[Epistemic status: low. You tell me if you think this works.]
Commenter no_bear_so_low has been doing some great work with Google Trends recently – see for example his Internet searches increasingly favour the left over the right wing of politics or Googling habits suggest we are getting a lot more anxious.
I wanted to try some similar things, and in the process I learned that this is hard. Existing sites on how to use Google Trends for research don’t capture some of the things I learned, so I wanted to go over it here.
Suppose I want to measure the level of interest in “psychiatry” over the past few years:
Looks like interest is going down. But what if I search for “psychiatrist” instead?
Uh oh, now it looks like interest is going up. I guess what I’m really interested in is mental health more generally, what if I put in “suicide”?
Now everything else is invisible, and the data are dominated by a spike in August 2016, which as far as I can tell is related to the release of the movie “Suicide Squad”.
I could try other terms, like “depression” and “anxiety”, but no_bear’s data already tells us those two are moving in opposite directions. Also, depression has a spike in late 2008, which must be related to the stock market crash and people’s expectations of an economic depression. This doesn’t seem like a great way to figure out anything.
I wondered if averaging a bunch of things might take away some of the noise. I chose nine terms that seemed related to psychiatry in some way: psychiatry, psychiatrist, psychotherapy, mental illness, mental health, suicide, depression, antidepressants, and anxiety. Google won’t let you combine that many terms in a single query, but that’s okay – I don’t want to see them relative to one another, I just want to get standardized data on each. There’s a button to download any individual Google Trends query as a spreadsheet:
