Digital Media and Traditional Political Participation ...

This article was downloaded by: [Lauren Copeland] On: 17 April 2013, At: 21:17 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Journal of Information Technology & Politics Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/witp20

Digital Media and Traditional Political Participation Over Time in the U.S. a

Bruce Bimber & Lauren Copeland

a

a

Department of Political Science, University of California-Santa Barbara Accepted author version posted online: 04 Feb 2013.Version of record first published: 16 Apr 2013.

To cite this article: Bruce Bimber & Lauren Copeland (2013): Digital Media and Traditional Political Participation Over Time in the U.S., Journal of Information Technology & Politics, 10:2, 125-137 To link to this article: http://dx.doi.org/10.1080/19331681.2013.769925

PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

Journal of Information Technology & Politics, 10:125–137, 2013 Copyright © Taylor & Francis Group, LLC ISSN: 1933-1681 print/1933-169X online DOI: 10.1080/19331681.2013.769925

RESEARCH ARTICLES

Digital Media and Traditional Political Participation Over Time in the U.S. Downloaded by [Lauren Copeland] at 21:18 17 April 2013

Bruce Bimber Lauren Copeland

ABSTRACT. Research shows that digital media use is positively related to political participation. However, this relationship does not appear in all studies. To date, researchers have generally treated inconsistent findings from study to study and from election to election as an empirical problem that reflects differences in measurement and model specification. In this article, we question the assumption that a consistent relationship between Internet use and political participation should exist over time. We test this expectation using 12 years of data from the American National Election Studies. Our findings support the expectation that a general measure of Internet use for political information is not consistently related to six acts of traditional political participation across elections.

KEYWORDS. Digital media, Internet, participation, political behavior, social media, voting

Since the late 1990s, researchers have published many studies about the relationship between digital media use and political participation.1 Most of this work focuses on whether people’s use of digital media increases the likelihood that they will engage in political participation offline. Many studies concur that people’s use of technology for political purposes has a small but positive effect on their level of traditional, offline political participation

(Boulianne, 2009). However, these studies have not resolved an important empirical problem: Are the relationships between the extent of digital media use and participation consistent across election years? Most of the literature relies on single cross-sections and rests on the implicit assumption that findings from one study generalize to other election cycles. However, the handful of studies that examine more than one cross-section find that the relationship between

Bruce Bimber is professor of political science at the University of California–Santa Barbara, where he is also affiliated with the Department of Communication. His research interests involve digital media and politics, with a focus on participation and the organization of collective action. Lauren Copeland is a Ph.D. candidate in the Department of Political Science at the University of California–Santa Barbara. Her research interests are in political behavior and political communication, with an emphasis on digital media, political participation, and political consumerism. Address correspondence to: Bruce Bimber, Department of Political Science, University of California, Santa Barbara, Santa Barbara, CA 93016-9420 (E-mail: [email protected]). 125

Downloaded by [Lauren Copeland] at 21:18 17 April 2013

126

JOURNAL OF INFORMATION TECHNOLOGY & POLITICS

digital media use and participation varies from year to year. Digital media use may predict voting or donating money in one election, for example, but not in another. Although we agree with Boulianne (2009) that measurement inconsistencies and differences in model specification contribute to this problem, we suspect that inconsistent findings other authors report from year to year are not entirely the result of differences in measurement or modeling. We also suspect that inconsistencies do not result only from a monotonic trend toward a stronger or more direct relationship over time, as several authors speculate (Cho et al., 2009; Xenos & Moy, 2007). In the discussion that follows, we respond to the widely employed assumption in the literature that relationships between digital media use and participation should, apart from measurement error, be consistent or strengthening from year to year. That assumption emphasizes technology use itself as the driver of behavior. Instead, we observe that the political content people encounter online likely drives their participation, and that measures of the extent to which people use technology are a poor proxy for understanding this. In principle, survey measures emphasizing what people see and experience online are possible, but historically surveys have emphasized use of the medium much more than the nature of the message. For this reason, we expect that the single-year relationships between technology use and participation that have been reported over the last decade or so will not be consistent and robust longitudinally. In this empirical project, we test our expectation using American National Election Studies (ANES) data from 1996, 1998, 2000, 2004, and 2008. Like other scholars, we find a variety of positive associations between using the Internet to obtain political information and offline political participation over these years; we also find that these relationships are not robust over time. Patterns in the relationship between political use of the Internet and participation vary across specific behaviors, with some behaviors (e.g., voting) showing idiosyncratic variation from one election to another. For at least one behavior, attempting to persuade others, we find some evidence of a trend toward a

stronger association between digital media use and participation over time. Overall, these findings support the idea that single cross-sections measuring general Internet use—of the kind that have dominated the literature—may not offer generalizable results.

