Pundits, Ideologues, and Ranters: The British Columbia Election Online

Harold J. Jansen (University of Lethbridge)

Royce Koop (University of British Columbia)

Abstract: This article examines the potential for Internet discussion boards to be a vehicle for political deliberation through a case study of the BC Votes discussion board during British Columbia's 2001 provincial election. Research reveals that the board was dominated by a relatively small number of users and that the favourite discussion topics were not issues but how parties and leaders were performing. The authors conclude that the perception of ideological homogeneity in online discussion may be overstated and that the first post in a discussion thread has an important agenda-setting function. They also find that the relative newness of discussion boards may play a role in shaping the nature of discussion there.

Résumé : Afin d'examiner le potentiel des babillards électroniques comme véhicules de débats politiques, cet article offre une étude de cas sur le babillard BC Votes lors des élections provinciales en Colombie-Britannique en 2001. Les auteurs démontrent que relativement peu d'usagers dominaient le babillard et que les sujets préférés de discussion ne touchaient pas aux enjeux de la campagne électorale, mais plutôt à la performance des parties politiques et de leurs leaders. Les auteurs concluent que l'homogénéité idéologique des babillards électroniques n'est peut-être qu'une perception exagérée et que souvent le ton polémique du débat est dicté par le premier message dans un fil de discussion. Les auteurs concluent aussi que le caractère relativement nouveau des babillards reflète sur la spécificité de la discussion qui s'y déroule.

The development of the Internet as a medium of mass communication has engendered much speculation about its implications for democracy. The ability of the Internet to allow for citizen interaction that is inexpensive, instantaneous, and not bound by geography opens up possibilities for citizen deliberation. The most optimistic theorists see the Internet as potentially ushering in an era where direct democracy is technologically viable, even in large states. More modest aspirants point to the possibility of "electronically enhanced democracy," where additional public spaces for democratic deliberation are created.

Although there is much speculation about the potential impact of new communications technology, there are relatively few studies of actual uses of the Internet for political communication. This is particularly true of the element of Internet political communication that is most relevant to the question of enhanced democracy: online political discussion. This paper will analyze one such "public space" through an analysis of a World Wide Web-based discussion board devoted to the 2001 British Columbia provincial election. The first part of this study will briefly examine some of the claims made about the transformative potential of the Internet for the political process. The second part will be a detailed study of the characteristics and dynamics of the BC Votes 2001 Web-based discussion board. The third part of this study will compare these patterns with those in another online forum discussing the 2000 Canadian federal election.

Internet discussion: Democracy's great future?

The transformation of the Internet from a tool for academic, government, research, and computer professionals to a mass communications medium has led to much speculation about the implications of this communications revolution for political communication. Previous developments in communications technology such as the development of radio and television led to changes in the nature of political communication (Carty, Cross, &Young, 2000, chap. 7). The development of the Internet promises changes that are as profound as those previous technological developments. Indeed, it could be argued that the Internet promises changes that are even more profound. While radio and television essentially provided alternative forms for the flow of one-way information, the Internet opens up the possibility of two-way information flows. One of the defining features of the Internet is that it is interactive. Users are more than mere consumers of information. They have the potential to interact with one another and with content providers in meaningful ways (Van Dijk, 2000).

The literature on the potential impact of the Internet on the democratic process is too vast to summarize here (see Hacker & van Dijk, 2000; Wilhelm, 2000, chap. 1, for useful summaries). However, we can identify a few central claims. First, it is argued that the Internet will increase the information available to citizens. With governments and various other political actors placing political information online, more information is available to a wider range of people than at any previous point. Second, it is argued that the Internet potentially opens up the decision-making process to a wider range of people. Third, the Internet increases possibilities for citizens to engage in deliberation (Tsagarousianou, 1999).

It is the third possibility - the deliberative potential of the Internet - with which we are primarily concerned in this article. The Internet provides a number of public forums for people to engage in political discussion and to deliberate about public issues. The oldest of these is Usenet, a collection of discussion forums that are shared between computer networks. The development of the World Wide Web generated different technological forms of online discussion. However, although online discussion increasingly occurs through a Web interface, most Web-based discussion groups mimic the structure of Usenet. The basic unit of discussion is the message or post, and these are organized into topics or threads. Users can post messages, either to initiate threads or in response to previous messages.

Although there is a sizeable and growing body of literature speculating about the political implications of the Internet and online deliberation, there are few empirical studies of the actual nature of Internet-mediated political communication. Certainly, Wilhelm's assessment of the state of the field is appropriate: "Sampling several of the more prominent works on the democratic potential of the digitally mediated public sphere reveals a field that is primarily driven by normative visions, shouldered by reified concepts and categories that ought to be revised if they wither under the light of a critical investigation of their relative merits" (Wilhelm, 2000, p. 15). That said, there have been a few important studies of online political discussion: Schneider's (1996) pioneering study of abortion discussions, Wilhelm's (2000) use of content analysis, Davis' (1999, chap. 6) more widespread analysis of a number of Usenet forums, and Hill and Hughes' (1997) analysis of the development of community on Usenet. We will discuss their findings in the analysis that follows. All of these studies, however, are based on Usenet and not on Web-based discussion boards. Presumably, the greater technical knowledge needed to access Usenet means that a more narrow class of user is attracted to these forums. Furthermore, much of this analysis focuses on American politics, not on Canadian discussion boards. Discussions of Canadian politics dominated by Canadians may presumably function differently than the Usenet forums of previous studies.

