With the craziness of this year’s ISA wearing down and the jetlag having finally retreated, a bit of reflection on the conference is in order. Much has been said prior and after the ISA, which is why I want to offer a somewhat different view of the conference: the ISA as it happened on Twitter.
Why should we care about Twitter at an academic conference like this? This year’s ISA was certainly extremely active on Twitter. The official program lists 5073 participants and I’ve counted 629 unique Twitter users mentioning the “official” hashtag #isa2014 at least once in their tweets (I’ll provide some background on the data below). Now, the number of participants is certainly higher, given the many exhibitors who don’t present on panels and other people simply attending the conference without presenting. Also the number of Twitter users mentioning the hashtag #isa2014 does not precisely reflect the number of tweeting ISA attendees – people can be tweeting about #isa2014 without actually being there. At the same time, not everybody tweeting from the conference necessarily includes the #isa2014 hashtag in her or his Tweet. So my numbers are skewed.
Nevertheless, a number of about 5100 conference attendees of which roughly 600 are using Twitter puts the percentage of tweeting ISAlers at around 11 per cent. Although this is just a ballpark estimate and may well be higher or lower, it is much higher than the average percentage of Twitter users in the world which is at roughly 2 per cent (if we are to believe Twitter user statistics and world population data). There was even a separate roundtable entirely devoted to discussing the role of Twitter in Academia. (More on that later.) And we’ve all seen the monitor next to the blackboard in the Sheraton Lobby, showing #isa2014 tweets in real time.
With Twitter playing such a prominent role at the ISA, I wanted to see if this role could somehow be quantified.
Data
This wasn’t as easy as I thought. I started downloading the all Tweets containing #isa2014 on Thursday (3 April) afternoon, using the excellent R package TwitteR. But when I took a look at the resulting list of tweets, I noticed that none of them were older than Thursday, 27 March, missing out the entire Wednesday of the conference. A quick look in the documentation of the Twitter API revealed that Twitter does not allow you to programmatically access Tweets older than 6 to 9 days. Seven days, in my case.
This was unfortunate, since I was missing the first day of the conference. So I took the names of all unique Twitter users from my initial, incomplete data and downloaded all the individual Tweets from these users, containing their user name and the #isa2014 hashtag. My hope was that the API’s search function reaches back the entire nine days if you’re looking for Tweets from individual users, rather than searching the complete Twitter timeline. This approach more or less worked and produced Tweets from Wednesday, too—but not very many, especially compared to the number of Tweets from the subsequent days. Unfortunately, it did not produce any Tweets older than Wednesday, 26 March.
This low number of Wednesday tweets questions the completeness of the data: at best, the following analysis is based on an incomplete reflection of Twitter usage. At worst, it is systematically biased, especially if there was heavy Twitter usage on Wednesday, which is mostly missing from the analysis. From constantly watching my TweetDeck’s #isa2014 column (yeah, I may have attended some panels, too), my personal impression was that Twitter usage took up heat only some time after the conference had started, so maybe the bias is not too severe. But I definitely know that I will set up a cron job for next year’s ISA to download each day for #isa2015 right at 0.00am and save it to get a complete snapshot. All in all, my procedure yielded 2940 unique tweets containing the #isa2014 hashtag.
Analysis
With these caveats in mind, we can move on to the actual analysis. Obviously, the most important question we all want to know is: Who was the most active Tweep of the entire ISA? The following graph lists the 20 busiest Twitter users using the #isa2014 hashtag.
With a stunning number of 190 Tweets, Annick Wibben clearly leads the field, followed by Raul Pacheco-Vega and FTGS (which is the official account of the Feminist Theory and Gender Studies section of the International Studies Association).
Some caveats are in order, though: this figure does not necessarily represent an account of the most active persons on Twitter at #isa2014, but rather gives us the list of people who used the #isa2014 hashtag most frequently. These two things are not the same. For instance, Charli Carpenter and PTJ’s epic debate about whether Star Wars constitutes sci-fi or not (BTW, it’s clearly sci-fi ;) ) is not included in these figures. They used #isa2015 to refer the debate to next year’s ISA, so, I guess we should include those in next year’s figures. Turns out their conversation was in the data, I just missed it. But Star Wars is still sci-fi. :) Also, other live tweetings etc. of panels that didn’t include the hashtag don’t make it into the analysis, which would change this overall figure dramatically. See the discussion of Twitter usage during the separate Twitter panel below. The numbers here also include retweets; they are not entirely made up of original Tweets by the persons on the list. But since a retweet represents a Twitter interaction, I included them in my analysis.
