Given some sharp comments after my last post about law professors using Twitter, I decided to see if there's anything the data has to say about it. The comments I've seen are generally of the type that professors merely use twitter for self-promotion that does not promote any scholarly or legal interest. Related comments from earlier census posts were essentially of the type, "Why do you care about tweets when your law school house is burning down?" Without accepting the burning house analogy, I want to provide the evidence I found and discuss how it bears on these criticisms. In a future post, I'll discuss my view of Twitter, as well as a way to obtain more granulated results than I'm presenting here.
(Note: some of the data here is incomplete. Updated data is presented in a later post).
I'm greatly aided by the Lounge's own Twitter census as well as Ryan Whalen's Twitter data gathering expertise. Mr. Whalen graciously agreed to gather all tweets by every professor on the census for the last month. My analysis is below, but I will admit here that this is a bit of a selective sample because it was winter break. It could be repeated during the middle of fall and spring semesters, as well as the summer. That said, the analysis below is based on more than 16000 tweets over 30 days by 500+ professors.
How Many Tweets?
The first chart is a histogram of the number of tweets per person. The chart shows that most users rarely tweet. Indeed, more than 20% of the users had zero tweets during the month, which means that their usage was for consumption of tweets only. Another 5% or so had just one tweet during the month. There is clearly a long tail of active users - some with almost 200 tweets in just a month. But this is a clear minority of users. More than 60% tweet 20 or fewer times in a month.
Who Is the Audience?
The next two charts shows followers and following by school US News rank, with all unranked US News schools listed as 147. Schools unranked due to foreign location or non- ABA approval are not included. The number of followers and following are log-transformed because outliers at the top were causing the remaining numbers to crunch together so much that it was difficult to tell the dots apart. It's hard enough even as is.
I take a few things away from these charts:
1. The distribution of followers and following does not appear to have an easily discernable pattern by US News ranking, either for general followers or law professor followers.
2. The number of law professor followers and followed is generally much lower than general following/followed counts. Whalen's blog post posts specific numbers that highlights this. For example, Santa Clara profs have 11159 followers, but only about 236 or so professor followers.
3. The number of tweeters is not discernably based on ranking, but as the rankings get lower, more schools appear to have no tweeters (these appear as empty bands in the dot chart). Some of the gaps are because there are no schools at a given rank because of ties, but some is not. I entered the ranking of each school from 1 to 146 down the rankings; while most schools had at least one tweeter, I found more schools that did not as I got below 100.
4. There were several (about 25, or 5%) professors that followed zero other professors, but that followed others. This could mean a variety of things in how they use Twitter.
Does Following/Follower Activity Affect Tweeting?
The following charts plot the number of tweets in a month versus the number of followers and following. As before, the first chart is all followers/following, and the second chart is all professor followers/following.
Here's my view of this data:
1. Activity appears to be roughly correlated with number of followers. In other words, those with more followers tend to tweet more, especially at the outliers. Or vice versa. The cause and effect is likely symbiotic, but it does reflect something I've read about for other social media, like blogging: knowing you have readers creates pressure to produce more content. Further, the more followers/following one has, the more likely there will be interactive tweets. More on this below.
2. The correlation appears diminished a bit with respect to law professor followers. The number of tweets grows at an earlier point and doesn't extend so far to the right for the high end of law professor followers. This implies that the drive to tweet heavily is diminished if based solely on law professor followers. Also, the law prof tweeters who follow no other law professors have a wide range of tweets, including near the maximum. It's hard to see these professors as trying to curry favor within the academy, so something else must be going on.
3. There's a large amount of clumping at low amounts of tweets. We know this from the histogram, but we can see here that the clumping is not based on any particular amount of followers or following. If you ignore the scattered outliers, it looks a bit like a normal curve - but there are a lot of outliers.
What do Professors Tweet About?
The last chart shows my classification of tweets. It is rough; a later post will discuss how we might refine it further with more work. Here is a description of each category:
- HasLink: Does the tweet include a link to other content?
- MentionsOthers: Does the tweet mention another twitter user?
- OthersWithLink: Does the tweet mention another user and include a link?
- Retweet: Is the tweet a retweet? That is, does it start with RT? Note that I don't know if the API returns "auto" retweets at all (or with an RT). I know I have retweeted 140 character messages just fine (that is, no reduction in characters for the "RT ", and the system says I retweeted, but no RT appears in the tweet on my feed.
- ModTweet: Is the tweet a modified retweet? This includes MT tags, as well as RT in the middle of the tweet.
- EgoWords: Does the tweet contain the words me, my, I'm, I?
- EgoWithLink: Does the tweet contain an Ego word and a link to other content?
Here are my thoughts on this chart:
1. More than half of the 16000 tweets included a link to other material. To the extent that law professors are supposed to share knowledge with the world, this may imply that they are doing so more than half the time they tweet.
2. However, they might be sharing their own work. I don't want to debate whether this is baseless self-aggrandizement or selfless knowledge-sharing (or maybe both) here. I do want to note that about 6% of the tweets included a link and ego word. This means that the other 50% of links are either not by the professor tweeting, or they are not clearly identified as such in the tweet. This tends to run counter to an argument that professors only use twitter to shill their own work. Even if I'm off by a factor of three, it is still only 15% of all tweets, leaving 40% of tweets as links to others.
3. Only about 11.5% of the posts have ego words without a link. Many of those may be conversations. (e.g. "@user I disagree"). About 12% of the posts have ego words and mention others, but only 3% of the tweets also have a link. This means that in 9% of the tweets ego words are used while others are mentioned. These are likely conversations.
4. The references to others is large (more than 50%), but indeterminate. These may be accolades/pointers to the work of other twitter users. Or they may be conversations, because the other handle must be used in a conversation. About 20% of the posts mention others and have a link.
5. This means that about 85% of the tweets either have a link or mention others, or both, leaving 15% for pontification without reference to any particulare person. Some of these (I didn't calculate) include ego words and some do not. I also did not count hashtags.
6. The number of retweets and modified tweets is surprisingly low. It makes me wonder if retweets were accurately captured. As presented, though, it implies that law professor usage of twitter is not simply to retweet other tweets. This cuts both ways. On the one hand, it implies that twitter is not being used for self-aggrandizement. Or, if it is, then it is failing at that. On the other hand, because numbers of followers are relatively small but more diffuse than just other law professors, and because network effects of sharing are important, more retweeting would actually be better for knowledge creation, sharing, and collaboration.
These are my initial results. If folks would like another cut at the data, I'm happy to take a look. This is already a long post; as noted above, my view of future examination of this data and of twitter generally will follow in another post.