British Columbia as a test case

British Columbia is an ideal place to examine how citizens are using the Internet for political purposes. B.C. is home to a wealthy, well-educated population, so it is not surprising to find high levels of Internet use there (Industry Canada, 2000; Statistics Canada, 2001). Because of the large user base, the political effects of the Internet are likely to be seen most clearly in provinces like British Columbia.

Our study focused on discussion during the 2001 B.C. provincial election campaign. The campaign was generally lacklustre and its result was predictable. Although the governing New Democratic Party (NDP) had experienced a slight increase in popularity following the selection of Ujjal Dosanjh as leader in 2000, the party could not escape the public cynicism created by Glen Clark's problematic tenure as premier. The Liberals under Gordon Campbell were consistently much more popular than the NDP, leading columnists to describe the period leading up to the election campaign as an NDP "deathwatch" (Smyth, 2001). Doubts about the NDP's ability to win even a single seat contributed to increasing media attention paid to the Green Party. Adriane Carr, the leader of the Greens, who had wrested the leadership from an older, more doctrinaire party stalwart the year prior to the campaign, projected a moderate centre-left image (Smyth, 1999). The fourth party was the new omnibus Unity Party, combined from several small right-wing parties, led by Chris Delaney, a low-profile but tough-talking conservative from New Westminster. The campaign was generally dreary; when it became clear that the NDP was headed for disaster, Ujjal Dosanjh conceded defeat in the middle of the campaign and ran negative advertisements warning British Columbians not to return "79 Gordon Campbells" to the Legislature. On election night, the Liberals took 77 of the 79 seats available with 57.5% of the popular vote, while the NDP took 2 in Vancouver's east end with 21.6% of the popular vote (Elections British Columbia, 2001). Even the premier lost his Vancouver-Kensington seat.

As is becoming more common, several media outlets cooperated in developing a common Internet portal to share news and content for the provincial election. BC Votes 2001 went live shortly before the actual election call at http://www.bcvotes2001.com. The Vancouver Sun, Vancouver Province, Victoria Times-Colonist, and Global Television provided content for the site. Besides the daily news and commentary, one of the features of BC Votes was an online, Web-based forum. This forum followed the traditional Internet practice of threaded discussions. Anyone with Web access could read the messages, but posting messages required registration, which was free and relatively easy. We do not know for certain how forum users heard of the forum; presumably, most found it as a link off of the main BC Votes site when looking for election news.

Methodology

Every day of the campaign, we went to the BC Votes discussion board and coded every message, noting its position in the thread, who posted it, as well as a number of other aspects of the message that will be discussed later. During the campaign, 164 unique user IDs (excluding the moderator) posted messages. One problem with relying on User IDs is that it is possible for a person to create more than one identity. While we have no way to ascertain the frequency with which this happens, we have no reason to believe it is a widespread phenomenon.

The first post on the board was created on March 29, 2001; the last message was posted on May 14, 2001, two days before the election. Over the course of the election, 650 messages were posted on the BC Votes board as part of 182 separate threads. We excluded those posts made by the board's moderator over the course of the campaign. This dropped the number of messages to 650.

On any given day, there was an average of 12.6 users active on the BC Votes board. They posted an average of 24 messages in an average of 16.9 active threads each day. Each day, the users started an average of 6.3 new threads. Each day generated just under 4,000 words of content. For this analysis, we excluded one message posted well before the election on March 29. The other messages were all posted during the actual campaign. There was significant daily variation in the posting behaviour on the BC Votes board, as Figure 1 indicates. Typically, participation in the boards would slow down on weekends, with the busiest periods of discussion happening during the weekdays. The second week of the campaign (from April 25 to May 2) generated the most discussion, in terms of messages, users, and threads.

Equality of participation

One of the central concerns in the analysis of message board behaviour is the equality of participation. In media-intense modern politics, relatively few people actually express opinions and ideas; most people are consumers of information and ideas. C. Wright Mills' classic formulation of the characteristics of mass society identified four problems, two of which are particularly relevant here: "In a mass, (1) far fewer people express opinions than receive them; for the community of publics becomes an abstract collection of individuals who receive impressions from the mass media. (2) The communications that prevail are so organized that it is difficult or impossible for the individual to answer back immediately or with any effect" (Mills, 1965, p. 304). The Internet in general, and message boards in particular, could potentially change this. Where TV or newspapers do not allow for viewers or readers to provide immediate feedback, Internet message boards potentially allow those with an interest in an action to participate with little delay.

Consequently, an important theme in the analysis of message boards is the extent to which a few participants dominate their discussions. The research thus far has been fairly unanimous in its findings. Schneider (1996) concluded that "posting behaviour is extraordinarily concentrated: very few authors account for a very high proportion of posts" (p. 381). Subsequent analyses have confirmed Schneider's findings. Davis (1999), for example, found a similar dominance of "regulars" in his analysis of several Usenet groups that deal with politics.

To gauge the extent to which BC Votes suffered from similar problems, we kept track of each user's posts. In our analysis, we treat each log-in identifier as a separate user. Besides the moderator, we identified 164 unique users. For each of these people, we kept track of how many posts (or messages) they created and how many words they contributed to the message board.

Our analysis suggests that the BC Votes forum suffered problems similar to those discovered in other studies. Table 1 categorizes users according to the frequency of their posting. We refer to these categories as "user class." One-time posters made up over half of the board's users, but they accounted for just over 13% of the total messages posted. The four users who dominated the board combined for almost a third (31.9%) of the total messages posted. Medium posters, however, accounted for the highest percentage of total posts (31.8%), an almost equal number to both the heavy and very heavy groups combined.