When do people Tweet?
The next question we all want answered is: when do people tweet? The following figure sheds light on this mystery.
We see that this is pretty much what we would expect, a lot of buzz during the day, less so late night, except for some jetlagged Tweeps from overseas. This is actually a hypothesis one could test with the data here (which I haven’t done): are Twitter users from overseas more likely to tweet later in the night (or earlier in the morning, given the direction of the time difference)? We also see a lot of goodbye Tweets following the conference as the numbers peter out in the end.
The figure indicates that something might be wrong with the data, as I mentioned above: is there really a steady increase of Tweets per day or did I simply download more Tweets from those days because of Twitter’s API restrictions? We won’t know, but I’ll be sure to test this hypothesis with more complete data next year.
The Twitter Panel
Also, this figure might be suspicious to anyone who actually attended the ISA: wouldn’t we expect a spike in Tweets on Thursday afternoon (27 March), since there was the dedicated Twitter panel? In fact, the attendees were insanely busy on Twitter during the Twitter panel:
But the Twitter panel used the separate hashtag #isatw14—and most attendees and panelists did not include #isa2014 in their #isatw14 Tweets. This is why these Tweets don’t show up in the overall #isa2014 statistics. If we look at the #isatw14 Twitter usage separately, a different pattern emerges. Take a look at the busiest Tweeps in the Twitter panel:
With 61 Tweets, Laura Seay (@texasinafrica) clearly leads the field of the Twitter panel tweeps. Need I mention that she was also a panelist and gave smart replies while tweeting at that pace? Multitasking FTW. If we add these numbers to the first figure, the overall ranking obviously shifts (figure not included).
What were the most popular Tweets?
Now, I’m already hearing the criticism of my stats-suspicious friends: enough with the numbers! What were people actually talking about? Although I don’t attempt any textual analysis of #isa2014 Tweet content (which might be a nice follow-up blog post idea for someone who actually knows how to do this), I’ve collected a list of the most popular Tweets of the conference. The idea here is that the most popular Tweets somehow reflect something like a shared agreement on statements about the ISA (an assumption which is, let’s say, at least questionable). But this being a blog post, I went ahead and did it anyway.
This obviously begets the question: how do you determine a Tweet’s popularity? I took the quick and dirty road and simply added the number of a Tweet’s retweets and its count of favorites. Now, this might not be the best index-generating process in the world (some people could say retweeting weighs more, because it reflects active endorsement or at least engagement, whereas a favorite is more like a Facebook “Like”). You are free to give me hell about this indexing process in the comments. But this simple procedure produces the following list of the 10 most popular tweets at ISA2014 in descending order:
Ah, now it gets interesting. The first Tweet, which received the highest number of retweets and favorites, gets at somewhat we’ve all experienced: questions that turn in to co-presentation which may or may not close with an actual question. This cartoon pretty much sums it up – it was actually shared quite often on Twitter during the conference:
The other top tweets reflect much of Twitter’s usual snarkiness, which is refreshing when you are hearing dead serious presentations the entire day. But the top 20 Tweets also reveal that this year’s ISA president Amitav Archya hit a nerve with his speech calling for an opening up of the study of world politics to voices from the Global South. Interesting stuff.
Again, this list is incomplete because my data does not reach back far enough. Because Laura Seay (@texasinafrica) tweeted – in my opinion – the most important information for the entire conference:
With 31 retweets and 30 favorites this tweet easily makes it to the second place in the previous list of most popular tweets, but it wasn’t included in my data because of Twitter’s time restrictions. Note that some of retweet and favorite counts have changed since I downloaded the data which is why the order of the Tweets from highest popularity score (retweets + favorites) to lowest is not entirely correct anymore.
Finally, the Twitter panel also collected the combined wisdom of the experienced Twitteratis-slash-academics. Thankfully, Simon Usherwood neatly summarized the major point in a list of Tweets which I’ve storified here. I guess the main take away of this panel was this:
Conclusion
I’ll leave it at this for the moment. Following ISA on Twitter while attending the conference is usually a lot of fun, sometimes provides context to the panels you’re attending (and a lot of snark) and gives you much needed distraction in others. It obviously doesn’t replace attending the conference, but gives you a meta-level discussion that nicely puts many of the interesting (or boring) events at such a conference into perspective.
