Analyzing a resolution to read more

Lots of little things happen in grad school that take you by surprise but really shouldn’t. One of those things for me happened after I finished all of my coursework. Without the deadlines of weekly seminars, I didn’t read journal articles or books nearly as often as I should have. I figured this out about this time last year and made it my New Year’s resolution to get better at keeping up with journals.

When I started working on this, one thing that became immediately apparent was that I lacked any sort of organized system for getting notified about new articles. Which seems like something I should have done year 1 and is rather obvious in retrospect. But hey, I was a young grad student.

I tested out two ways of getting notifications. One was with email alerts. All the publishers are more than happy to email you about new issues/articles. This worked, but it was annoying to get all these emails in my inbox. So I then switched to using RSS.1 That’s my current setup.2 All of the major publishers publish RSS or Atom feeds for their journals, so it’s easy to get subscribed. Well, all of them except Cambridge. Luckily, there is a webapp that lets you add RSS posts via email. So I signed up for Cambridge’s email alerts but gave them an email address that that company gave me.

Now that I get notifications about new articles/issues, it certainly feels like I read more. But having some empirical evidence would be nice.

So I analyzed my .bib file in R.

You can import a bib file with the “bibtex” package. I then did a little wrangling to get a data.frame with one row per citation. The code is available on github, and I’ve embedded it at the bottom of this post.

When taking classes I did an OK job at keeping track of what I read, but a lot of stuff I read for coursework will be missing. Also missing are things that I read but didn’t put in my bib file (yet), like articles I reviewed or working papers that colleagues had me read. So the data actually undercount what I read. Since we’re relying on citation information, we’ll also be plotting the year the article was written, not the year I read it.

First, I looked at a plot of all the years of the articles/books I had. 2016 is the rightmost bar.

Count by Year

You can see that my suspicion was probably right - the lines for 2014 and 2015 are pretty short. 2015 was the last year I took classes, and 2014 was a methods-heavy year so not much reading there.

It looks like I did a pretty decent job with my new years resolution. 2016 is pretty clearly the tallest bar (and it’s not over yet! In fact, I still have 23 articles to read on my todo list).

Of course, now that we have the data in R, we can make all kinds of interesting plots. I know that I read a lot of AJPS/APSR/JOP (the top three journals in political science) articles when I was taking coursework, so I was interested to see the distribution of the things I’ve read by journal (only plotting journals where I have more than 5 articles)

Suspicion confirmed again. We read a lot from the same three journals in coursework. I think that since I’m reading from many more journals now that this discrepancy will lessen over time though.


  1. And yes, of course you can read RSS in Emacs. gnus is built-in and there’s also elfeed, which is what I use. ^
  2. If you want a list of all the RSS subscriptions I currently use, including the journals, you can find that on github. ^