While I was reading tweets about elections for the U.S. Senate and for the Ontario provincial government in Canada, it struck me how political speech was different between the United States and Canada. I couldn’t quite put my finger on it, but it seemed to me that Canadian political speech was more toned down than its American counterpart. With over five million tweets coming out of these elections, I set out to analyze just how political speech on Twitter is different –not just from everyday tweets, but across political parties between Canada and the U.S.
I collected 2,618,015 tweets during the recent U.S. senate races (September 18, 2014 to November 4, 2014) from Arkansas, Colorado, Georgia, Iowa, Kansas and North Carolina. I also collected 2,500,821 tweets from the June 2014 Ontario provincial election (April 4, 2014 to June 12, 2014). Both of these elections were focused on the state/province level and were held within six months of each other. While Twitter usage morphs over time, there were no significant changes in Twitter speech in the time between the two elections.
Annotations and Citations
But how is political speech characterized? Reading over my collection of political tweets, it occurred to me that the speakers were using hashtags and links in a very specific way: as annotations and as citations.
Annotations serve as bread crumbs in the conversation. By including hashtags, people were linking their contribution to a larger conversation. One such use is mapping out a conversation on a single topic, as shown by the hashtags #onpoli, #ARSen, or #NCSen for example. By searching for these annotations, a larger conversation can be reconstructed from the Twitter history. Another use is a commentary within the tweet, such as #DefeatHagan or #SayNoToHudak. Including these types of annotations allows people with similar sentiment to discover and connect through Twitter. There is the additional goal of causing such a hashtag to trend on Twitter, increasing the chance that their message is heard.
Citations, on the other hand, are used to back up a statement. People use citations in many ways. One use is corroborating a statement of fact, such as a photo showing a candidate at a specific event. Another use is as an appeal to higher authority, as is the case when citing a news article or political pundit to support an opinion. Citations can also be used to spread information, such as upcoming political events or political messaging directly from the candidate.
With such specific usage in political speech, hashtags and links were the obvious choice for my analysis.
Analysis of Hashtags and Links
Before I started analyzing hashtags and links in political speech, I needed to set a baseline by which to compare. Looking at all of the sampled tweets, I found that 0.6±0.7% of tweets from the U.S. and 0.8±0.6% tweets from Canada use links, while 2.1±0.4% of U.S. tweets and 2.1±0.4% of Canadian tweets use hashtags. This tells me that Americans and Canadians use links and hashtags in the same way, and that a tweet is more likely to contain a hashtag than a link.
With the baseline measurement set, I looked next at how the two big American parties (Republican and Democrat) and the three main Canadian parties (Conservative, Liberal, and NDP) use hashtags and links to see if there is a difference between the countries. Figure 1 summarizes this analysis.
Figure 1: Analysis of link and hashtag usage for tweets mentioning a political party.
This data contains many interesting observations. Let’s look first at the American parties. Unlike the general population (where hashtags are more prevalent than links), both political parties are as equally likely to use links as hashtags (Republican — link: 48.1±2.2%, hashtag: 46.1±2.2%; Democrat — link: 54.2±2.0%, hashtag: 53.6%±2.0%). Since a hashtag is nearly twice as likely than a link in the general sample of tweets, the equality of hashtag and link usage suggests that in the U.S., links are used more frequently than hashtags in political tweets.
Looking at the difference between tweets from Republicans and Democrats, Democratic tweets are more likely to have a hashtag or link than a Republican tweet. Including extra information like hashtags and links requires an advanced understanding of Twitter. Seeing a statistically significant difference in links and hashtags between the two parties points towards Democrat tweeters being more Twitter savvy than their Republican counterparts.
Over to the Canadian results. Statistically, the Conservative and Liberal parties use Twitter almost identically (Conservative — link: 45.4±1.1%, hashtag: 56.1±1.0%; Liberal — link: 42.7±1.4%, hashtag: 54.6±1.3%). Both political parties also follow the general population trend of using more hashtags than links. The NDP is the real outlier (link: 55.5±1.7, hashtag: 41.6±1.8%). Tweets mentioning the NDP are more likely to include a link than a hashtag.
Looking at the difference between the two countries, American political speech is more likely to include a link while Canadian political speech is more likely to include a hashtag (U.S. — link: 46.9±1.5%, hashtag: 44.8±1.5%; Canada — link: 40.3±0.8%, hashtag: 49.8±0.7%).
What Does it Mean?
I established earlier that links in political tweets serve a citation purpose, while hashtags are for annotation. Citation is used to support an opinion. Annotation is used to participate in a larger community. Tweets that use a lot of citation are therefore supporting an opinion, while tweets with a large number of annotations are participating in a larger community.
To recap: U.S. political speech is more likely to include a citation, and Canadian political speech is more likely to include an annotation. Without analyzing the tweet content in detail, I can infer that U.S. political speech, based on the increase in citations, is more opinionated than Canadian political speech. Some examples of U.S. tweets that include links are:
While #ISIS grew, Sen Hagan kept quiet. Leadership is about taking a stand. And it starts with showing up. http://…
Democrats BaROCK the vote. Republicans BLOCK the vote. #UniteBlue #PDMFNB #TurnRedStatesBlue #LibCrib http://…
And some examples of Canadian tweets that include links are:
Hudak has a low-wage agenda, but workers are fighting back. #StopHudak campaign continues in Windsor May 13 http://…
Tim Hudak’s plan for 1 million jobs doesn’t add up: Cohn #onpoli http://…
In both cases, the tweet states a strong opinion that is supported by the attached link. On inspection, this pattern is consistent for tweets with links. More link-rich tweets implies more opinionated speech.
On the other hand, Canadian political speech is more likely to be community- and conversation-oriented. By inserting hashtags, Canadian political tweeters are making it easier for others to find and follow their discussion. This does not mean that tweets with just hashtags are un-opinionated, but they do tend towards a more civil discourse than tweets lacking a hashtag.
There are probably many more insights to be gleaned from this comparison of Canadian and American political tweets. My analysis is just scratching the surface, but does indicate there is a clear difference. Americans are more opinionated in their political speech and they like to include extensive citations to back up their claims. Canadians are also politically opinionated, but they express that opinion in a more community-based conversation.