An interesting error

I'd last week wondered whether long term decline in celebrity search interest predicts release of private photographs.

Turns out, my results were wrong. I'd provided search traffic data on "Scarlett Johannson", not on "Scarlett Johansson". The former, a typo, had long term decline, followed by a surge in interest with the release of artistically-shot nude photos from her cell phone; the latter has a noisy but flat long term curve with, again, a massive spike in search interest with the photo release.

In my defence, I'm not the only one that got the name wrong. This one's funny as the misspelled name in the Washington Post headline is matched with a cover of GQ (as part of the WP story) where the name is spelled right.

The graphs are up as an update to the original post.

It's interesting that the version with the typo has long term decline while the correct spelling has noisy but stable interest: the proportion of people interested in searching for Scarlett is increasingly made up of folks who know how to spell her name. But, with the photo release, we get a spike in all searches on the name, including those spelling it incorrectly. Unfortunately, you can't compare search volumes unless you put them on the same graph; when you put them on the same graph, volume on the typo normalizes to zero for the period prior to the photo release, after which it increases to 1 (searches on the properly spelled name normalize to 100 with the photo release). It would be mildly interesting to see if a greater proportion of searches post-photo include the typo.