Methodology
This post is based on data gathered using the following methodology:
- Develop a list of Australian and New Zealand related Twitter accounts and sort them by sport.
- Get a lot of the followers for all those sport related Twitter accounts. (This process takes a while. I can only use one Twitter API code set. This limits the number of calls I can make to something like 150 or 350 per hour. For each of those individual calls, I can get information on one page of a account’s followers. Each page consists of about 20 followers. This means the whole run took about a week and it started on December 18.)
- Cross reference User inputted locations to actual Australian city locations .
- Combine all followers by sportinto one file. Remove duplicate entries so that if, for example Person X follows @AFL, @stkildafc, @zacd_6, @harry_o, Person X gets counted once for Australian Rules, not four times.
- Count the total number of followers by sport and by city.
- Combine the total number of followers by sport with all the cities in the electorate the city is from. For this analysis, that data can be found at Australian city location to electorates . This list was created using the official list of Australian polling places by electorate. The following cities were left off the totals: Brisbane, Perth, Sydney, Melbourne, Adelaide, Canberra. These cities were left off because they would skew the results as people that say they are living in Melbourne are actually probably living in a different suburb. (The CBD of these centers is very tiny. There are probably more people claiming to live in Melbourne on Twitter than actually living in Melbourne CBD proper.) If a town is included in multiple electorates, it is counted in all those electorates: Brunswick and Brunswick East are counted for both Melbourne and Wills.
The complete totals for the results above are available at Australian Twitter Sports Electorates Report – December 28.xls. The raw data is 74megs when zipped. If you would like access to it, please leave a comment. When there was a tie and if cricket was involved, I chose cricket. The single blacked out electorate is one where I did not have any people from that electorate.
I suspect I have under counting some where in this data, merely because the total number of sport fans by electorate is smaller than the total number of political followers by electorate. I just haven’t explored the raw data enough to understand why… or there is the issue that the crossover between Australian sport related accounts is huge. It seems like something is just off with my totals and I can’t figure out what.
