Tuesday, July 17, 2012

the geography of klout scores - or why are the French so influential?

Most Twitter users have heard of Klout scores. These scores which fall between 0 and 100 supposedly measure influence (higher scores indicating that a person is more influential). This isn't to say that such quantification of a person's influence based on online activity is entirely unproblematic. The entire endeavour is worrying on a number of levels, and it is highly unlikely that a single number (especially a number generated using Klout's methods) could ever sum up the various ways in which influence is perceived and enacted. 

Nonetheless, I wanted to map the service in order to see how the geography of online influence (according to Klout) might vary over space. With the help of Devin Gaffney, I did just that:

Over the course of four consecutive days of polling for 30 seconds every 5 minutes from Twitter's spritzer-level of access, we collected a total of 3,598,060 geotagged tweets via the random public timeline. These geotagged tweets were then bundled into their respective countries of origin, and the resulting set of country-bundled tweets were sampled randomly for up to 1000 users. The resulting sample of users were queried using Klout's API. 

The map above shows only countries with a user sample size of more than fifty users (who publish geotagged tweets). Looking at the data, we see a very interesting amount of variation. The average score, globally, is just 26. 

France has the highest score with an average of 37.8 (taken from a sample of 837 users in the country).  
The UK (34.9), Sweden (34.8), Brasil (34.8), and Indonesia (34.2) all follow closely behind (Brazil and Indonesia are incidentally some of the world's most prolific tweeters). 

The US, which normally excels in all metrics of online visibility/power/reach comes in at 10th with an average Klout score of 33.

This isn't to say that tweets emanating from the US as a whole are not influential. The US is the world's largest source of content on Twitter. This massive amount of information, pushed through the platform, undoubtedly means that American users in the aggregate have a large amount of visibility. 

Yet it remains that they have a relatively weak average 'influence.' Nonetheless, despite the strong scores of Brasil and Indonesia, it remains that we (perhaps unsurprisingly) see that most countries in the Global South have less 'influence' than their Northern counterparts. In the list of top-20 Klout scores, there are only two countries with a GDP per capita below the world average (Indonesia and Egypt).

Kenya scores highest in Sub-Saharan Africa (in 22nd place globally) with an average Klout score of 31. Most other Sub-Saharan nations are then much lower down in the list of average influence.

This doesn't mean that there is a clear relationship between GDP (or level of 'development') and Klout scores (Australia, for instance, is in 52nd position on our list). However, with a few exceptions, poor countries tend to have relatively low scores. 

Is this because we are picking up traces of the cultural dominance of the North even in a supposedly decentralised network? (i.e. Northern tweeters might tend of have greater reach and amplification than their Southern counterparts) This finding doesn't mean much for any particular person attempting to communicate or spread a message, but still potentially sheds light on the issue of voice in the world's margins. 

On the other hand, perhaps we are just reproducing and amplifying opaque and highly problematic data. We should therefore certainly not overreach in any interpretations of these data. 

Nonetheless, I still want to know if the French truly are more influential on Twitter than everyone else? And, if so, why?


inlocoveritas said...

Interesting-- if you buy into the value of Klout, that is. French Guiana, a French territory or département, stands out in South America. I wonder how many tweet there, and whether they are disproportionately more influential than their countryfolk in France. There are a fair number of interesting geographic anomalies, after all.

Mark Graham said...

I certainly don't buy into the value of Klout. But do find it quite fascinating to map and interrogate. You're seeing the French Guiana value because it is officially part of France (thus having the same average Klout score as France). I suppose I could have separated it out. If we did that, it would have been shaded grey as it has fewer that 50 geocodable users in our sample.

Alan Bryant said...

Their algorithm definitely sucks, Argentina has more Klout than Japan... Need I say more?

Mark Graham said...

@ Alan: more Klout per capita

Breizhou international said...

Are you taking into account that Klout users can add their Facebook, LinkedIn and more accounts on Klout ?

Which would mean that your study is considering only users having one twitter account linked to their Klout ?

Also, as long as we all know more or less that Klout is not really mesuring inlfuence, why should we really consider doing some studies about it ?

Meanwhile, you could be interested in this recent study about Facebook users influence tend to be a bit more... scientific ?

Mark Graham said...

We're aren't, and that is a very good point. We're simply measuring Klout scores based on Twitter sampling (scores that could well have been influenced by Facebook and other accounts).

I'm not sure we should be putting a huge amount of resources/time into study Klout, but I find it worth deconstructing a little.

korosion said...

You should definitely add the Klout Style on the stat, in order to define what component of influence makes the difference !

Unknown said...

I think there's a very simple explanation, especially since (I'm guessing) Twitter is used as a large proportion of the sampling:
In France, we have a smaller percentage of people that are active online, but they are usually very adept and capable in that field. Hence a higher average Klout per "counted" capita.
If you compare it to the US, where everybody is on Twitter, the Klout per capita is lower because everyone and their brother (who don't necessarily do much other than browse Yahoo news) is counted as well.

Basically, we win because we our "non-Klouty" population isn't on the net much and isn't taken into account.

At least that's the way I'd explain it.

Patrick Beja said...

"Unknown" above is Patrick Beja by the way (http://twitter.com/notpatrick); I'm French, and my Klout kind of sucks. :)

D.Gabs' said...

@Patrick no no, we we are the best; that’s is. :D

François Merlin said...

I think your test is wrong. You are not measuring the klout score of a country but rather the klout score of people who tweet. It means that people who send tweets in france are, more likely, important people. So your map only shows how twitter is used around the world.