It’s Gavin Newsom versus the out-of-nowhere political star that is John Cox, or “John H. Cox American Accountant” as he’s referred to on Google Trends. And that leads to the raison d’etre for this post: to determine if Google Trends got it wrong, or if this blogger, me, got Google Trends wrong, or something in between.
I want to thank a viewer for pointing out, in the comments to another related video, that if I tried just “John Cox” as a search term, rather that the “John H. Cox American Accountant” choice that I selected on Google News, then a different result would appear. But before I continue, let me catch you up on what I am writing about.
In brief, I was intrigued with the idea that Google Trends (which attempts to show what search terms a population is looking for using the Google Search Engine at any given time) could be a predictor of elections other than that for POTUS. When I ran my initial tests of this theory, I ran Google Trends using the top candidates (not all of the 33, as Google Trends can only fit five at a time, and the others listed failed to move the search needle to even be worth noting) and with “John H. Cox American Accountant”, the Google Trends feedback had “John H. Cox American Accountant” with low search intensity.
But then the viewer, who goes by a YouTube name that’s not his or her real name, and is xxTheGhilliedGhostxx chimed in with “I noticed your results show “John H. Cox.” If you put simply “John Cox” into Google trends, his numbers are much higher. I just compared the top 4 candidates on Google Trends, and as of right now, the results are: Gavin Newsom:100 John Cox: 35 Travis Allen: 31 Antonio Villaraigosa: 26 I predict it will be Newsom vs Cox in November.”
So I tried his suggestion, but used john cox, and for a California-focused Google Trends search test, and received the same low result as for “John H. Cox American Accountant”. Then, it occurred to me to replicate what the viewer did exactly as it was done. I didn’t have a screen shot of the YouTube’s result, so I had to guess that I should capitalize the J and the C in John Cox. That made the difference, and the result is below.
With that chart, we see a result more in line with what the UC Berkeley Institurte of Governmental Studies department learned after their poll published May 31st: that Cox was competitive with Former LA Mayor Antonio Villaraigosa, Travis Allen, and over John Chiang. But that said, Cox never had a giant bump that was in Newsom territory. Moreover, the chart, for some reason, only goes to Monday, June 4th, but even then, it’s a pure assumption to think Cox’s 34 intensity (which still put him below Allen and Villaraigosa but by five points and two points respectively) would have jumped to say “50”. It brings the question, how could Cox emerge to almost beat Newsom in the real election, but not have that reflected in Google Trends.
One possible answer has to do with the low voter turnout. Google Trends captures what people search for, but for politics in local elections is not the perfect predictor of what will happen in an off-year local contest. That said, the overall fact is, Google Trends got it mosly right – but not turnout. And while the Berkeley-IGS poll didn’t predict voter turnout directly, it did indirectly via the result. Berkeley-IGS calling mostly traditional landlines rather than cell phones told us that older voters mostly likely to get to the polls, were going to make Cox the second place choice – and that’s what happened. But even then, it did not capture the pure enthusiam for Cox.
So, while Google Trends didn’t get it exacly right, it did reflect the degree of interest in the candidates and it predicted Newsom’s win. Google Trends can’t replace a good Berkeley-IGS poll, but should be used in conjunction with it: Google Trends shows degree of interest in a candidate, the Berkeley-IGS poll methodology shows the degree that turnout, low or high, will effect an election.