#DataHack2: 2014 South African Election Twitter Interactions

This is a continuation into an analysis dive into South African election related Twitter data. Previously I created DataHack1 and after reading this post you can also see DataHack3.

UPDATE: This post made in onto iAfrikan (in a nicer edited format): Link, Evernote Mirror

Twitter Interactions

Political party or leader engagement is always an election time fascination of the media. Whether it’s all of those photo ops kissing babies, shaking hands or those one night stints sleeping in a shack in Diepsloot. So how can we use Twitter data to measure this engagement? We can track the number of interactions that the leaders have with other users. Her we an interaction as either a retweet of the party/leaders tweet(s) or a Twitter mention (@partyleader,@partytwitteraccount). Lets look at a timeline of these interactions. First the leaders:

Party Leader Interactions

The leaders I chose to examine are:

• Jacob Zuma: ANC (@presidencyza)
• Helen Zille: DA (@helenzille)
• Julius Malema: EFF (@julius_s_malema)
• Mamphela Ramphele: AGANGSA (@mamphelar) (note I had an error in the filter so this handle might be underreported)

Obviously there are a lot of political parties in South Africa, with even more leaders between them, but to keep it readable I only examined those above. The good news though is that full dataset is available for anyone to delve into and do more analysis (detailed at the end of post). Below are the aggregated interaction counts from the 11th of April to present.

Here its clear to see that @helenzille is always talk of Twitter town. She dominates with the shear amount of interactions users are having with her. On the 24rd of April, both Hellen Zille and Julius Malema both started increasing their interactions. Not all interactions are positive though. More on that later. Jacob Zuma has low interactions because he rarely uses his Twitter account to interact with users. Especially in matters relating to ANC politics.

Party Twitter Account Interactions

To mirror the above, I examined:

• ANC (@myanc_)
• DA (@DA_News)
• EFF (@EconFreedomZA)
• AgangSA (@AgangSA)

The results for the parties are presented below.

Here we start to see that the political party Twitter accounts are used more by the ANC, DA and EFF. AgangSA is not really getting much traction. The DA and EFF have been swapping their daily leads in terms of totals but ANC is also in there. There is more to explore here but that will be for another day or the user can dive in by visiting the data repository detailed below.

Interactive Plots

You can view interactive versions of the above plots on the GitHub project page.

So what causes the fluctuations?

A natural question that arises when looking at the trends is what causes the movements for each account we are monitoring. So for example if looking at Julius Malema on April 23rd one notices an increase in total engagement. This becomes even more evident when analysing the word frequencies in the tweets on that day. You can see this in the graph below.

The histogram even reveals that there was a big conversation around an “open letter”. When diving into this it reveals a big topic on that day regarding this tweet:

Open letter to Julius Malema. Dear Julius Malema, I wronged people in this country two decades ago when I (cont) http://t.co/UKhZ8shzu3

— Gayton McKenzie (@G_XCON) April 23, 2014

Which was the most retweeted tweet in the dataset for the day. This trend was simple to find and is a good heuristic to use to annotate the original graphs. But then a next question could be “Are there other people who are garnering more interactions than the politicians on these days?” The answer is yes.

Here are the top 5 most retweeted on the day (in decreasing order of retweets):

• @g_xcon (# retweets 642)

• @da_news (430)

• @vusani_ (389)

• @khayadlanga (151)

• @djsbu (145)

• @city_press (144)

•  @myanc_ (120)

• @econfreedomza (117)

• @news24 (109)

•  @mbalulafikile (107)

As always there is more that can be done with this data. This is just a taste of what’s possible.

Data

As always, I made available the Twitter JSON/CSV dumps at my GitHub. Grab the continuously updated data here ->  github:za-2014-election-tweets

iPython Notebooks

• How to create the Timelines

Bonus: Google Trends

Google also has their own search trends that they are compiling. See below