financial resources, we may be capturing one kind of motivator in the adoption decision (e.g. an external one like norms and expectations surrounding high-visibility, high level of office campaigns) and the consultants who comingle among them or are historically associated with that state or district. In the innovation’s implementation, however, that same measure of financial resources may be capturing quite a different driver, namely an internal organization-specific facilitator or constraint in ways analogous to the findings and reasons that large firms (Rogers, 2003) − or states (Hage and Aiken, 1967; Mohr, 1969; Walker, 1969) − innovate more and better than smaller ones in the same industry. Our results also demonstrate that the same independent variable may be a significant influence on one decision (adoption) and not the other (implementation), or work in the opposite direction, as we saw with competitiveness of the race.

This study has identified two new independent variables that are important to early adoption: geographic proximity contagion and propensity to adopt campaign innovation technologies. Both measures could be expanded or refined in ways that the limitations of our data did not allow. For example, geographic proximity defined in terms of state borders may not be as appropriate or finely measured as a physical proximity distance measure. And it may be that social interaction networks are additionally or more important. Furthermore, the low percentage of variance explained in this and similar studies of campaign websites underscores the point that we have a long way to go in identifying variables that appropriately and fully specify our extent of usage models. As Fichman (2004: 54) observes, ‘Research that goes beyond the dominant paradigm holds more promise to tell us things about the IT innovation phenomenon that we do not already know.’ Part of the reason is that internal organization-specific attributes are difficult for political scientists to observe directly or publically and to measure quantitatively. Campaigns are reluctant or unwilling and too busy to reveal this information, and it is difficult to convert their responses and data into standardized measures comparable across campaigns. Our interviews of staff members offer tantalizing insights that enrich the interpretation of the findings in this study, but require more systematic and extensive replication before they can be employed to adequately test our explanatory models.

Social networking sites like Facebook go beyond simply communicating the campaign’s theme and providing information on how to participate in it. Active engagement by the candidate and a well-maintained site can make the candidate more accessible and seem more authentic. It also can encourage a more professional discussion among supporters. ¹⁷ In addition to personalizing the candidate, Facebook puts a face on the other supporters and facilitates interpersonal connections around activities other than politics. And because Facebook organizes members by regional and organizational networks and gives greater access to profiles in one’s own networks, offline meetings and connections are possible.