Livejournal population statistics, round 2.
It seems like every time LJ does something stupid I get inspired to write another population post. I tried tackling LJ's gender balance back in November in this post in lj_research, but the raw data makes no sense and I sort of gave up. Oh well.
Today I'm going to update the population post I made in August of last year. I'm only going to show data from Jan 1 2005 to March 15 2008. This time period includes the entire length of 6A's ownership of LiveJournal.
My methodology this time is roughly the same as that of before. I used the stats.bml and stats.txt files archived on the Internet Archive. Since the Internet Archive's archive of those pages has grown sparse lately, and has no data after May 28, 2007, I supplemented it with numbers recorded in this post in no_lj_ads and my own downloads of files. The raw data is available here if you like. (Yes, I'm using Excel now to make the graphs. Excel sucks, but it sucks less than the program I was using.)
It may be easier for you to open the image gallery containing the figures here in another tab or window to follow along.
Let's first look at the rate of account creation. I think some phenomenon here explain what comes later.
Figure 1: New accounts per day since Jan 2005
Woah! There's a big spike on Jan 28, 2008, when 41463 new accounts were created. The next day had 11779 new accounts created. For comparison, the largest spike in LJ's history before that was 17630, shortly after the invite-only period ended. I believe this spike falls under the dictionary definition of ginormous. What caused it? I don't know, but I'll theorize about it later.
Let's ignore that big spike for a moment.
Figure 2: New acounts per day since Jan 2005, huge spike off the scale.
What we see is a general trend of decline, which seems to flatten out around May of 2007 and stay between 8000 and 6000. There's a drop in account creation right after SUP took over, but if you look at past years, the account creation and posting rate usually slumps around time of year due to the winter holidays. It's too early to say if there's a meaningful trend in the data yet.
Now to the accounts numbers! Everything's defined here the same way it was before.
Figure 3: Accounts numbers.
There's a little upward tick in the Total Users where the big spike in account creation was. This isn't surprising, but it corresponds to an increase in the activity and updated numbers. Accounts Ever Updated does not show this tick, but that may be because of there wasn't data taken often enough near that date to show it.
Aside from that, the usual story: Total users and accounts ever updated go up, everything else trends down. However, after the SUP deal there seems to be some stabilization. This may be a result or an artifact of the account creation spike.
Figure 4: Activity and updated numbers.
Even at this detail level, it's hard to tell if the trend has changed to flat or not. We're going to need to wait a while before we can tell what the trend in LJ's activity rates in the future will be. I hope that whether it does or not will depend on how SUP treats LJ, but I have a sinking feeling that it won't.
A number of people have posited the theory that while there are now less LJ users posting regularly, the ones who are are posting more. If this was true, the 24 hour updating number would increase against the 7 day and 30 day updating number. I don't see evidence of that that from the graph. There's also the theory that people are commenting more and posting less, which may be more likely, since the 30 day updating number has decreased faster than the 30 day activity number.
Final question: what caused the spike? I dunno, but I have a theory. GreatestJournal started dying around Jan 13; i'm not sure when it finally started disallowing new account creation, but it was probably some time after that. The people leaving GJ probably either went to InsaneJournal or back to LJ. GJ had something like 45000 users updating in the past 30 days back in August and has only 11133 updating in the past 30 days now, so considering that GJ's peak was probably later than August, it's possible that combined with normal account creation that could do it. I can't get any hard numbers on GJ's population at the moment because they blocked the Internet Archive from archiving http://www.greatestjournal.com/stats.bml. C'mon, you guys, cooperate. I'm not sure if the dates all line up well enough for this to be true.
Next time: Either age, or country origin. Hopefully the data for that will make more sense than the gender data. Thanks again to brad for leaving the code open, everyone in no_lj_ads who collected data, and Viewers Like You.