Geolocating the #Occupy movement – surprising results and importance of scale!

Posted by on Mar 1, 2012 in Mapping, Social Networks | 0 comments

We’ve been working hard at SoMe Lab to begin processing some of the Twitter data we’ve pulled.  We’ve found some surprising results!  Click through to view maps and read interpretations of Twitter data as they relate to particular #occupy movements (#ows, #occupyseattle, and #occupyoakland / #oo).

One of the fascinating aspects of Twitter information to a geographer is the presence of a “geolocation” tag.  This allows Twitter users to tag their locations with a latitude/longitude coordinate pair and thereby designate their tweet as originating from a specific geographic location.  The SoMe Lab team has been collecting a series of over 175 #occupy related hashtags since early October, and we’ve just finished compiling, processing, and analyzing those tweets in a very exploratory fashion.  Remember: you can click on any of these images to see a larger version.

A global map of Occupy tweets, where one green dot equals one tweet.

Global map of #occupy related hashtags; 1 dot = 1 tweet from Oct. 19 - Nov. 5

This is a very “quick and dirty” map of the Tweets we’ve collected from Oct. 19 until Nov. 5.  We find it interesting that despite some representations in mainstream media, #occupy has garnered international attention and activity.  Keep in mind that geolocation is an “opt-in” technology — that is, folks using Twitter are intentionally choosing to share their location associated with the information that they share. After opting in, tweets will be geolocated until the feature is turned off, so while we can’t say that users are transmitting their location because of #occupy, we can make some observations. There are some caveats with this dataset: we haven’t yet discovered a way to remove “noise” from the dataset, and even if someone uses an #occupy-related hashtag, it doesn’t necessarily denote sympathy with the movement itself.  All the same, we can see that there is something inherently “spatial” about the dataset, even if we can’t yet explain why that might be.  There is also a danger to mapping at this scale: at a global level, we may only be mapping centers of high population density.  It makes sense then to examine the data at a local scale, focusing on particular #occupied locations.

Tweets for New York City, Oct. 19-Nov. 5; yellow star = Zucotti Park

Map showing point location of tweets for Oakland

Tweets for Oakland, Oct. 19-Nov. 5; yellow star = Frank Ogawa Plaza

Map displaying geolocated tweets for Seattle

Tweets for Seattle, Oct. 19-Nov. 5; yellow star = Westlake Park


 

Looking at the #occupy related tweets at the scale of the city shows a large cluster of geolocated tweets surrounding the physical site of occupation for that city. We use a spatial analytic method called “average nearest neighbor” to statistically determine whether or not this distribution could have occurred by random chance. It’s probably no surprise (you can tell by looking) that there is significant clustering within each of the cities’ data we explored (p < .001 for the nerds out there).  But what if we zoomed in further?

Map displaying geolocated tweets for Zuccotti Park

Zuccotti Park - red line = 1000' away

Map displaying geolocated tweets for Frank Ogawa Plaza

Frank Ogawa Plaza - red line = 1000' away

Map displaying geolocated tweets for Westlake Plaza

Westlake Park - red line = 1000' away


 

This is where things start to get interesting!  We already know that very few tweets within our dataset are actually geocoded (roughly 0.5%). However, when we look at the number of geolocated tweets within a city, we find that a remarkable percentage fall within 1000′ of the physically occupied location for that city (37% of NYC’s geolocated tweets, 47% of Oakland’s, and 48% of Seattle’s).  Given that Twitter’s geolocation function is “opt-in”, we can assume that geolocated Twitter users involved in the #occupy discourse are likely intentionally attaching their statements to the occupied location — or perhaps they just left geolocation turned on after using it previously.  Regardless, it shows that there is likely a spatial component to social media that we’re unable to explain (yet).

But what does it all mean?  Beats us!  That’s what’s so interesting about this work; there’s an awful lot of exploration to do before we can even begin asking questions about meaning.  But this is the first time we’ve had to examine this information spatially, and we’re very excited to share it with you.

Have any questions?  Want to tell us what you think this might mean?  Leave something for us in the comments!

Leave a Comment

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>