Table 1: Posting behaviour by user class
Classification Number of users Number of posts Number of words Median words/user Median average post length
One-time (1) 85 (51.8%) 85 (13.1%) 25,101 (23.3%) 94 94
Light (2-5) 55 (33.5%) 150 (23.1%) 16,008 (14.9%) 236 80
Medium (6-20) 20 (12.2%) 207 (31.8%) 31,830 (29.6%) 1,676.5 138
Heavy (21-79) 3 (1.8%) 127 (19.5%) 25,677 (23.9%) 9,547 194.8
Very heavy (80+) 1 (0.6%) 81 (12.5%) 8,917 (8.3%) 8,917 110
Total 164 (100%) 650* (100%) 107,533 (100%) 162.5 94.2

* Total after elimination of moderators' posts.

We also recorded the number of words in each post in order to analyze differences in levels of user commitment on the board. It takes relatively little effort to post several one-sentence responses - the number of words contributed by users is an indication of the extent to which they are substantially contributing to discussions. The last two columns of Table 1 summarize the median number of words for users within each user class and the median value of the users' average post length. On balance, the typical heavy user contributed the most words to BC Votes (9,547), closely followed by the single member of the very heavy user class (8,917). The other users generally trailed significantly after those two user classes.

As discussed earlier, online forums have generally followed the Usenet practice of organizing discussion along the lines of threads, or discussion topics. The first post in a particular thread assumes a special significance, because it performs an agenda-setting role. Besides setting the topic for any subsequent discussion, the first thread can set the tone for the debate. Table 2 indicates that although heavy and very heavy users were a tiny fraction of the users (2.4%), they initiated over half of all the threads on the board. Interestingly, the one-time users were most likely to initiate a thread; just over half of their posts were of the "post and run" variety. They initiated a thread and then never posted again.

Table 2: Thread initiation
Classification (% of users) Total posts Thread-initiating posts Thread-initiating posts
as % of total posts
One-time (51.8) 85 (13.1%) 44 (25.9%) 51.8
Light (33.5) 150 (23.1%) 31 (18.2%) 20.7
Medium (12.2) 207 (31.8%) 37 (21.8%) 17.9
Heavy (1.8) 127 (19.5%) 25 (14.7%) 19.7
Very Heavy (0.6) 81 (12.5%) 33 (19.4%) 40.7
Total 650* (100%) 170 (100.0%) 26.2

* Total after elimination of moderators' posts.

When we break down participation on the basis of the number of posts created by each participant, we find significant inequality in participation. The users who were in the bottom half in terms of the number of messages they posted contributed only 10% of the total posts over the election. At the other end of the spectrum, the top 10% of users contributed over half of the total number of posts on BC Votes. The pattern becomes even more pronounced when participation is measured by the number of words posted. The half of the users who posted the fewest words posted only 6% of the total words submitted to BC Votes, while the top 7% of users wrote over 50% of the total words written on the message board. If we calculate the Gini coefficient of inequality based on the participation patterns on BC Votes, we find that the measure based on the number of posts is 0.623, while the Gini coefficient based on the number of words is 0.747.1 Figure 2 illustrates this inequality by plotting the cumulative percentage of posts and words against the cumulative percentage of users. The further the curve deviates from the line plotting perfect equality (where, for example, 40% of the users contribute exactly 40% of the posts), the more we can see inequality in message participation. From this, we can note that only 10% of the people participating on BC Votes contributed over half of the posts and over 60% of the words posted.

In summary, BC Votes 2001 suffered from the same problems of participation that plague other Internet-based discussion boards. If equality is an important element of democratic deliberation, the BC Votes discussion board was sadly deficient in this respect. Even when presented with technology that lends itself to the participation of a wide number of users and opportunities for people to express ideas and not just receive them, a small group of users dominated the discussion.

Issues, parties, and leaders: What did users talk about?

The topics of discussion for online users are an important consideration because one of the supposed advantages of online discussion is the ability for people to discuss what matters to them, unfettered by media gatekeepers. Common criticisms of the mass media are that they tend to focus on personality and leaders rather than substantive issues in the election and that they tend to ignore candidates and parties that the public would like to hear about.

Given the opportunity to set the agenda themselves, what did the users talk about? Table 3 summarizes the most popular topics of discussion on BC Votes. Surprisingly, the most popular topics of discussion were not issues, but rather the performance and fortunes of the parties and leaders (37.3%). BC Votes 2001 appears to have provided an opportunity for users to become political analysts and have an audience for their analysis. Given the prevalence in the mainstream media of pundits who make a living pronouncing on the political process, the message board seems to have encouraged "amateur punditry."

Table 3: Top 10 discussion topics on BC Votes, 2001
Topic Number of posts % of total posts*
Party performance/fortunes 183 28.2
Health care 102 15.7
Personal comments directed at other users 97 14.9
Taxes 73 11.2
Economy 68 10.5
Leader performance/fortunes 59 9.1
Institutional reform 57 8.8
Scandals 54 8.3
Media coverage of election 53 8.2
Education 49 7.5

* Numbers in column will not add to 100% due to multiple mentions possible in each post.

One consistent concern over the Internet and political participation is the fact that Internet users are not representative of the general population (Davis, 1999). As mentioned earlier, Internet users are generally better educated and wealthier than the general population, factors that consistently correlate with political interest and participation. In particular, participants on a political message board are more likely to be politically interested, as reading and writing messages requires an investment of time and interest. BC Votes users were thus likely members of the "attentive public." This does not necessarily mean, however, that their concerns were substantially different from those of the general public. BC Votes users talked about the same issues that concerned the general public (see Table 4). According to the polls, the dominant two issues were economics and health care. Our coding separated economics and tax cuts, but those issues were clearly the dominant issues for BC Votes users, too. The one issue that was more important to BC Votes users than to the general public was institutional reform. Among other things, this reflected an interest in electoral reform and the challenge of incorporating Aboriginal representation in a proportional representation system.