THEORETICAL ISSUES Many studies report a positive relationship between the extent of digital media use and political participation (e.g., Bakker & de Vreese, 2011; Bimber, 2001, 2003; Boulianne, 2009, 2011; Cho et al., 2009; DiMaggio, Hargittai, Neuman, & Robinson, 2001; Gainous & Wagner, 2011; Gibson, Wainer, & Ward, 2005; Mossberger, Tolbert, & McNeal, 2008; Prior, 2007; Shah, Cho, Eveland, & Kwak, 2005; Shah et al., 2007; Tolbert & McNeal, 2003; Xenos & Moy, 2007). In her meta-analysis of this literature, Boulianne (2009) examines 38 studies, finding on the whole a small, positive effect of Internet use on participation. Generally speaking, these studies conclude that the more people use digital media, the more likely they are to participate in politics. Moreover, political interest and political discussion often mediate these relationships (Cho et al., 2009; Shah et al., 2005, 2007; Xenos & Moy, 2007). An implicit assertion in much of the literature is that findings from single cross-sectional surveys generalize well over time, providing meaningful empirical support to general theories of digital media and politics. This is due to the assumption that there is a stable underlying relationship between general use of digital media and participation. However, a handful of researchers note that relationships appearing in one election cycle may not be present in another. For example, Bimber (2003) finds that political use of the Internet predicted voting in 2000 but not in 1998. Tolbert and McNeal (2003) examine three election cycles and reach the same conclusion, but add that digital media use predicted voting in 1996 and in 2000. They also find that Internet use predicted an index of eight types of participation in 1998 and 2000, but not in 1996. Similarly, Cho et al. (2009) find that online political advertising exerted a greater


Bimber and Copeland

effect on participation in 2004 than it did in 2000. These discrepancies have not been addressed adequately. The most common explanation for inconsistent findings has been variability in measurement or model specification. Boulianne (2009) makes a persuasive case that some yearto-year variation in the relationship between extent of digital media use and participation results from such causes. Her meta-analysis documents considerable differences in findings across studies. While some researchers find a small positive effect of Internet use on participation, others find no effect. Still other studies find a negative relationship between digital media use and political participation. Her analysis shows that studies are much more likely to show a positive relationship between Internet use and participation when they do not control for political interest. Although Boulianne’s (2009) analysis suggests that differences in data quality and model specification explain inconsistent findings among studies, this explanation does not address those studies that have examined more than one year using identical models with controls for political interest (Bimber, 2003; Cho et al., 2009; Tolbert & McNeal, 2003). Another explanation suggested by some observers is that digital media use has become more strongly associated with participation over time, such that studies examining more recent elections should produce different results than those that examine earlier ones. This explanation makes sense intuitively, especially given the revolution in social media in the mid2000s. Although several authors speculate about this hypothesis, none have attempted a systematic test. For instance, Xenos and Moy (2007) write that “as new media become more deeply integrated into everyday life, previously contingent relationships may give way to more direct effects as different facets of Internet communication continue to diffuse and usage patterns evolve” (p. 715). Similarly, in a comparison of results from 2000 and 2004, Cho et al. (2009) suggest “the growing role of the Internet between these elections” (p. 79). Boulianne (2009) sees increasing effects of the Internet on participation over time, but finds

127

that these effects are not monotonic: the proportion of studies with significant effects is higher in 1998–1999 and 2004–2005 than during the intervening period. It is unclear whether this reflects an interrupted upward trend in the relationship between digital media use and offline participation, or whether it reflects variability across periods or individual elections. Tolbert and McNeal (2003) have suggested the latter; they argue digital media use plays a larger role in presidential election years than it does in midterm election years. At this point, one can summarize the state of the literature as follows. A decade-long search has been undertaken to identify a relationship between digital media use and participation. Researchers have generally assumed that if such a relationship exists, it should be stable over time and across studies and therefore should provide a valid indication of underlying causal mechanisms. This assumption has been central to many theoretical discussions and much empirical research about the Internet. There are, however, a number of clues in the literature that this assumption may be incorrect; it may be that relationships change substantially from year to year, or that they exhibit an upward trend over time. These possibilities have not been subject to a test that is designed to look for consistency over multiple elections. If it is the case that relationships vary from year to year, then future survey work would have to take this variation into account; by the same token, future theorizing about the Internet would benefit from a refined understanding of the possibility that relationships vary from year to year. We believe that the assumption of a stable, or growing, relationship between general measures of digital media use and participation is unwarranted in the U.S. That assumption prioritizes technology itself as the driver of behavior, rather than as a channel for the communication that is associated with behavior (Vaccari, 2010). Overall use of digital media tools does not fluctuate greatly over time; however the nature of communication and information to which any one person is exposed through digital media can indeed vary greatly across time and political events. In contrast, surveys have generally measured only such factors as whether people obtain


