2017 Radboud Summer School – Social Media in Political Communication and Journalism

So, last year (2016) my course on social media in political communication and Journalism premiered in the Radboud Summer School. This year (2017), I’m glad to report, I’ll be teaching the course again!
My experiences of last year’s course were very positive, and the students thought so too! It’s an intensive week – entire days of lectures and assignments – and many social activities in the evenings. There will be some tweaks to the course’s design, but the core of the course will remain the same.

In short, the morning sessions focus on theories about social media in political communication and journalism. The afternoon sessions are hands-on sessions on data collection, and data analysis. The afternoon sessions will particularly focus on what you can and can not do with social media data. You will also perform some analyses yourself. The main tool for social media analysis will be R. Why? In short, it’s free, flexible, has special packages for social media data, works on Windows, Apple and Linux computers. There are many more advantages and of course some disadvantages (there’s no perfect software package).

Regarding the afternoon hands-on sessions, we will use several English language social media data sets, available for empirical analysis. These are tweet data, Facebook posts, profile data as well as network data.

DAY 1 (AUGUST 7, 2017)

  • Kick-off meeting of the course


  • Introduction to social media, theory and data
  • Short presentation of student’s project interests

DAY 2 (AUGUST 8, 2017)

  • Approaches to studying social media data in journalism and political communication


  • The collection of social media data: API’s and software tools
  • Empirical descriptions and visualization of social media data
  • Approaches to longitudinal analysis of social media data

DAY 3 (AUGUST 9, 2017)

  • Social media content analysis: theory and methods
  • Content analysis: political and news issues, sentiment analysis


  • Different approaches to analyzing social media content data
  • Analyzing social media content

DAY 4 (AUGUST 10, 2017)

  • Social network analysis on social media data
  • Different approaches to networks on social media: social relations, retweets and texts


  • Analyzing social networks and communication networks on social media

DAY 5 (AUGUST 11, 2017)

  • Advanced approaches to analyzing social media data
  • Setting the agenda for social media research in journalism and political communication


  • Student presentations
  • Reflection

Further general information about the Radboud Summer School can be found here. For specific information about my course, please click here. And … maybe we’ll meet in Nijmegen :-).

Below are some pictures of the course.

Norris’ dimensionality of web features on political party websites tested

When analysing the data on web campagining of the EP elections of 2009, I re-analysed the data Pippa Norris was so kind to let me use. My intention was to show the need for testing on multi-dimensionality of a set of indicators. Originally I intended to include the following in the article about EP web campaigning (Vergeer, Hermans & Cunha, 2013), but this was a classic case of ‘killing your darlings’, because the article became too long.
Norris distinguished the two dimensions of Information and Communication a priori, while I wasn’t sure when reviewing the indicators. Some of the indicators didn’t seem to fit the dimensions. For instance, the search function is seen as an indication for communication, whereas this automated feature might also be included to measure informing. To see whether there is different view on the set of indicators, I performed a multiple correspondence analysis (i.e. multivariate cross tabulation (cf. Greenacre, 2007). This resulted in a different dimensional structure. The slide shows that only a small number of web features indicators compose a single dimension[i]. The remaining indicators do not show significant (co-)variance for a second dimension.

Interpreting the dimension shows us that it refers to the presence (right from the origin) and absence of specific web features (left from the origin): the more to the right the more features political parties have on their websites. The second finding is that looking at the labels C (for communication) and I (for information; cf. Norris 2001) these features are randomly scattered across the dimension. This suggests that web features that were assigned to two different dimensions should be merged according to the correspondence analysis. More close inspection also suggests that these features refer to enabling the website visitor to enlist to participate in parties’ activities.
Below the horizontal line, so-called passive variables are presented for descriptive purposes[ii]. Looking at which parties score low or high on the degree of enabling people to participate, we see that it is in particular a) the major and the minor (and not the fringe parties), b) the Green, extreme left, and conservative parties that offer these features more than average. In particular, the liberals and the center parties show below average presence of participation web features.

Greenacre, M. J. (2007). Correspondence analysis in practice. Boca Raton, Fla. ; London: Chapman & Hall/CRC.
Norris, P. (2001). Digital Divide? Civic Engagement, Information Poverty and the Internet Worldwide. Cambridge: Cambridge University Press.
Vergeer, M., Hermans, L., & Cunha, C. (2013). Web campaigning in the 2009 European Parliament elections: A cross-national comparative analysis. New Media & Society, 15(1), 128–148. http://doi.org/10.1177/1461444812457337

[i] Eigen value = 6.560, Inertia = .205, Cronbach’s α = .875.
‘C’ and ‘I’ indicate the original communication and Information functions as distinguished by Norris (2001).
Categories above the horizontal line belong to variables that influence the dimensional structure. The categories below the horizontal line are supplemental variables that do not influence the dimensionality and are merely included for descriptive purposes.
[ii] Supplementary variables do not influence the dimensionality which arises on the from the analysis on web features (cf. Greenacre, 2007).