Table 4: Dominant election issues, Compas polls
Issue % of respondents
April 2001 May 2001
Improve or protect health care 18 29
Economy; economic growth; tax cuts 27 24
Honesty or integrity in government; Corrupt, unethical, dishonest NDP 11 8
Improve or protect education 5 4
Miscellaneous anti-NDP feeling 5 4
Improve or protect the environment 3 4
Competent management in government 2 4
Source: Adapted from URL: http://www.compas.ca/html/archives/growthofthegreens_surv.html.

Given the emphasis in BC Votes on analyzing events in the election - what we have called "amateur punditry" - we would expect parties and leaders to be a favourite topic of conversation. Table 5 reports which parties and leaders were most frequently mentioned in the users' election discussions. Political parties were mentioned by name in almost 60% of the total messages. Surprisingly, leaders were not mentioned by name very often - only 17.1% of the messages actually mentioned a particular leader. Given the importance of leadership in structuring the vote, this has to rank as one of the most surprising findings of our analysis (McAllister, 2000).

Another striking finding is the dominance of the two major parties. Despite the leveling potential of Internet discussion, the two major parties clearly were favourite topics for conversation. Just over 50% of all the messages mentioned the Liberals and/or the NDP; 15.7% of all the messages mentioned Ujjal Dosanjh and/or Gordon Campbell. Nearly twice as many messages named the NDP than named the Liberals. For online discussants, this suggests that the election was more about scrutiny of past NDP performance than prospective Liberal performance.

Table 5: Mentions of parties and leaders
Parties Leaders
Party # of posts % of total Leader # of posts % of total
NDP 267 41.1 Dosanjh 55 8.5
Liberals 177 27.2 Campbell 69 10.6
Green 51 7.8 Carr 8 1.2
Unity 25 3.8 Delaney 14 2.1
Other 50 7.7 Other 5 0.8
Any party(s) 371 57.1 Any leader(s) 111 17.1

This conclusion is given further weight through the analysis of positive and negative mentions of parties and leaders. We coded a positive evaluation of a party or leader in a message as a +1 and a negative evaluation of a party or leaders as a -1. A neutral reference was coded as a 0. Table 6 reports the scores for each party, and Table 7 reports the same information for leaders. Over three-quarters of the messages that mentioned the NDP did so in a negative light. Combined with the previous findings, we can conclude that the participants saw the election not only as a judgment on the NDP, but an overwhelmingly negative judgment at that. This negative assessment of the NDP, however, was not counterbalanced by a positive assessment of the Liberals.

Table 6: Party evaluations
Party Positive (+1) Neutral (0) Negative (-1) Mean
NDP 29 (10.9%) 35 (13.1%) 203 (76.0%) -0.65
Liberals 53 (29.9%) 50 (28.2%) 74 (41.8%) -0.12
Green 20 (39.2%) 16 (31.4%) 15 (29.4%) +0.10
Unity 5 (20.0%) 7 (28.0%) 13 (52.0%) -0.32
Other 24 (48.0%) 9 (18.0%) 17 (34.0%) +0.14
All parties 131 (23.0%) 117 (20.5%) 322 (56.5%) -0.34

Table 7: Leader evaluations
Leader Positive (+1) Neutral (0) Negative (-1) Mean
Dosanjh 9 (16.4%) 13 (23.6%) 33 (60.0%) -0.44
Campbell 17 (24.6%) 23 (33.3%) 29 (42.0%) -0.17
Carr 2 (25.0%) 0 6 (75.0%) -0.50
Delaney 4 (28.6%) 5 (35.7%) 5 (35.7%) -0.07
Other 1 (20.0%) 3 (60.0%) 1 (20.0%) 0
All leaders 33 (21.9%) 44 (29.1%) 74 (49.0%) -0.27

Based on this analysis, the lasting impression of the BC Votes discussion board is its overwhelming negativity. The only party with a favourable rating overall was the Green party (as were the other parties, taken as a group). While this undoubtedly reflects something of the nature of the 2001 provincial election and the state of British Columbia party politics, these findings are consistent with analyses of other online discussions. Indeed, federal Liberal pollster Michael Marzolini dismisses online discussion as the haven for the "angry" and "frustrated" who need to "let off steam" (Marzolini, 1999).

Ideology and patterns of agreement and disagreement

In assessing the suitability of Internet message boards for political deliberation, the central debate in the literature has been over whether Internet discussions provide the give and take of political deliberation. The wide range of public spaces means that users can search out forums where they feel comfortable and where underlying assumptions are not challenged. Sunstein forcefully develops this critique of the Internet. Arguing that the idea of a public forum requires access to and subsidy of "shared exposure to diverse speakers with diverse views and complaints," he argues that the Internet fragments public discussion to the extent that citizens no longer come into contact with unwanted arguments (Sunstein, 2001, p. 31). While Sunstein's book is primarily theoretical, there is some supporting empirical evidence in previous analyses of online discussion. Hill and Hughes (1997) demonstrate how leadership maintenance and recruitment efforts in Usenet sustain a level of ideological purity in discussion threads. Davis (1999) and Wilhelm (2000) echo those conclusions, finding that users tended to reinforce opinions, rather than challenge them.