128


political information online, or how frequently people use social media. The content-based drivers of variation in online political experience are not necessarily strongly correlated with how frequently or intensively people use digital media in general. Bakker and de Vreese (2011) provide some empirical support for this idea, showing that whether people use the Internet, as well as other media, is related to political participation, but frequency of use is not. Stronger evidence for the inability of simple survey measures of technology use to capture the influence of online experiences comes from the remarkable study of Bond et al. (2012) reported in Nature. In collaboration with Facebook, political scientists conducted a randomized, controlled experiment with every U.S. adult who accessed Facebook on election day in 2010, for a total N of 61 million. In the first treatment, one-third of subjects received a message urging them to vote. In the second treatment, one-third of subjects received the voting message along with images of six of their own friends who reported having already voted. A control group received no message. Using public voting records to measure actual voting, the researchers show that the information-only group was no more likely to vote than the control group, but the social message group with friend photos was 0.39% more likely to vote than the control group. Their study demonstrates one mechanism by which the nature of people’s experience online, in this case the social context of political messages, can affect participation in ways that are not captured by measures of how often they use the Internet or Facebook. It is beyond the scope of this project to hypothesize or test all the myriad potential influences of the content of digital media on participation, such as the interaction between receipt of a political message in Facebook and the extent to which social networks provide a context for that message. Presumably a wide range of mobilization effects, network effects, information effects, and selectivity effects are at work, and presumably these vary in complex ways across people who are online as well as across time. We do not expect that traditional measures of the extent of Internet use or even political Internet use do not tap into these effects. For that reason, we doubt

that whatever associations between general measures of Internet use and participation appear in survey data actually tap into underlying relationships that are stable and meaningful. The campaigns of Barack Obama for president illustrate another aspect of the contextual character of digital media. In 2008, Obama’s two major opponents, Hilary Clinton in the primary race and John McCain in the general election, both used a variety of digital media tools, including social media such as Facebook and YouTube. Obama was vastly more successful with technology for several reasons: because of superior expertise with and investment in digital media, because of his personal appeal as a candidate, and because of his campaign’s capacity to channel much of the social movement–like excitement of his supporters through digital media. The fact that many Obama supporters used digital media may have been a necessary condition for their involvement with this campaign, but it was not sufficient. As Vaccari (2010) writes, the impact of digital media on Obama supporters’ behavior was “contingent upon the candidate’s personality and message, and in turn by their ability to generate movement-like enthusiasm” (p. 329). Digital media provided channels for political communication but did not themselves drive participation. The 2012 presidential race further illustrates this idea. In his re-election bid, Obama employed a similar array of technologies to those of 2008, enhanced by a massive data analytics effort. Yet the social movement– like enthusiasm of 2008 had dissipated. While Obama won the 2012 election, he did so with fewer votes than he received four years earlier, in a context of reduced turnout overall and a contest that was highly concentrated in a few swing states. Nationwide, Obama took a smaller share of younger voters in 2012 than in 2008, yet in swing states where he focused his technology-driven mobilization efforts, his share of voters under 30 increased over four years earlier (Bimber, 2012). Thus, the state in which younger citizens resided may have had an effect on whether their digital media use was connected with turnout in 2012. These observations lead us to our primary expectation in this study: when measurement

Bimber and Copeland

and model specification are consistent, and multiple cross-sections are compared, there should be no stable relationship between use of the Internet for political information and political participation over time in national samples.


METHOD AND DATA We use American National Election Studies (ANES) data to test our expectation. The discovery of an association between Internet use and traditional political participation in the U.S. is attributable in large part to the existence of the ANES. These data have been important in support of years of theorizing about mechanisms linking digital media use to participation. Although the ANES measures of Internet use are quite thin and have not been elaborated upon over time by survey managers, we believe the ANES has the best available data to examine over-time change in the relationship between digital media use and participation. The ANES data provide a high degree of consistency across years, such that we can employ identical models and measures for multiple elections across a variety of political actions. Consequently, we can reduce error from measurement inconsistency to the smallest practical level, since data come from identically worded questions over time.2 In our analysis, we revisit the basic approach to modeling Internet use and traditional political participation that others have used when analyzing ANES data, expanding on that work by examining all five elections from 1996 to 2008 for which data on Internet use are available. These are 1996, 1998, 2000, 2004, and 2008. In 2002, ANES did not include Internet measures, and in 2006 no time-series survey was fielded.

Dependent Measures We modeled six dependent variables using logistic regression: displaying a button or bumper sticker; attending a meeting or rally; working for a party or campaign; donating money to a candidate, party, or other political group; attempting to persuade others; and voting. The wording of the question about displaying a message is as follows: “Did you

