Numerical Just So stories

Continuing from The vox populi of Google Trends.

So what do the numbers mean?

It will vary by the person viewing them, and that person's priors.

To continue the example of "Global Warming," I suspect that part of the decline in the use of the search item "Global Warming" is that people are now using the search term "Climate Change" more often (up fifty percent.) The strong claims of the early Global Warming advocates have turned out to be false, not yet proven or unprovable. In that environment, the broader claim of Climate Change is much easier to prove and is now the preferred term (Climate Change is indexed at 5.3 in December 2017 versus 4.0 for Global Warming.)

You can stare at these numbers for a long time, musing. How you interpret them depends to a great extent on your contextual knowledge.

A handful of items stood out to me.

There is clearly a power law in force here with a steep declination. The top five search items (Home, Sex, Love, Food, Family) constitute nearly 60% of this population of 97 search items. The next five (War, Security, Education, China, Disease) are nearly 17% of the searches. The top ten of ninety seven search items are 76% of all searches.

Issues that are of overwhelming interest among the chattering classes (Racism, Gay Marriage, Implicit Bias, Rape Culture, etc.) occupy vestigial interest among all Americans (as reflected in their searches.)

Biology seems to trump romance with "Sex" searched for about 35% more often than "Love." Disappointing perhaps, but useful to know.

Foodies are validated. I experience frustration with the amount of attention dedicated to food in conversations and in the media. All those food shows. But apparently my disinterest in the art of food is a minority view with food being among the top five searches.

I was surprised that certain items were not higher. Abortion at 0.5% of all searches, Terrorism at 0.1%, Social Media at 0.2% of all searches, Drug Overdose at 0.2%. This suggests to me that these issues are not broadly pertinent to most people (low percentages) but are likely of intense interest to select populations (relatively large standard deviations) in the right context. Even given that rationalization, the Drug Overdose result still surprises me. We are losing some 65,000 people a year to drug overdoses and neither Drug Overdose or Opioid Crisis are much searched. I have no explanation.

I was also surprised at how high some items were. I have considered the amount of column inches dedicated to the issue of Transgender to be way out of proportion to what people are focused on. Granted it absorbs only 0.1% of searches but that is higher than I would have expected.

It is rather like staring at the gravel in a zen garden. The longer you stare at the data, the more you see. While interesting, it's power is in suggestion possibilities rather than proving anything.