The literature thus far seems to be fairly conclusive, but the evidence rests on analyses of Usenet, rather than Web-based boards. Given that Web-based boards are presumably open to a wide range of users, it is possible that these discussions are different in character than those found on Usenet.

Our analysis involved coding messages as to whether they expressed what could be considered a right or left ideological position, as these are conventionally understood.2 Just over 70% of the 650 messages posted to BC Votes could not be categorized as expressing a particular ideology. Twenty percent of the messages were coded as to the right; 9.5% were coded as being to the left. The overall rightward tendency of BC Votes is certainly consistent with the findings of previous analyses of online discussion. Hill & Hughes (1997, p. 12) report that "on the USENET, conservatives outnumber liberals." Davis (1999, p. 156) concludes, "Usenet posters likely are more right-wing ideologically than the general public." Unlike other boards, however, BC Votes 2001 was somewhat unusual in that the vast majority of posts were non-ideological. Hill & Hughes' analysis of Usenet found that 44.2% of the messages were non-ideological, significantly fewer than was the case on BC Votes.

Given the dominance of the board by relatively few posters, to what extent was the ideological character of the board determined by the heavy users rather than infrequent posters? Table 8 reports the ideological character of messages by user class. Although the relationship between user class and ideology is not particularly strong, there are a few notable patterns. One-time posters were less likely to post an ideologically oriented post than were more frequent posters. The most interesting pattern is the behaviour of the one very heavy poster. Nearly a third of this user's posts were ideologically right-wing. In fact, this one user accounted for 42% of all of the right-wing messages on BC Votes. If we exclude this user's messages from the analysis, BC Votes appears to be more balanced ideologically: 13.0% of the posts are left-wing, while 15.7% of the posts are right-wing; the remainder are non-ideological. To a significant extent, then, one user determined the overall ideological "feel" of this discussion board.

Table 8: Ideology of posts by user class
User class Number
of users
Left None Right Total
One-time 85 7 (8.2%) 67 (78.8%) 11 (12.9%) 85 (100%)
Light 55 11 (11.2%) 69 (70.4%) 18 (18.4%) 98 (100%)
Medium 20 21 (15.9%) 90 (68.2%) 21 (15.9%) 132 (100%)
Heavy 3 23 (14.1%) 115 (70.6%) 25 (15.3%) 163 (100%)
Very heavy 1 0 (0%) 117 (68.0%) 55 (32.0%) 172 (100%)
Total 164 62 (9.5%) 458 (70.5%) 130 (20.0%) 650 (100%)

Chi-square: 44.6 (p < 0.001) Cramer's V = 0.19

Given that BC Votes did not have a particularly strong ideological orientation, we thus might expect that its patterns of agreement and disagreement would be different from those observed on Usenet groups. Our analysis suggests that BC Votes was substantially different from those other forums in that there was a significant level of debate. Nearly half of the messages disagreed with the message posted before it, while about a quarter of the messages expressed agreement. Clearly, BC Votes was not a message board that was the exclusive domain of the like-minded.

Our analysis of the role of ideology and patterns of disagreement highlights the importance of the first message in a thread. On BC Votes, first-post ideology was associated with the amount of discussion generated. Left-wing initial messages started messages that lasted, on average, 6.53 messages, nearly twice as long as right-ring or centrist first posts. Clearly, posts that went against the "ideological grain" of BC Votes generated higher levels of discussion.

Table 9 supports this conclusion. Non-ideological messages tended to generate non-ideological follow-ups. Left-wing follow-ups were equally likely in response to any ideological initial post. Right-wing follow-ups were more likely to be found in response to any ideological initial post, but especially if the initial post was a left-wing message. In other words, a user posting a left-wing message could anticipate a flurry of right-wing responses - and only a little bit of support from other left-wing messages. Right-wing messages were more likely to follow up right-wing initial posts.

Table 9: Ideological follow-ups in threads
Ideology of first post Mean length
of thread
(posts)
Mean proportion of ideological follow-ups (%)
Left Right
Left 6.53 16.2 43.3
None 3.29 4.7 10.6
Right 3.34 16.8 28.1

N=118

The BC Votes message board thus provides a brighter picture about the deliberative potential of online political discussion than previous analyses. BC Votes was less ideological and more driven by disagreement and debate. Still, we do not want to overstate the degree to which BC Votes fulfilled the ideal of deliberative democracy, where strong cases for opposing points of views are presented and debated. To a significant extent, much of the disagreement on BC Votes was the product of left-wing users facing a lengthy response from right-wing users. There was a conversation, but it was primarily in one direction. This tendency may have resulted from the atmosphere of the election campaign, where the left-wing NDP was very unpopular.

Reasonable debate

Of course, simple disagreement does not make deliberation. A certain level of thoughtfulness and reasoned debate is required to keep political debate from degenerating into a shouting match. Critics of this feature of online discussion abound. Charles McGrath advances a typical critique: "The Net is full of ranters standing on invisible soapboxes, and a great many exchanges essentially come down to: Enough about you. Let's hear from me" (McGrath, 1996, p. 84). Davis' analysis of Usenet (1999) found that users tended to rely on their own authority and rarely provided external evidence and corroboration for their views.

A surface reading of the BC Votes forum was that it was full of angry ranting about everything from leaky condominiums to fast ferries. To more systematically judge the extent to which BC Votes could be seen as living up to the ideals of reasonable debate, we coded each message on a scale from 1 to 4, with 1 being the lowest level of debate and 4 indicating a high level of debate. We tried to mitigate the inherent subjectivity of this approach by providing a standardized set of guidelines for coding messages that would aid in the categorization. While other readers of BC Votes may dispute these guidelines, we believe they constitute a useful means of categorizing posts on widely held conceptions of what constitutes reasonable debate. The categories and the frequencies appear in Table 10.

