A real-life tipping point?

Last night, following Steve Jobs’ passing away, a college classmate of mine posted the following status on the Chinese Facebook: “The United States went through two things this year: 1. They lost jobs; 2. They lost Jobs.” A good one.

This post became popular immediately after its appearance online, and has been forwarded many times ever since. My friend documented the counts of forwarding this message over a period of time after the post, and with the help of STATA, she was able to draft the diffusion of her own message in a realistic figure below:

After posting this interesting real-life figure, she and I, joined by a group of fellow college friends, had an online discussion about the diffusion of this message. From the curve, there seems to be a “tipping point” around 800-1000 minutes after the posting. Everybody seems to be interested in whether there is a real “tipping point” and why it happens. But we all have doubts.

The first stream of doubts, naturally, lies in suspecting the sparse area around the “tipping point”, where few observations are reported. It is fair to question whether the data is convincing enough to reveal the underlying pattern.

The second stream of doubts, brought up by me, is about the potential time-confounding. My friend told me that she posted the message at 9:43 EDT at New York, suppose the graph tells us that the tipping time, according to the graph, happens after around 10 hours after its posting (this is a very rough estimation, though), which is the time at which our fellow friends in China (young people who understand English) begin to surf the net after an entire day of work. So I can argue that the message got spread faster from that time on simply because more people are online and have free time to read and forward messages. Alternatively speaking, the intrinsic rate of diffusion should be much faster around that time due to people’s varying exposure to social networking websites along the timeline of a day. Thus, even if we do observed a convincing pattern of “tipping point” with denser data, we still need to be very cautious about the intrinsic heterogeneity over time.

As a follow-up to the two streams of doubts, let me remind the readers to take another look at this figure – it is quite clearly visible from the dots a time period between 500 minutes to 1000 minutes, where there are not much data points collected. This is the period where U.S. users are at work in the afternoon, and Chinese users are sound asleep…This pattern is not surprising because the “data collector” (who is the author of this message) assembled the data by inviting people on this SNS to report to her “on the spot” counts of forwarding as they saw on their screen. Interesting.

Caveats aside, it was just fun to record a real-life information diffusion case!

Hat tip to the author of this message.

Post a comment or leave a trackback: Trackback URL.


  • yxysamurai  On October 7, 2011 at 2:51 am

    She posted two pictures, is this one the final one?

  • yxysamurai  On October 7, 2011 at 9:42 am

    I am still a little confused about her two graphs. This one counts around 900 within 20 hours, while the other one counts nearly 1,200 in 13 hours.

    • mumamme  On October 7, 2011 at 9:50 am

      The Excel graph has an incorrect horizontal axis. The time line is not right for that one.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: