I’ve been writing here, on this blog, for a little more than eight months. What started as a way to take a break from finishing my thesis and put down some thoughts about economics has turned into a big part of how I spend my time thinking about economics. Responding to tweets and news articles and other bloggers helps me formulate my thoughts about teaching and my research, and gives me a place to keep track of papers I’m reading. I find it much more useful than EndNote, but that’s perhaps more indicative of the way my head works than anything.
One partly unintended consequence is that I’ve gained a little notoriety. The first time Modeled Behavior tweeted one of my posts, my site stats shot up and I was so confused. I thought someone had made a mistake, then became nervous that Gary Becker had read it and was going to end my career, or something. When discovered the source and tweeted them (him? it? how do we refer to a hivemind?) to say thanks, the hivemind confirmed they had similar fears that elevated site stats were a result of pissing someone off.
With that, I’ve done a very non-scientific survey of bumps. My site stats have, on average, risen over time, but there’s still a noticeable difference when one of the more established economics bloggers tweets or reblogs my stuff. Also, I don’t always know who reblogs or retweets, so if you did and I didn’t mention you, it’s nothing personal, wordpress just didn’t indicate very well to me who you are.
By way of methodology, I wanted to calculate a percentage increase in day-over-day page views, as displayed by WordPress on the day of a tweet or mention. But stats will only let me go back far enough to see three of the bumps. So, the others are cobbled together from my memory. These posts all occurred between February 10 and April 10, 2012.
The biggest bump so far came from a combined DeLong/Modeled Behavior bump. I can’t separate them out with great confidence, but given the second biggest bump came from Modeled Behavior and the magnitude of daily hits was more than twice the sole MB day, so I’m going to give it to DeLong. It’s close though, for sure. Without further ado, my list of blogging bumps, in descending order of magnitude of percentage change (or as best I remember it) in hits on day I was tweeted/reblogged/whatever.
- Brad Delong (est. 1005%)
- Modeled Behavior (est 610%)
- Justin Wolfers (304%)
- Tyler Cowen (144%)
- Marc Bellemare (est 50%)
- Brett Keller (20%)
As I see it, my analysis suffers from a few big problems:
- heterogeneity of tweets/posts might change click-through rates (did they retweet/reblog because I said something antagonistic about Gary Becker, or just mentioned them, or something else entirely? Did the retweet or reblog contain a link to this blog?).
- Serial autocorrelation (If hits are high on one day, they’re bound to be high on the next as people read through recent blog entries and tweets, and when retweets were close together, I could be attributing hits to one when they belong to another).
- Trend over time is also partially due to people coming back because they found me interesting (different, but related to 2, and impossible to know how big or small it is).
- the time of day. It’s pretty well established that tweets in the morning and mid-afternoon get the most views (or so I am told–please don’t quote me), so retweets/blogs will have differential effects given when both I and the retweeter publish the post. I don’t control for this. (also, days on this blog are on Mountain Time. Colorado, I just don’t know how to quit you. No, really, I don’t know to change it.)
- unknown retweets/reblogs
- Popularity of other blogs. (For instance, MB bump came before the Time list of top tweeters came out, so their bump may be even bigger now)
Please don’t judge me (for not controlling for obvious variables. You can judge me for writing this post; that’s fine.)