Table 10: Reasonableness of messages
Score Description N
0 Not applicable 2 (0.3%)
1 No thought, insults 10 (1.5%)
2 Opinion: Very little thought, but no blatant insults 375 (57.7%)
3 Opinion backed up with some thought and reasons but no evidence to back up opinion 251 (38.6%)
4 Involves substantial thought and evidence to support position 12 (1.8%)

Our analysis thus far has demonstrated the importance of the initial post in a thread in setting a tone for the subsequent discussion. Table 11 explores the question of whether the reasonableness of the first message in a thread affects the character of the subsequent debate. Based on a very small number of cases, on BC Votes, the most reasonable posts (a rating of 4) generated a much longer discussion than messages that were of lower reasonableness. Furthermore, at the low end, threads that began with an insult-oriented thread (a rating of 1) quickly died out.

Table 11: Consequences of first post reasonableness
First post reasonableness N Mean length of thread (std dev.) Mean reasonableness of follow-ups (std dev.)
1 4 2.50 (0.58) 2.38 (0.75)
2 81 3.91 (3.97) 2.29 (0.33)
3 78 3.22 (2.68) 2.41 (0.42)
4 7 6.43 (6.05) 2.55 (0.27)
Total 170 3.66 (3.5) 2.36 (0.39)

A longer thread does not necessarily mean a higher quality discussion; the final column of Table 11 shows a very modest relationship between first-post reasonableness and the average reasonableness of the follow-ups. The magnitude of the differences is small, but the direction of the relationship between the reasonableness of first posts and of subsequent posts is positive. Although this is suggestive, the modest nature of the relationship is such that we are not able to conclude that a more reasonable first post will lead to a more reasonable and thoughtful subsequent discussion.

Was BC Votes typical of online discussions of elections?

Because the study of online discussion is in its infancy, there is very little research that can serve as a benchmark for comparison. It is difficult to determine on the basis of the data presented thus far whether BC Votes is representative of online discussion or reflects unique features of its technology, board culture, and context. In order to provide some basis for assessing the significance of our findings, this section will compare our BC Votes finding with a similar analysis of the Bourque discussion board during the 2000 Canadian federal election.

Like BC Votes, the Bourque board is a Web-based board. There were some significant differences, however. First, the BC Votes discussion forum was linked to a media site that provided original news coverage; the Bourque board is a feature of a Web page (http://www.bourque.com) that provides links to other stories, but no original content. Secondly, unlike the BC Votes Board, the Bourque board did not require users to register their names before posting. Third, the BC Votes board had a very limited focus: the BC provincial election. The Bourque board is a general discussion of Canadian politics; during the federal election it obviously focused on electoral matters, but that was not its exclusive focus. Fourth, and arguably most importantly, the BC Votes board was a relatively new creation at the time of coding and existed only temporarily during the B.C. election. In contrast, the Bourque board has existed for several years as a fixture on Pierre Bourque's news site and is characterized by discussion among long-term users.

Our methodology in analyzing the Bourque board was similar to that used to analyze BC Votes, except that the sheer volume of traffic on the board required us to sample messages and threads, rather than code every message. Each day from October 23 until December 2, 2000, a random sample of six new threads was selected and all the posts within these threads (no matter how many) were coded. In addition, all the original posts in new threads for the period were coded.

Table 12: User class on the Bourque and BC Votes boards
Classification Number
of
users
Users (%) Total posts Posts (%)
Bourque One-time (1) 77 43.02 77 3.83
Light (2-5) 56 31.28 159 7.91
Medium (6-20) 27 15.08 274 13.63
Heavy (21-79) 13 7.26 547 27.21
Very heavy (=>80) 6 3.35 953 47.41
Total 179 2010
BC One-time (1) 85 51.83 85 11.89
Light (2-5) 55 33.54 150 20.98
Medium (6-20) 20 12.20 207 28.95
Heavy (21-79) 3 1.83 127 17.76
Very heavy (=>80) 1 0.61 81 11.33
Total 164 650

One of our central findings on BC Votes was the dominance of a small number of users. To what extent is this typical of online discussion? The Bourque board also fell well short of the ideal of equality of participation, as seen in Table 12. There are some interesting differences, however. Fewer one-time posters and a larger number of heavy and very heavy posters used the Bourque board. Furthermore, in terms of total messages, the heavier user classes dominated the Bourque board much more completely than on BC Votes. To a greater extent, the Bourque board was more of a club, dominated by a small cadre of users. The explanation for this difference likely lies in the more established nature of Bourque. While the community on BC Votes was only temporary in nature, the lasting nature of Bourque presumably rewards the investment of time and posting effort.

Another central element of our analysis was the emphasis on two parties and leaders in BC Votes. Despite the opportunity for Internet discussion to broaden the subjects of discussion, BC Votes users did not do so. Table 13 presents the same information from the Bourque board: 43.5% of total posts to the Bourque board mentioned parties; and while leaders were mentioned in only 17.1% of posts to the BC Votes board, 29% of total posts to the Bourque board mentioned leaders. The Bourque board, in this case, better fulfills our expectations about the importance of leadership.

