The role of sentiment - exploring citizen engagement with police via ...

Building on a two-month data set (08 April 2019 – 07 June 2019) for the 11 largest police stations across Germany that use twitter, the present paper looks at whether and how media  content related tweet features (i.e. content type, affective valence of tweets) affect three different citizen engagement dimensions. Moreover, we assess and compare the level of sentiment as well as direction of polarity (negative, netural, positive) in police and citizen tweets and the relationships between them. Given that many tweets stopped at the level of information, i.e. did not prompt any retweet or comment, the analysis builds on hurdle models which consider excessive zeros,. The results reveal that the included factors affected the three dimensions of citizen engagment (likes, tweets ,or comments) differently. With regard to the affects conveyed in a message, emotionally charged but not necessarily negative tweets are more likely to be passed on or to receive likes. However, both emotionally charged and negative tweets receive more comments. In contrast, informational tweets increase the number of likes and comments, but they are less likely to be retweeted. The impact of two media richness factors (hashtags and mentions) varies across the different engagement dimensions, in some cases even showing opposite signs. Tweets with links show a negative impact on all engagement dimensions. Finally, publishing tweets during workdays has a negative impact on the number of likes and retweest but seems to have no effect on the number of comments. The senitment analysis revealed that citizen initated tweets, on average, are of a morenegative affective valence than tweets by the police or citizens’ reactions to them.