Tuesday, February 12, 2013

Mapping tweets about the Kenyan presidential debate

Yesterday saw the first ever televised presidential debate in Kenya. The debate was not only carried across a range of tv and radio stations, but was also live-streamed on YouTube and was actively discussed on Twitter under the #KeDebate13 and #KeDebate hashtags (and perhaps others).

Because Kenya is one of Africa's most active sites of Twitter usage, I thought it would be useful to map out the geographies of geocoded tweets about the debate. With a bounding box around the continent, we collected 2321 geocoded tweets mentioning #KeDebate13 and #KeDebate (between the afternoon of the 12th and the morning of the 13th). 

Most of these tweets were unsurprisingly in Kenya. Of those tweets, only Nairobi saw a significant cluster of activity. In the maps below, we map tweets mentioning #KeDebate13 and #KeDebate in Nairobi: revealing some of the urban geographies of participation in the Twitter debate.  

We see that some parts of the city (downtown, Madaraka, Kileleshwa, Kimathi, Royasmbu etc) are clearly central in the online discussion, while others barely show up (e.g. take a look at Kibera). I'll dive into the data a bit more to see whether we are seeing clusters of friends tweeting in these areas (and look at what they are saying), but in the meantime wanted to share the maps in case anyone has initial insights into these patterns.

See also:
Mapping Twitter in African cities


Anonymous said...

No tweets from Kibera. Socioeconomic reach of Twitter...

Erica Hagen said...

I personally know people who were tweeting about the debate from Kibera, like @steveKibera and @Ngitok. I think you need to look at the idea that geotagging as a reliable way of analyzing tweets. You have to have a phone with that feature, and then choose to have it turned on. Also many would be using desktops to tweet which won't accurately locate in many areas. Certainly don't disagree with socioeconomic specificity of Twitter but be careful about the assumptions as well.

Mark Graham said...

Thanks Erica. We are only mapping geocoded tweets. And aren't really making any assumptions about the reliability of geocoding. In other posts and papers, I point out that geocoding will usually only pick up far less than 5% of all tweets. The map is simply showing geocoded tweets using particular hashtags - and I'm pretty sure we are capturing most of those.

alex mann said...

Hi Mark, I wonder whether something like
could be used for this? I'm part of the dev team and we love to hear from people who have had to create a platform to view Tweets as you have. The platform is in early alpha but is certainly useable as it can filter keywords etc.