Leaders ’ ‘ Green ’ Posts
A Facebook application called Netvizz was used to conduct this analysis . Netvizz allows an automatic download of all posts published in the Facebook pages of the leaders under consideration . A total of 99,234 posts were discovered . Among these 99,234 posts , 25,151 posts were manually coded and analyzed . N The guiding criteria applied for the sampling process were the following : for each leader , we analyzed one post every 2 days and one post every day when the leader was involved in an election campaign . O The coding process concerned all leaders with a Facebook page that had published at least 50 posts . 89 out of 127 cases matched this criterion . Therefore , the content analysis concerns only a portion of the cases included in Table 2 .
Our analysis focused on two main variables : the topic covered in the posts and the type of policy issue covered in the posts . The first variable examines if the leader has talked about a policy issue and differentiates between posts about policy issues and other posts where leaders do not mention at all a policy issue , such as those cases where they talk about their private life , a news story or campaigning issues . For the second variable , we looked at leaders ’ posts with an explicit reference to a policy issue . The type of policy issue has been coded in 10 different categories : ( i ) environmental policies ; ( ii ) primary sector ; ( iii ) secondary sector , energy and infrastructure ; ( iv ) tertiary sector and public administration ; ( v ) economy ; ( vi ) social policies ; ( vii ) security and immigration ; ( viii ) foreign policy ; ( xi ) human and civil rights ; and ( x ) institutional reforms and justice . This second variable could have also been categorized by using the Party Manifesto Project ( Volkens et al . 2012 ). However , the use of this typology has been confronted with some criticism over the years ( for example , see Zulianello 2014 ). Therefore , we opted for a more parsimonious and autonomous categorization .
To ensure inter-coder reliability , a random selection of 523 posts was independently coded twice . Krippendorff ’ s Alpha turned out to be 0.82 for the first dependent variable and 0.89 for the second one , indicating satisfactory results for the statistics .
We conducted the content analysis of Facebook pages of 89 leaders and the percentage of posts about environmental issues contained in these Facebook pages represents our dependent variable . After a descriptive outline of our data , we conducted bivariate analyses in order to test the research hypotheses from which we specified six independent variables such as : ( i ) gender ; ( ii ) age ; ( iii ) party orientation ; ( iv ) economic well-being of country ; ( v ) effect of an economic crisis ; and ( vi ) environmental quality of a country . These variables were taken into consideration to identify the
N
To tackle the different languages in our data set , we used online translation services to code posts in languages such as Finnish or Greek .
O
Among the pool of coded posts , there can be the case of a hypothetical leader A who published 150 posts over a period of 400 days with no election campaigns in the period under consideration . In this case , all the 150 posts would have been coded . There can also be the case of a second hypothetical leader B who published 2000 posts over the period of 400 days with no election campaign . In this second case , 200 posts would have been coded by following a casual sampling process . There can also be the case of a third hypothetical leader C who published 2000 posts in 400 days and had been involved in an election campaign . In this case , we would have coded 230 posts . More precisely , 170 posts would have been coded during the 340 days when the leader was not involved in an election campaign , plus at least 60 further posts during the 60 days before the corresponding election day . The choice to double , if necessary , the number of posts coded during the 60 days before an election day , follows the empirical observation of the data . With the unique exception of Werner Faymann ( Austria ), all the leaders included in the dataset increased significantly the number of posts published in their Facebook pages during election campaigns .
78