129

wear a campaign button, put a campaign sticker on your car, or place a sign in your window or in front of your house?” For attending a meeting or rally, the ANES asked: “Did you go to any political meetings, rallies, speeches, dinners, or things like that in support of a particular candidate?” Respondents could also indicate whether they worked for a political party or campaign by answering the following question: “Did you do any (other) work for one of the parties or candidates?” For donating money, we combined three measures of donating to create a dummy variable coded “1” if the respondent donated money to a candidate, a party, or a group.3 Finally, to assess whether respondents voted in each general election, the ANES asked: “In talking to people about elections, we often find that a lot of people were not able to vote because they weren’t registered, they were sick, or they just didn’t have time. How about you—did you vote in the elections this November?” All questions for all years had dichotomous responses. Rather than using a single aggregate dataset with variables for “year” or “year × Internet use” interactions, we created separate models for each year. This approach multiplies the number of models, but permits easier comparisons with results of other studies that model years separately, provided proper consideration is given to experiment-wise error. For each year we created six models (one for each specific action), and we repeated this for each of the five years of unweighted ANES data. Table 1 shows the frequency of participation over time. In the table, the upper row for each of the six acts displays the percentage of all respondents in the ANES who reported participating. The figures show generally flat or increasing participation over time. There is little variation from 1996 to 2000 in the frequency of attending events (6%), followed by a modest rise to about 9% by 2008. The pattern is similar for working on a campaign, which ranges from 2% to 3% through 2004, with a slight rise in 2008 to about 5%. The frequency of displaying a message rose more substantially, from about 10% through 2000 to roughly 20% in 2004 and 2008. Attempting to persuade others also grew during this time, from about 29% in 1996 to a high of 49% in 2004. For donating money, frequencies vary throughout the period between 10%

130


TABLE 1. Frequency of Participation in Six Political Acts, 1996–2008


1996 Voted All people Saw political info. online p Displayed message All people Saw political info. online p Attended event All people Saw political info. online p Did campaign work All people Saw political info. online p Donated money All people Saw political info. online p Persuaded others All people Saw political info. online p

1998

2000

2004

2008

76.6 90.2 0.001

53.8 64.2 0.014

76.1 82.5 < 0.001

78.5 86.6 < 0.001

76.3 86.0 < 0.001

10.2 19.6 0.001

6.2 9.7 0.087

10.0 11.2 0.047

20.6 25.1 0.001

20.2 24.5 < 0.001

5.9 16.7 < 0.001

5.6 10.6 0.010

5.5 6.5 0.024

7.6 10.6 < 0.001

9.3 14.0 < 0.001

2.7 5.9 0.045

2.0 3.3 0.274

2.8 3.3 0.105

3.4 4.8 0.015

4.7 7.6 < 0.001

12.9 23.5 0.001

10.3 22.0 < 0.001

12.2 14.8 < 0.001

14.9 20.4 0.004

11.7 19.7 < 0.001

28.9 41.2 0.004

20.0 32.3 < 0.001

35.1 39.9 < 0.001

48.5 57.9 < 0.001

43.0 55.4 < 0.001

Note. Values are percentage of all people engaging in each political act and percentage of those people who saw political information online who engaged in each act. p values are from χ 2 tests. See the text for interpretation of multiple significance tests. Source is ANES. Note that no survey was fielded in 2006, and in 2002 no Internet measures were included in the survey.

and 15%, with no rise in 2008 despite the effectiveness of the Obama campaign at soliciting a large number of individual donations that year. For voting, the well-known over-reporting phenomenon in ANES data masks an actual upward trend in voting over this period: about 76% to 78% of ANES respondents report voting in the four presidential election years, while actual voting-age turnout figures for presidential contests from 1996 through 2008 are 48.1%, 50.0%, 55.4%, and 56.9%, respectively (McDonald, 2011).

Independent and Control Measures Following the general approach of the crosssections of Bimber (2003) and Xenos and Moy (2007), our key independent variable is use of the Internet for political information. To measure whether respondents saw political information online, the ANES asked, “Have you seen any information about this election campaign

on the (Internet/the Web)?”4 In the final models reported here, we included the following as control variables: education, age, income, gender, party contact, political interest, and internal political efficacy. We also experimented with other variables to test for sensitivity of our findings to changes in model specification, and we found that our results are robust.5 We employed the standard ANES variables in which respondents were asked indicate their level of education, their age in years, and their annual household income. To assess whether respondents were contacted by a political party, the ANES asked: “As you know, the political parties try to talk to as many people as they can to get them to vote for their candidate. Did anyone from one of the political parties call you up or come around and talk to you about the campaign this year?” The ANES political interest question is: “Some people seem to follow what’s going on in government and public affairs most of the time, whether there’s an election going

Bimber and Copeland

on or not. Others aren’t that interested. Would you say you follow what’s going on in government and public affairs most of the time, some of the time, only now and then, or hardly at all?” For internal political efficacy, the ANES asked respondents to indicate the extent to which they agreed or disagreed with the following statement: "People like me don’t have any say about what the government does."


