Drawing conclusions from proxy data during extraordinary events

Note to the reader: this post refers to traffic data published by an adult video website. I do not link to that site directly, but pages linked in this post link to said website. I try to keep the content of this post safe for work, but use your own judgement.

How do people alter their behavior when they think they’re minutes from death? We got a chance to see just over a week ago when officials in Hawaii made a mistake. A very big mistake. They accidentally sent out a warning that a ballistic missile was inbound. For 38 minutes, people in Hawaii thought they were about to die in a nuclear attack.

An adult website that rhymes with “Corn Bub” found that their traffic dropped off dramatically during the time between the initial alert and the announcement that it was a false alarm. They compared traffic from Hawaii on that Saturday with the preceding two Saturdays and found at one point it was 77% below normal. After the all clear, page views spiked to 48% above normal.

One blogger used this to conclude that “nobody [engages in self-gratification] while waiting to die in a nuclear fire.” I cannot accept that conclusion without more supporting evidence. Under normal circumstances, I’d buy traffic to a mainstream adult website as a proxy for self-gratification. Not everyone uses such material, and those who do don’t all use the same website, but it’s probably a relatively stable representation of patterns. But the imminent threat of a fiery death is a (thankfully) unusual event that could drastically alter behavior.

Popular culture is rife with jokes and “sure, that totally happened’ tales of sexual activity when death looms (e.g. a crashing airliner). It’s entirely reasonable to believe that people figured “well, if I’m going to die, I might as well take care of this first” and immediately got down to business. Given the time constraints and the general stress of the situation, I would not be surprised if they didn’t partake in online content as they normally would.

Now I’m not saying that people did or did not engage in self-gratification during that time. I am saying using a particular website’s traffic as a proxy for that behavior is questionable. Even if it is reliable in normal circumstances. This is a particularly awkward example, but it’s a good lesson. If you’re not directly measuring something, don’t assume your proxies are valid in unusual situations.

Leave a Reply

Your email address will not be published. Required fields are marked *