From analysis point of view, this quick exercise raised for me the question if and how we can capture a more complete snapshot of ISA tweets (besides getting more complete data by downloading earlier)? Because taking the hashtag as anchor point misses much of the sideline discussions between conference attendees tweeting at each other. Maybe one approach would be to locate Tweets geographically, but only very few users turn on their geotagging function of their Twitter client. I’ll give this more thought, but I’d be interested in the opinions of the Twitter-savvy readers on this.
As I said, I’ll try to get a more complete snapshot of ISA tweeting next year—in the meantime I invite everybody to have a look at the data themselves. You can download the .csv of the 2940 ISA Tweets here. I’ve also uploaded my R code to Github, but given the time restrictions of the Twitter API you probably won’t be able to replicate the graphs and tables. Maybe it’s useful for anyone trying to replicate this for another conference or event.
Bretterblog 
Felix, this is a great post and extremely interesting. I’m already looking forward to next year’s results. Maybe I will have set up a personal twitter account by then..
This leads me to a question that I have been thinking about for a quite a while now: How about German tweeps using #ISA2014?
And in general: (Why) Do German polsci / IR scholars (not) use twitter in general? Along this line and since you mentioned doing this analysis for other conferences yourself: What about the ‚IB Nachwuchstagung‘ in Tutzing and ‚Offene Sektionstagung IB‘ in Magdeburg? (for non-Germans: these are the IR junior scholars conference and the IR section’s conferences)
Excellent questions, Sören. Thanks.
Re German Twitter users at #isa2014. Analyzing German Twitter users isn’t easy, because Twitter doesn’t have a country code for its Tweets. It does have a „location“ field, but you’d need to write a script looking up the entry in that field (which many users leave empty, however) and locating whatever string is in that field in a country. This possible, but tedious and probably very imprecise. So, no obvious and easy way to do this programmatically, at least AFAIK. But maybe someone else has an idea?
Re why there are few German poli-sci/IR scholars (not) using Twitter: Honestly, I don’t know. Probably for the same reason they don’t blog. We did a non-representative survey among several German political scientists some time ago which might give you some ideas: https://bretterblog.wordpress.com/2012/08/29/umfrage-die-deutsche-politikwissenschaft-im-web-2-0-teil-1/ One reason might be that social media hasn’t really arrived in German political science because it hasn’t really arrived in other fields of science either (I might be wrong here). But I can only speculate. Again, maybe someone else has more / better insights on this.
Re using this kind of analysis for other conferences. This should be easy, as long as tweeting conference attendees use the same hashtag as an anchor, such as #ibtutzing14 etc. It’s simply a matter of adapting some lines of the code I wrote for this post. So if this is happening, let me know (I won’t be at either conference) and I’ll see if I can do something similar along the lines of this post.
[…] Haas has written a monster blog post over at Bretterblog with interesting descriptive statistics from the #ISA2014 hashtag. Among other fascinating […]
Awesome post! However I didn’t understand why the bulk of PTJ and my Star Wars tweets were excluded. Almost all included the #ISA2014 hashtag; only the final one about a roundtable at the next conference included the #ISA2015 hashtag. Just saying… I wonder what it would do to PTJ’s prominence on the most prolific tweeter list if those were included. :)
Thanks for the shout-out at the Duck, Charli! I just checked; your Star Wars conversation is indeed in the original data. I must have missed it somehow while sifting manually through the large number of tweets. I’ve corrected the text in post. The original figures do include the count of the Star Wars tweets for both of you, however, because the computer didn’t miss them when aggregating the data. So no changes here, but you could view it that way: in the data I have (with all the mentioned caveats about completeness in mind), the SW tweets make up roughly a 1/3 of PTJ’s #isa2014 tweets. So you can imagine the changes to the most prolific tweeter list if he hadn’t tweeted about Star Wars. (FYI, if I’m counting correctly, the Star Wars tweets make up also roughly 1/3 of your tweets).
Great post, two quick suggestions:
1) Start downloading a couple weeks *before* the meeting; excitable tweeps often start tweeting while preparing and travelling to make sure their friends are coming.
2) People who tweet a lot at conferences but know not all their followers want to read that will often direct their tweets at the main conference account. This means only followers who also follow the conference see the tweets, which is perfect. But due to character restrictions, you often don’t include the #tag in this case. So search for the text of the #tag OR the conference account, leaving off the # or @. This way you’ll get a lot more tweets.
Good luck!
Thanks! Extremely good points. I’ll keep them in mind when next year’s ISA is approaching.
[…] to #ISA2014 – which Felix Haas has thoroughly analyzed – the #MPSA2014 Twitter activity was underwhelming. @EvilMPSA and @DrunkMPSA were great, […]