RESULTS We begin with bivariate correlations to establish which political acts are correlated with seeing political information online over time. There are several ways to examine this association; we report here participation rates for people seeing political information online. These figures are reported in the middle rows for each act in Table 1. The data show that in every year, seeing political information online is positively associated with participation. For instance, in 1996, about 17% of those who saw political information online attended a political event, compared with the figure of 6% for the population as a whole (p < 0.001). In 2008, about 8% of people who saw political information online worked on a campaign, compared with about 5% of the population as a whole (p < 0.001). Chi-square tests are significant at the p < 0.05 level in every year for four acts: voting, attending an event, donating money, and persuading others. For working on a campaign, the correlations are significant in every year but 1998 and 2000, and for displaying a message in every year except 1998. Because of the large number of tests in Table 1 (30), it is helpful to consider statistical significance using a correction for experimentwise error. The most conservative approach with respect to avoiding Type I error is the Bonferroni technique of treating the study as an experiment containing a single family of 30 comparisons. In this case, the experiment-wise (family-wise) error rate is 1 – (1 – 0.05)30 = 0.79. We can correct for this by employing a more stringent significance level of 0.05 / 30 = 0.00167 for each test. Using this conservative correction, most of the correlations remain significant, with the exceptions of persuading others in 1996,

131

voting and attending an event in 1998, attending an event and displaying a message in 2000, and working on a campaign in 1996 and 2004. With so many positive results (i.e., rejections of the null hypothesis), we are not concerned with false negatives and conclude that there is a robust bivariate association across time between seeing political information online and these acts. This comes as no surprise. Next we move to the multivariate analysis for 2008, which shows that seeing political information online is a significant predictor (p < 0.05) for four of the six actions: attending a political event, working for a candidate or campaign, donating money, and attempting to persuade others. These single-year results are generally consistent with reports in the literature from single cross-sections in other years, such as that of Xenos and Moy (2007) who analyze 2004 ANES data, and Prior (2007) who examines 2000 ANES data. Testing our expectation about lack of consistency over time requires comparing these results with those for 1996, 1998, 2000, and 2004. We find that the set of relationships present in 2008 does not appear in any of the other years. In fact, there is no pair of years in which the Internet measure predicts the same set of political acts. For example, in 2004, seeing political information online predicts persuading, attending an event, and voting (p < .05). In 2000 it predicts only voting, in 1998 only donating, and in 1996 just voting and attending an event. Table 2 shows a simplified version of the results, reporting the coefficients and p-values for the Internet variable for each year and each act. Full models are available from the authors upon request, but are omitted here for brevity. Correction for experiment-wise errors due to the large number of models only strengthens this result. If we employ a conservative approach with respect to Type I error by treating the entire experiment as a single family and employ 0.0016 as a significance threshold, as above, eight of the ten relationships are no longer significant, leaving only persuading in 2008 and voting in 2000. In light of the possibility of Type II error, it is useful to explore a less stringent test. This can be done by treating each year as a single family of experiments. With five families

132



TABLE 2. Coefficients for Seeing Political Information Online in Models Predicting Six Actions, 1996–2008

Vote Coef. (SE) p Display message Coef. (SE) p Attend event Coef. (SE) p Campaign work Coef. (SE) p Donate money Coef. (SE) p Persuade others Coef. (SE) p

1996

1998

2000

2004

2008

0.79 (0.39) 0.041

0.24 (0.24) 0.308

0.74 (0.22) 0.001

0.45 (0.21) 0.031

0.32 (0.21) 0.126

0.52 (0.30) 0.086

0.59 (0.38) 0.119

0.07 (0.29) 0.811

0.35 (0.19) 0.060

0.22 (0.19) 0.232

0.72 (0.34) 0.035

0.52 (0.36) 0.144

0.3 (0.46) 0.071

0.57 (0.29) 0.050

0.62 (0.27) 0.021

0.40 (0.50) 0.422

0.34 (0.60) 0.568

0.86 (0.55) 0.118

0.060 (0.43) 0.160

1.21 (0.41) 0.003

0.49 (0.29) 0.091

0.81) (0.23 0.004

0.44 (0.33) 0.183

0.24 (0.21) 0.262

0.7 (0.24) 0.004

0.20 (0.24) 0.411

0.27 (0.23) 0.242

0.31 (0.20) 0.119

0.46 (0.15) 0.003

0.53 (0.16) 0.001

Note. “Coef.” is logistic regression coefficient for the variable Seeing Political Information Online in models predicting the six acts. Control variables in the models are omitted from the table: education, age, income, sex, contact by party, political interest, efficacy. Results are not sensitive to small changes in model specification, such as inclusion of variables for following campaigns on television or in newspapers. Diagnostics indicate that multicollinearity is not a problem in the models. Bold indicates significance at 0.05 level; bold italics indicates significance at Bonferroni-corrected 0.00167 level. Source is American National Election Studies, 1996–2008.

of experiments, a simple correction for familywise error involves using p = 0.05/5 = 0.01 as a significance threshold in each hypothesis test. At this level, the variable for seeing political information online is significant in six cases: persuading others in 2004 and 2008, working on a campaign in 2008, donating in 1998 and 2008, and voting in 2000. This approach produces a greater number of positive results, but does not produce a consistent pattern in the relationship over time for any of the six acts. Interpretation of the data is aided by examining predicted probabilities of engaging in each political act. Figure 1 displays these probabilities, with all other values held at their means, over time and across the political acts. The graphs compare predicted probabilities for people who saw political information online with those who did not. In all years, probabilities

of acting politically are higher for those who saw political information online. Only some of these differences reach statistical significance, and the pattern is generally consistent with that of the significance of coefficients from the logistic regression models displayed in Table 2. The graphs in Figure 1 reveal three main classes of results across the six acts: (a) the absence of a stable relationship between Internet use and participation, (b) idiosyncratic variation from year to year, and (c) a possible trend toward a stronger relationship over time. The first result, namely the absence of a relationship between Internet use and participation, appears in the case of displaying a message. The ANES measure for displaying a message asks respondents whether they wore a campaign button, put a campaign sticker on their car, or