Table 13: Mentions of parties and leader, Bourque board
Parties Leaders
Party # of posts % of total Leader # of posts % of total
Liberal 492 24.5 Chrétien 297 14.8
Alliance 531 26.4 Day 277 13.8
NDP 84 4.2 McDonough 51 2.5
BQ 45 2.2 Duceppe 26 1.3
PC 224 11.1 Clark 153 7.6
Other 22 1.1 Other 7 0.4
Any party(s) 874 43.5 Any leader(s) 582 29.0

The focus of the BC Votes board on the two main parties and leaders was similar to that of the Bourque board. Unlike the BC Votes board, however, there was no great difference in the amount of discussion of the Liberals and the Alliance or Chrétien and Day. While the BC Votes board was characterized by criticisms of the unpopular NDP, the Bourque board was characterized by equal attention paid to both main parties.

Table 14: Party evaluations, Bourque board
Party Positive (+1) Neutral (0) Negative (-1) Mean
Liberal 23 (4.7%) 193 (39.2%) 276 (56.1%) -0.51
Alliance 80 (15.1%) 209 (39.4%) 242 (45.6%) -0.31
NDP 10 (11.9%) 49 (58.3%) 25 (29.8%) -0.18
BQ 4 (8.9%) 37 (82.2%) 4 (8.9%) 0.00
PC 38 (17.0%) 127 (56.7%) 59 (26.3%) -0.09
Other 1 (4.6%) 18 (81.9%) 3 (13.6%) -0.09
All parties 156 (11.2%) 633 (45.3%) 609 (43.6%) -0.32

The overwhelming impression of the BC Votes board was its negativity. It is plausible that this could simply reflect the character of the British Columbia election, where voters were not happy with the NDP but were not particularly enamoured with the Liberals and Gordon Campbell either. Our analysis of the Bourque board suggests that this negativity is, in fact, typical of online discussions. Table 14 and Table 15 present an analysis of party and leader mentions similar to that performed on BC Votes. The Bourque board was roughly equal to the BC Votes board in its treatment of parties. Differences exist, however, in the evaluations of different parties. Although no federal party was rated as low as the B.C. NDP (-0.65), the Liberals were not far behind this score. The Bourque board was much more critical of leaders than the BC Votes board, producing a mean score of -0.42 compared to -0.27 for the BC Votes board.

Table 15: Leader evaluations, Bourque board
Leader Positive (+1) Neutral (0) Negative (-1) Mean
Chrétien 8 (2.7%) 86 (29.0%) 203 (68.4%) -0.66
Day 39 (14.1%) 96 (34.7%) 142 (51.3%) -0.37
McDonough 3 (5.8%) 26 (51.0%) 22 (43.1%) -0.37
Duceppe 4 (15.4%) 21 (80.8%) 1 (3.9%) 0.12
Clark 35 (22.9%) 58 (37.9%) 60 (39.2%) -0.16
Other 0 (0.0%) 3 (42.9%) 4 (57.1%) -0.57
All leaders 89 (11.0%) 290 (35.8%) 432 (53.3%) -0.42

One of the findings that set BC Votes apart from analyses of Usenet discussion was that it was less ideological. Surprisingly, the Bourque board was even less ideological than BC Votes. Only 17.5% of the messages were ideological in nature, compared to 29.9% in the case of BC Votes. However, even though the number of ideological posts was lower on Bourque, the ratio of right-wing to left-wing posts was similar to BC Votes (roughly 2:1). Given that this ratio was similar to that found on Usenet, it appears that this right-wing bias is a pronounced feature of online discussion, and not something unique to BC Votes (Hill & Hughes, 1999).

Our analysis of BC Votes also found that there was less unanimity than was found in previous analyses of Usenet. Bourque board users were slightly more likely to agree or to neither agree nor disagree, but BC Votes board produced significantly more disagree posts (48.9% compared to 41.5%). This is consistent with our argument that the difference reflects two boards at different stages of development.

Conclusion

Our analysis of the BC Votes board focused on a number of issues that relate to the viability of the Internet for the discussion of politics. One of our most significant findings is not revolutionary, but confirms the findings of previous scholars. A small group of users dominate Internet discussion. On BC Votes, four users (out of 164) posted almost half of the messages. Although these users do not completely dominate the choice of topics discussed, they significantly influence the tone of the board. On BC Votes, one user was almost single-handedly responsible for the ideological perspective of the board. If encouraging many people to share opinions and ideas is an important goal, our analysis suggests that Internet message boards do not live up to their potential.

The users of BC Votes tended to take on the role of amateur pundits in their discussions of the election. They spent much of the time talking about parties, leaders, and media coverage of the election, without reference to issues. When they did talk about issues, users tended to be concerned about the same issues as the general public, as revealed by Compas polls during the election. This may reflect the agenda-setting role of the media, but it reveals that online users were not substantially different from the general public, at least in this respect. Their discussions of parties and leaders were overwhelmingly negative, however. Simply put, BC Votes users did not have much nice to say about anyone. Again, this may simply reflect the mood of the general population in the 2001 provincial election, but it also reveals the role of discontent as a motivation for participating in discussion forums. Our study thus confirms previous depictions of online discussion as a haven for complainers.

Although our conclusions tend to confirm many of the negative assessments of online political discussion, our analysis suggests that concerns over the fragmenting nature of the Internet and the tendency of Internet users to be drawn into little enclaves of self-referential agreement may be overstated. We found that BC Votes did have a substantial level of disagreement. The right-leaning ideological tone of BC Votes meant that posts that were ideologically left-leaning sparked a reasonable level of debate.

