Paper by Axel Bruns, Darryl Woodford, and Troy Sadkowsky presented at the Association of Internet Researchers conference, Daegu, Korea, 22-25 Oct. 2014.
Axel BrunsProfessor at Digital Media Research Centre, Queensland University of Technology
1. Exploring the Global
Demographics of Twitter
Axel Bruns, Darryl Woodford, and Troy Sadkowsky
ARC Centre of Excellence for Creative Industries and Innovation
Queensland University of Technology
Brisbane, Australia
a.bruns / dp.woodford / t.sadkowsky @ qut.edu.au
@snurb_dot_info / @dpwoodford / @tsadkowsky
http://socialmedia.qut.edu.au/ / http://mappingonlinepublics.net/
2. THE GLOBAL TWITTER USERBASE
• Limited information:
– Some details from Twitter in IPO documents and updates to shareholders
– Some commercial research (e.g. Socialbakers), using unknown methods
– Some extrapolations based on surveys, usually country-specific (e.g. Pew
Centre, Sensis)
• Significant gap:
– Difficult to assess global and national Twitter patterns without background data
– Impossible to assess relative size of local Twitterspheres without global data
3. OUR APPROACH
• Data gathering and analysis:
– Slow crawl through Twitter account ID numberspace, from ID 0 to ID ~2,000,000,000
(as of 31 Aug. 2013)
– Retrieval of all publicly available profile information for each ID from Twitter API
– Storage and processing in Google BigQuery database
– Data analysis via Tableau Desktop
• Limitations:
– Accounts, not users (!)
– All dates in AEST
– Single snapshot, blurry edges: accounts created/deleted during data gathering
– Twitter API not always a completely reliable data source
– Data for most recent accounts will have changed substantially since gathering
– No information available about deleted historical accounts
– Privacy concerns (rightly) prevent any detailed profiling
7. ACCELERATING GROWTH?
• Oldest surviving accounts:
– @jack (ID #12), @biz (ID #13): 22 March 2006
• Daily growth:
– From ~200k (March 2009) to ~900k (August 2013)
• Monthly growth:
– From ~5m (March 2009) to ~26m (August 2013)
• But:
– Dataset only includes accounts which still exist:
growth = accession – attrition
– Attrition is likely to increase with time: most recent accounts least likely to be
deleted soon
17. http://mappingonlinepublics.net/
@snurb_dot_info
@dpwoodford
@tsadkowsky
@jeanburgess
@timhighfield
@socialmediaQUT – http://socialmedia.qut.edu.au/
This research is funded by the Australian Research Council through Future
Fellowship and LIEF grants FT130100703 and LE140100148.