Bimber and Copeland

133

FIGURE 1. Predicted probability of participation for people who saw political information online, 1996–2008. Solid lines show predicted probability of engaging in each political act for people who saw political information online, with all other variables held constant at their means. Dashed lines show predicted probabilities for those who did not see political information online. Vote

1.00 0.90

0.50

0.80

0.40

0.70

0.30

0.60

0.20

0.50

0.10

0.40 1996

1998

2000

2004

2008

Attend Event

0.60


0.60

0.00 1996 0.60

0.50

0.50

0.40

0.40

0.30

0.30

0.20

0.20

0.10

0.10

0.00 1996

1998

0.60

2000

2004

2008

Donate Money

0.00 1996 0.60

0.50

0.50

0.40

0.40

0.30

0.30

0.20

0.20

0.10

0.10

0.00 1996

1998

2000

2004

2008

0.00 1996

Display Message

1998

2000

2004

2008

Campaign Work

1998

2000

2004

2008

Persuade Others

1998

2000

2004

2008

Saw Political Info.Online Did Not See Political Info.Online Difference is Significant at 0.05.

displayed a sign at their home. Seeing political information online does not predict displaying a message in any election year, and the difference in predicted probabilities of engaging in this act between those who did and did not see political information online are not significant at any point in the time series. Variation in Internet use for political information is not associated with this form of traditional political participation. The second result in the predicted probabilities is a relationship between Internet use and participation that appears in some years but not others, absent any trend over time. Donating

money is the first example of this idiosyncratic variation. The Internet variable predicts donating in 1998, a midterm election year, but not again until 2008, the year of the Obama/McCain contest. In both elections, the change in predicted probability between people who saw and did not see political information online is six percentage points. There is no obvious general explanation for why the relationship appears in 1998. The Federal Election Commission did not authorize online donations until 1999, and so 2000 was the first election in which people could donate money online (Bimber & Davis, 2003).


134


It is likely, therefore, that this association in the 1998 ANES data reflects some underlying relationships that the ANES measure did not capture. The election of 2008 was special with respect to online campaign donations because of Barack Obama’s abandonment of federal election funds and his massive success at fundraising from individuals online. This may account for some of the reason that 2008 stands out with respect to the Internet and political donations. Like donating, voting shows an unpredictable relationship with Internet use. Seeing political information online increases the predicted probability of voting in a statistically significant way by 8% in 1996, 10% in 2000, and 6% in 2004. Although 2008 was an important election year for the use of digital media, the difference in predicted probabilities is not significant. We attribute this finding to some particular distribution of campaign effects from persuasion and mobilization efforts across demographic groups and locations, and other unmeasured factors in the elections. The third result in the predicted probabilities appears in the case of persuading others. The variable for attempting to persuade others exhibits a statistically significant relationship with the Internet variable in the last two elections, 2004 and 2008, suggesting an upward trend. Differences in predicted probabilities grow monotonically over time, from a nonsignificant 4% in 1998 and 7% in 2000 to a statistically significant 11% in 2004 and 13% in 2008. This trend is likely the result of at least two factors. One is the emergence of social media during the mid-2000s, which are especially conducive to political discussion. Some social media were available in 2004, such as Friendster, Flickr, Meet Up, and MySpace. By 2008, the list had expanded dramatically to include Facebook and Twitter, as well as many others. The efforts of the Obama campaign in 2008 to reach citizens through digital media is a second factor that likely contributed to the increasing importance of digital media use for political discussion for some citizens in some places. The results for attending a political event and working on a campaign support the idea that 2008 was a special year. For both acts,

the change in predicted probabilities between those who did and did not see political information online was significant only in 2008. This may reflect the emergence of social media in the mid-2000s in a way that changed the landscape of political participation. If so, then in future elections Internet use is likely to remain associated with these acts. On the other hand, it may be that 2008 was a special case due to the novelty of social media and the nature of the Obama campaign. If so, data from future elections may show that persuading, attending events, and working on campaigns exhibit more idiosyncratic variation from election to election, as has been the case with donating money and voting.