Our analysis has been primarily exploratory and descriptive, an appropriate research standpoint given the relative infancy of the study of online discussion. It points to the possible importance of two variables. Our brief comparison of BC Votes with the Bourque board suggests that the relative newness of the BC Votes board may have played a significant role in shaping its character. Our findings suggest that a board with a longer history and more established user base may be less ideologically heterogeneous and feature lower levels of disagreement. Initially, message boards are the subject of a contest to determine their ideological orientation and character. People who do not feel comfortable with this orientation and character will either scale back their participation or leave altogether. Our findings on this point are not conclusive and point to the need for further research.

Our analysis also suggests that the first post in a thread may play an important agenda-setting function besides determining the topic of the discussion. We found that the ideological standpoint of the initial post in relation to the rest of the message board and its reasonableness helped to determine whether a discussion topic would "take off," sparking a lengthy discussion, or would die out. There was some evidence (although not strong) that the reasonableness of the initial message shaped the intelligence of the subsequent discussion.

The study of online communication is still in its infancy. Although the BC Votes 2001 board was a temporary discussion board with a limited focus, our analysis drew a number of conclusions that need to be tested in studies of other, more mature discussion boards. In particular, the role of patterns of agreement and disagreement in online discussion need to be studied further, as this is central to the question of democratic discourse. As new forms of online interaction and discussion emerge, such as weblogs, these questions need to be at the centre of future studies.

Acknowledgments

The authors gratefully acknowledge the financial support of the Centre for Cybercitizenship and the University of Lethbridge Research Fund, the research assistance of Adam Drew and Tony Christensen, and the helpful comments and suggestions of the anonymous reviewers and the editor of this journal.

Notes

  1. The Gini index is usually used to indicate inequality in national income distribution, where 0 indicates perfect equality and 1 indicates perfect inequality. Since the formula for calculating a Gini coefficient is complicated, we adopted the estimation formula provided by Gardner (1988). To give a point of comparison, Canada's Gini coefficient is 0.315. The Gini values here exceed those of all countries for whom data were available in the United Nations Human Development Report of 2001. The highest values were Swaziland at 0.609 and Nicaragua at 0.603. See URL: http://www.undp.org/hdr2001/indicator/indic_105_1_1.html.

  2. Messages that advocated greater government involvement in the economy or a more liberal position on moral issues such as abortion or the rights of gays and lesbians were coded as being to the "left," while messages that advocated less governmental intervention in the economy or a more traditional position on moral issues were coded as being to the "right."

References

Carty, R.K., Cross, William, & Young, Lisa. (2000). Rebuilding party politics. Vancouver, BC: UBC Press.

Davis, Richard. (1999). The Web of politics: The Internet's impact on the American political system. New York, NY: Oxford University Press.

Elections British Columbia (2001). Statement of votes: 37th provincial general election, May 16, 2001. URL: http://www.elections.bc.ca/elections/sov01/sov_tocfinal.html [April 5, 2002].

Gardner, H. Stephen. (1988). Comparative economic systems. Chicago, IL: Dryden.

Hacker, Kenneth L., & van Dijk, Jan (2000). What is digital democracy? In K.L. Hacker & J. van Dijk (Eds.), Digital democracy: Issues of theory and practice (pp. 1-9). Thousand Oaks, CA: Sage.

Hill, Kevin A., & Hughes, John E. (1997). Computer-mediated political communication: The Usenet and political communities. Political Communication, 14, 3-14.

Industry Canada. (2000). Internet use in Canada. URL: http://strategis.ic.gc.ca/SSG/ri00122e.html [April 5, 2002].

Marzolini, Michael. (1999, December). Party politics and the Internet. The Liberal Times. URL: http://www.pollara.ca/new/Library/SURVEYS/ltimesDec_99.htm [April 5, 2002].

McAllister, Ian. (2000). Leaders. In L. LeDuc, R.G. Niemi, & P. Norris (Eds.), Comparing democracies: Elections and voting in comparative perspective (pp. 280-298). Thousand Oaks, CA: Sage.

McGrath, C. (1996, December). The Internet's arrested development. New York Times Magazine, 84.

Mills, C. Wright. (1956). The power elite. Oxford: Oxford University Press.

Smyth, Michael. (1999, October 10). Green boss in the fight of his life. Vancouver Province, p. A6.

Smyth, Michael. (2001, March 13). MLA clings to faint hope it won't be a wipeout. Vancouver Province, p. A6.

Schneider, Steven M. (1996). Creating a democratic public sphere through political discussion: A case study of abortion conversation on the Internet. Social Science Computer Review, 14, 373-393.

Statistics Canada. (2001). Household Internet use survey. The Daily. URL: http://www.statcan.ca/Daily/English/010726/d010726a.htm [July 26, 2001].

Sunstein, Cass. (2001). Republic.com. Princeton, NJ: Princeton University Press.

Tsagarousianou, R. (1999). Electronic democracy: Rhetoric and reality. Communications: The European Journal of Communication Research, 24, 189-208.

van Dijk, Jan. (2000). Models of democracy and concepts of communications. In K.L. Hacker & J. van Dijk (Eds.), Digital democracy: Issues of theory and practice (pp. 54-69). Thousand Oaks, CA: Sage.

Wilhelm, G. (2000). Democracy in the digital age: Political life in cyberspace. New York, NY: Routledge.

  •  Announcements
    Atom logo
    RSS2 logo
    RSS1 logo
  •  Current Issue
    Atom logo
    RSS2 logo
    RSS1 logo
  •  Thesis Abstracts
    Atom logo
    RSS2 logo
    RSS1 logo

We wish to acknowledge the financial support of the Social Sciences and Humanities Research Council for their financial support through theAid to Scholarly Journals Program.

SSHRC LOGO