Tweet, Click, Vote: Twitter and the 2010 Ottawa Election

It’s 1:30am on October 26, 2010. The Ottawa municipal election is over. I will spend the next week reviewing and analyzing almost 10,000 archived Tweets about the election, but wanted to share some of the preliminary findings now.

OVERALL TWEETING ACTIVITY (01 – 25 Oct)

Total tweets: 9409 (it’s noteworthy that 70 percent of all tweets during this period were generated during the final week of the campaign)

Total tweeters: 1344

URLs shared: 1125 (top URL goes to The Bulldog Ken Gray)

80% (7527) of all tweets generated by 22% (308) of all tweeters

Top 10 tweeters (0%) generated 22% (2094) of all tweets

45% of all tweeters posted only 1 tweet

#OTTvote Tweets (Oct 1-25, 2010)

In the last 4 days of the campaign, 809 new Twitter accounts contributed to the #ottvote feed. It’s difficult to verify the authenticity of these accounts since most of them used non-identifiers. Yet, since a vast majority of them were used to besmirch the online reputations of candidates and their supporters it’s likely illustrative of the role of astroturfing in local politics. Indeed, there were almost 300 more new accounts in the final 4 days of the campaign than the total number (535) that had contributed to the Twitter feed during the previous 20 days.

TOP 10 TWEETERS (volume of activity)

CliveForMayor (356)

willsamuel (322)

ottawasun (237)

OniJoseph (231)

MacDoaker (198)

davidreevely (180)

DenVan (155)

SunCityHall (149)

jchianello (147)

josh_greenberg (119)

TOP @ REPLY RECIPIENTS OR MENTIONS

CliveDoucet (286)

JimWatsonOttawa (266)

CliveForMayor (239)

LarryOBrien2010 (172)

denvan (153)

willsamuel (152)

OniJoseph (104)

Ottawateaparty (100)

josh_greenberg (93)

ctvottawa (92)

TWEET SOURCE

The vast majority of #ottvote tweets were posted from the main Twitter website or via one of the many popular applications (e.g. Tweetdeck). Although only 15% of the tweets were added from a mobile device, I expect to see more use of iPhones, Blackberries and other smartphones in future campaigns not only for contributing content but for following results as well.

Desktop/Laptop – 85%

iPhone/Blackberry/Android – 15%

TWITTER SENTIMENT: TOP 3 MAYORAL CANDIDATES (18-25 Oct)

In the final week of the campaign a significant amount of Tweeter energy was devoted to negatively framing the incumbent and perceived frontrunner. The sharp increase in new Twitter accounts contributing to the #ottvote thread helps explain the heavy negative sentiment scores reported below, particularly for mayor-elect Watson.

Larry O’Brien (1134 opinions)

12% [+]

62% [-]

26% [+/-]

Clive Doucet (318 opinions)

46% [+]

10% [-]

44% [+/-]

Jim Watson (1177 opinions)

20% [+]

50% [-]

30% [+/-]

TWEET THEMES (Sept 1 – Oct 25)

Using the Crimson Hexagon data mining program I developed a coding grid to map the frequency and distribution of primary themes in the #ottvote feed between September 1 and October 25. As can be seen, over this period tweets discussing various aspects of the transit issue (e.g. light rail, ring road, OC Transpo strike, cycling) were predominant, followed closely by tweets which focused on a personality characteristic of a candidate. With a few notable exceptions, these personality mentions were almost entirely negative and focused in almost every case on one of the mayoral candidates.

The next two most common themes mentioned in tweets relating to the election focused on some element of the media coverage (e.g. announcing or commenting on media endorsement of a candidate) or on some type of electioneering strategy (the latter came almost entirely from tweets by candidates, e.g. “I’ll be canvassing in the Byward Market, come say hi!”). Although mentions of development were consistent and modest, I was surprised that more tweets relating to intensification, infill, environmental impacts of new infrastructure, etc. did not garner more attention. Equally surprising was the significantly lower numbers of tweets mentioning taxes or finance compared to other issues. Tweets referring to a range of social issues (housing, parks/recreation, childcare, etc.) netted only 11% of the mentions and tended to cluster around particular dates in which there were higher than normal levels of tweeting about these topics (i.e. Social Issues Mayoral Debate on Oct 8). Finally, although I coded for mentions of crime these did not yield significantly high numbers of mentions.

