Twitter, TV Audience and Italian Political Talk Shows


From the end of August 2012, I started collecting Tweets containing the official hashtags of the eleven political talk shows (Ballarò, Servizio Pubblico, Otto e Mezzo, Porta a Porta, Piazzapulita, In mezzora, L’ultima parola, In onda, L’infedele, Agorà e Omnibus ) aired by the Italian free-to-air broadcasters during the 2012/2013 TV season.

I ended up with a dataset of 2,497,885 Tweets retrieved between 30th of August 2012 to 30th of June 2013. As far as I know, this collection is unique because it covers an entire season of a TV genre and because it is complete as it was acquired from the Twitter Firehose (via GNIP).

Most of the Tweets (76%) were published during the airtime of one of the 1,077 episodes broadcast during the season. When compared to the whole dataset (51%), Tweets published during the airtime tend to be more frequently (59%) original (not ReTweet or Reply).

For each aired episode we extracted the following metrics: number of Tweets created during the airtime, the number or replies, reTweet and original Tweets — Tweet-(ReTweet+Reply) -, the number of unique contributors, the reach (total sum of followers for each non unique contributors), the average rate of Tweets (TPM) and contributors (CPM) per minute, the ratio between unique contributors and average audience. For each episode, we also counted the Tweets containing a reference to one of the three main political parties (PD, PDL and Five Stars Movement or M5S) and collected the average audience and rating as estimated by Auditel.

Ballarò, with an average audience of more than 4 million viewers per episode, was the most viewed political talk show of the season. However, Servizio Pubblico, with an average of more than 17,000 Tweets and 5,000 contributors per episode, was, by far,the most popular on Twitter. It is worth to note that the average values for Servizio Pubblico are deeply affected by the numbers registered during the episode aired the 10th of January 2013 with Silvio Berlusconi as the main guest (Figure 1).

Figure 1. Distribution of Tweets published during the airtime by day

While the dominance of Servizio Pubblico extends to most of the metrics, there are two interesting exceptions: Piazza Pulita score the highest rate between average contributors and average audience for episode. This ratio represents the inclination of the audience base of a show to contribute to the conversation with the official hashtag while the show is on air. The second exception is Agorà ranked n. 1 for average percentage of replies per episode.

The three main political parties were mentioned regularly during the season. The most talked about was the Democratic Party with an average of 12% of Tweets per episode. PDL and M5S accounted respectively for 9% and 8% of Tweets per episode.

However breaking down the season in four parts according to the main political events (pre-campaign and primaries, campaign, elections and formation of the new government) clearly points out how the conversations around M5S increased significantly (Figure 2) after the election days (24th and 25th of February 2013).

Figure 2. Percentage of Tweet per episode mentioning the three main political parties

It’s interesting to note that the Tweets mentioning the parties are not equally distributed across the different shows. In 1/2 h is by far the program evoking more discussion around the Democratic Party. Agorà and Piazza Pulita stimulated more discussion on M5S while again Agorà and Omnibus elicited more conversations around PDL.

A table with a summary of this metrics per show is available here.


Drawing on this data, I’ll publish in the next few days a working paper titled “Exploring Correlations between TV Viewership and Twitter Conversations in Italian Political Talk Shows”. The study confirms, in the context of Italian political talk shows, the existence of a statistical correlation between Twitter activity and TV audience.

On this topic I just finished reading the press release issued by Nielsen concerning the study on Twitter driving TV viewers and the other way round. I’m not entirely sure about the methodology used (the available documents are not in depth enough to clarify it). My guess is that they identified the direction of the causality by looking at spikes in Tweet-per-minute preceding spikes in TV rating. I’d love to get more details on this study and maybe, with the help of someone from Italian Auditel providing the audience data by minute, attempt to replicate this study on my dataset.

Email me when Next Media & Society publishes stories