Negativity in Parties' Press Releases in the 2015 Canadian Federal Election
Description
To collect the raw data, I used parties’ press releases during the 2015 campaign, starting from August 1st to October 19th. I count every single attack in each press release and add them up, to calculate the number of negative statements against other candidates or parties in a press release. However, if in one statement, there is more than one attack toward other leaders or parties, each statement counts only as one negative attack. Also, to capture the difference between attacks against the leaders and the parties, if there is direct mention of a party leader’s first or last name in a statement, it is considered as a negative attack against a leader. For example, if in a single press release, there are two criticisms against Justin Trudeau, but in one sentence he is called “Justin” alone, and in a different sentence he is called “Trudeau,” the number of attacks in this press release is counted as two. Also, since the unit of analysis for this study is days of the campaign, in case of having more than one press release in each day, the number of attacks in each press release is added to others to obtain the overall number of attacks per day. Thus, if there are two attacks in one release and three in another on the same day, I count five attacks on that day. Also, to make better sense of data, I calculated the proportion of attacks against each party or leader per day by dividing the number of attacks per day against each case by the total number of attacks that a party published per day. Since two primary factors in our analysis, which act as independent variables, are voting intention for parties and party leaders’ popularity, I use publicly available polls to create my dataset. In order to track the federal voting intentions, I use CBC Poll Tracker (CPT) which aggregates all public polling data during the election, including Nanos Research, EKOS Research, Léger, Ipsos, Angus Reid Institute, Abacus Data, and Forum Research. Although this is the most comprehensive dataset for the polls during the 2015 Election, it does not include polling results of every single day of the campaign. To fill this gap, I use the average of +/-1 day which reflects the closest estimate possible, considering little possible fluctuation in a day, particularly at the beginning of the campaign. For capturing the second crucial factor, party leaders’ popularity, I use Nanos Research data, since it is the most consistent and available data about the leaders’ popularity during the campaign. However, unfortunately, these data are only available on a weekly basis. In order to align my dataset with this weekly leaders’ popularity rate, I aggregated all the attacks per week and used the total number and their related proportion in a week. Contextual analysis of parties’ press releases means that the determination of negativity in every single sentence has been done subjectively, and thus, is open to a certain degree of interpretation, particularly in a complex context.