DISCUSSION Was there a relationship between using the Internet for political information and political participation in the U.S. during the period of Internet diffusion into politics since 1996? As the first longitudinal report to examine more than three elections, and one of only a few to examine more than a single election, we find that the answer is “no” for the period from 1996 to 2004, and “it depends” for the period from 2000 to 2008. Our findings suggest that inconsistencies between data sets or differences in model specification do not fully explain why other studies have not reported consistent results across years. As we expected, our analysis show that relationships between seeing political information online and some acts of participation, namely voting and donating money, vary idiosyncratically. This finding casts doubt on the utility of using single-year or two-year cross-sections to test general theories about the underlying relationship between digital media use and voting or donating money. By contrast, we find evidence for a strengthening relationship between Internet use and persuading others for the 2000–2008 period. This is consistent with the speculation of Cho et al. (2009) and Xenos and Moy (2007), although it does not appear for any other political act. Our findings on the relationship between seeing political information online


Bimber and Copeland

and attending a political event or working on campaigns are unclear; Internet use predicts these acts in 2008 only. Clearly the 2008 election is an important case. Seeing political information online in the national sample was associated with more political acts in 2008 than in any previous year: persuading others, donating money, attending a political event, and working on a campaign. With the exception of persuading others, it is not clear whether 2008 constitutes an exceptional election or evidence for an upward trend, because it is the last year in our series. One difference between 2008 and previous election years was the arrival of social media, which are here to stay and which will continue to be important in political campaigns to come. Another difference is the unusual character of the Obama campaign, which employed a variety of then-novel strategies and which generated social movement–like excitement among many voters. Turnout also peaked in 2008, reflecting the enthusiasm of many voters for change. Data from future election cycles will be necessary to establish whether 2008 was an outlier or the first election of a new era of digital media in politics. Whichever is the case, our results suggest that the most compelling possibilities for finding meaningful, persistent relationships between digital media use and traditional participation involve behaviors that entail social actions like talking about politics, volunteering, and attending events or meetings. This possibility is not surprising because the content of digital media use increasingly centers around social information and networks. The evidence in ANES data suggests that these kinds of actions are the best candidates for influence by digital media, and this suggestion is consistent with claims that the nature of citizenship is changing toward an increased emphasis on expression, individual voice, volunteership, and other non-institutional actions (Bennett, 1998; Dalton, 2008; Inglehart, 1997). Nevertheless, our findings are confined to ANES data. These data offer the opportunity for a five-election comparison to test whether findings in the literature based on individual cross-sections are generalizable. We are aware of no other dataset in the U.S. or elsewhere

135

that offers this opportunity. But, the ANES is of increasingly diminished relevance to contemporary media because it has failed to incorporate sufficiently robust measures of digital media use. It may be that some of the failure of the ANES data to show consistent relationships is due to poor measurement of a changing phenomenon. One concern associated with studies of the effect of digital media use on participation is endogeneity: use of digital media for political purposes is endogenous with political interest and offline participation. This constitutes a problem of statistical inference for research concerned with isolating the causal influence of digital media use on participation. The ANES measure we employed in this study, seeing political information online, is especially endogenous in these participation models—more so than would be measures of social media use or generic Web use, for example. In the present study, we accept the fact of two-way influence: that digital media serve as tools of action for those who are already inclined to participate, and that they also may exert a positive influence on participation. We include control variables for political interest and whether people were contacted by a candidate or party, but we do not anticipate that these resolve the concerns of those scholars who are interested specifically in causation in one direction. Our goal has been different: to inquire whether stable relationships exist between use of the Internet and participation over time, regardless of whether causation works one direction, the other, or both. We expected to find that positive associations do exist but that these reflect underlying, unmeasured influences of the specific nature of people’s experiences online. The data are consistent with this expectation. Fortunately, better longitudinal data sets for the purposes of studying digital media are being built in the U.S., as well as in other countries. These datasets are designed to capture the nuances and varieties of contemporary digital media use as well as new forms of participation and civic engagement that exploit them. We are intrigued to know what the data will say after a few more election cycles. On the basis of the present study, we view the extant literature in a new light, namely that


136


the observed associations between digital media use and traditional participation that others have reported in fact vary from year to year and event to event, most likely because they are driven by underlying, unmeasured factors. It has yet to be the case, at least in the ANES data, that an association exists between Internet use for political information and any political act for three elections in a row. We suspect that while one can safely predict that various forms of Internet use will be associated directly and indirectly with various forms of political behavior in the future, it will be difficult to predict these reliably using general measures of digital media use. This suggests new ways forward. Both empirically and theoretically, researchers could benefit from approaches to Internet use that address the content and context of online experience, rather than the broad-brush concepts of the past, such as frequency of Internet use or of social media use, or whether citizens have seen political information online. We believe it is important to conceptualize digital media not as a steady, continuous, or uniform influence on behavior that can be measured reliably at the individual level, but in terms of changed context for political communication and information. Conceptualizing and measuring that communication and information, rather than the channels through which it flows, are likely to lead to improved insights about how the experiences of digital media use affect political behavior.