SUMMARY

It’s important to put case study findings in the appropriate comparative context. Looking at the final week of Twitter activity alone, it is noteworthy that #ottvote contributors generated 6642 tweets about the municipal election. However, their counterparts in other closely watched Canadian municipal campaigns were far more active. Using the hashtag #yycvote, Calgary tweeters generated 18,692 tweets about that city’s election in its final week. Calgary is a reasonable point of comparison to Ottawa – based on 2006 census data, it is Canada’s third largest city (population 988,193) while Ottawa is the country’s fourth largest (population 812,129). In Toronto, where the most exciting and controversial race occurred, tweeters contributed an impressive 33,504 tweets to the #VoteTO hashtag in the final 7 days of the campaign.

I will spend the coming week trying to make sense of these findings. What do these data say to you? Please share your thoughts about the role Twitter played in the election. What did you think were the most and least interesting and effective uses of Twitter? What impact, if any, might it have had on the election process and outcome?

ACKNOWLEDGEMENTS

The findings reported above were generated with two very useful open source data mining programs (Twapper Keeper and The Archivist). The sentiment analysis was performed using an algorithm developed by Crimson Hexagon. I want to thank Melyssa Plunkett-Gomez, VP Sales & Business Development, Crimson Hexagon, for providing access to this excellent program. I also wish to recognize the outstanding research assistance provided by Vincent Raynauld, PhD candidate in Communication at Carleton University.

Advertisements

7 Comments

Filed under Politics, Social Media, Technology

7 responses to “Tweet, Click, Vote: Twitter and the 2010 Ottawa Election

  1. My first reaction to your posting is to say “Jeez, that Will Samuel guy sure can’t shut up”, but then I realize that the sarcasm might be lost in translation.

    Thank your for tracking the metrics of the election over the last 25 days or so. It was a lot to learn and even more to work with.

    Cheers,

    Will

  2. Anonymous

    I have several friends in Calgary and visit frequently and although the population size may be the same I am willing to guess (and it shouldn’t be too hard to determine as fact) that the demographic is not the same, the population is overall younger.

    • Josh Greenberg

      The significantly higher volume of #yycvote tweets cannot be explained solely by the age of voters. There are also other mitigating demographic factors that could have turned the tide in another direction if you look at social categories deterministically (e.g race and ethnicity). The major difference is that Naheed Nenshi represented a significantly newer, fresher perspective on local politics than any of the other candidates. Although Clive Doucet was the most progressive candidate in the Ottawa race, he is a well known quantity, did not offer a platform that was not already familiar to people in this city and does not connect with voters in the same way as Nenshi. Political watchers, particularly youth, do not contribute the volumes of content we are dealing with here unless they are motivated and inspired to do so.

  3. Pingback: End of Elections « Matthew Kupfer

  4. Cool stuff and thanks for the references on the open source data mining sources!

    To me, the most interesting part is the negative framing that seems to be going on. Not surprising. I know the election junk mail I’ve been getting from snail mail over the past few weeks has been pretty much all negative.

    I wonder if there’s a reinforcing spirals effect. The idea basically is people seek out particular news that reinforces their point of view, thereby spiraling (and bolstering) people’s world view. Michael Slater advanced this idea in a seminal article in Communication Theory in 2007 (get link to cite here: http://osuhcrg.blogspot.com/2008/10/dr-michael-slater-wins-intl-award-for.html)

    The question becomes if the people following the political candidates on Twitter are only following those whom they agree with/like (which seems reasonable), do the tweets work in concert with other media to reinforce their (perhaps inaccurate) worldview.

  5. Thanks for the amazing analysis. It’s great to see the data supports what most of us had observed.

  6. Pingback: Twitter in the Tower | Carleton Now

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s