NOTES 1. Original data from the American National Election Studies are publicly available from the International Consortium for Political and Social Research at http://www.icpsr.umich.edu/icpsrweb/ICPSR/. In addition, replication data for only those variables employed in this study are available at http://dvn.iq.harvard.edu/dvn/ dv/jitp. 2. See Note 4 for a minor exception. 3. The question wording for these three binary questions was as follows: “During an election year people are often asked to make a contribution to support campaigns. Did you give money to an individual candidate running for public office? Did you give money to a political party during this election year? Did you give any money to any other group that supported or opposed candidates?” We coded

our measure 1 if the respondent answered “yes” to one or more of these questions, and 0 if the respondent answered “no” to all three. 4. There was a wording change in 2008, when the measure for seeing political information online changed from a dichotomous question to a frequency question. 5. In one variation, we included measures of watching programs about the campaigns on television and reading about the campaigns in newspapers, although these are not available for 1998. Including these measures produces only trivial differences in coefficients on the Internet variable and in predicted probabilities. We also tested our model for only people aged 35 and under. For this group, there are fewer significant relationships with the Internet measure, and the overall patterns are similar.

REFERENCES Bakker, T., & de Vreese, C. (2011). Good news for the future? Young people, Internet use, and political participation. Communication Research, 38(4), 451–470. doi:10.1177/0093650210381738 Bennett, L. (1998). The uncivic culture: Communication, identity, and the rise of lifestyle politics. PS: Political Science & Politics, 31(4), 740–761. Bimber, B. (2001). Information and political engagement in America: The search for political effects of the Internet. Political Research Quarterly, 54(1), 53–67. doi:10.1177/106591290105400103 Bimber, B. (2003). Information and American democracy: Technology in the evolution of political power. New York, NY: Cambridge University Press. Bimber, B. (2012, November). Digital media in the Obama campaigns of 2008 and 2012. Keynote lecture presented at the Comparing Online Democracy and Elections Conference, London, United Kingdom. Bimber, B., & Davis, R. (2003). Campaigning online: The Internet in U.S. elections. New York, NY: Oxford University Press. Bond, R., Fariss, C., Jones, J., Kramer, A., Marlow, C., Settle, J., & Fowler J. (2012). A 61-million-person experiment in social influence and political mobilization. Nature, 489, 295–297. Boulianne, S. (2009). Does Internet use affect engagement? A meta-analysis of research. Political Communication, 26(2), 193–211. doi:10.1080/10584600902854363 Boulianne, S. (2011). Stimulating or reinforcing political interest: Using panel data to examine reciprocal effects between news media use and political interest. Political Communication, 28(2), 147–162. doi:10.1080/10584609.2010.540305 Cho, J., Shah, D. V., McLeod, J. M., McLeod, D. M., Scholl, R. M., & Gotlieb, M. R. (2009). Campaigns, reflection, and deliberation: Advancing an o-s-r-o-r model of communication effects. Communication


Bimber and Copeland

Theory, 19(1), 66–88. doi:10.1111/j.1468-2885.2008 .01333.x Dalton, R. J. (2008). Citizenship norms and the expansion of political participation. Political Studies, 56(1), 76–98. doi:10.1111/j.1467–9248.2007.00718.x DiMaggio, P., Hargittai, E., Neuman, W. R., & Robinson, J. P. (2001). Social implications of the Internet. Annual Review of Sociology, 27, 307–336. Retrieved from http://www.jstor.org/stable/2678624 Gainous, J., & Wagner, K. (2011). Rebooting American politics: The Internet revolution. New York, NY: Rowman & Littlefield. Gibson, R. K., Wainer, L., & Ward, S. J. (2005). Online participation in the UK: Testing a “contextualised” model of Internet effects. British Journal of Politics and International Relations, 7(4), 561–83. doi:10.1111/j.1467-856X.2005.00209.x Inglehart, R. (1997). Modernization and postmodernization: Cultural, economic, and political change in 43 societies. Princeton, NJ: Princeton University Press. McDonald, M. P. (2011). Voter turnout. United States Elections Project. Retrieved from http://elections.gmu. edu/voter_turnout.htm Mossberger, K., Tolbert, C. J., & McNeal, R. S. (2008). Digital citizenship: The Internet, society, and participation. Cambridge, MA: MIT Press.

137

Prior, M. (2007). Post-broadcast democracy. New York, NY: Cambridge University Press. Shah, D. V., Cho, J., Eveland, W. P. Jr., & Kwak, N. (2005). Information and expression in a digital age: Modeling Internet effects on civic participation. Communication Research, 32(5), 531–565. doi:10.1177/0093650205279209 Shah, D., Cho, J., Nah, S., Gotlieb, M., Hwang, H., Lee, N.-J., . . . McLeod, D. (2007). Campaign ads, online messaging, and participation: Extending the communication mediation model. Journal of Communication, 57(4), 676–703. doi:10.1111/j.1460-24662007.00363.x Tolbert, C., & McNeal, R. (2003). Unraveling the effects of the internet on political participation? Political Research Quarterly, 56(2), 175–185. doi:10.1177/106591290305600206 Vaccari, C. (2010). Technology is a commodity: The Internet in the 2008 United States presidential election. Journal of Information Technology & Politics, 7(4), 318–339. doi:10.1080/19331681003656664 Xenos, M., & Moy, P. (2007). Direct and differential effects of the internet on political and civic engagement. Journal of Communication, 57(4), 704–718. doi:10.1111/j.1460-2466.2007.00364.x