Ontario election: What is Google Trends showing?

Using searches on Google can be useful during an election. There seems to be at the very least a correlation between the quantity of searches and votes. With that said, it's at best iffy and not something I'd use over polls. It can also be super finicky. Last year during the mayoral election in Montreal, I discovered that I'd get different results if I use the accent on the e of Valérie Plante (seriously, if I was looking at Google Trends for "Valérie Plante", she was losing, but she was winning under "Valerie Plante"...). Still, it can be interesting to look at this.

First, let's look at which leader is most searched. All three leaders (Wynne, Ford and Horwath) are famous enough that Google knows about them. It means we can do a (superior) "topic" search instead of a term search (the latter means we'd simply be looking literally at the words "Kathleen Wynne". It's inferior because it can get polluted by people searching for another Kathleen. It also doesn't capture related searches like "Premier of Ontario". At least that's my understanding). Results are below for Ontario during the last 90 days.


Note: I don't choose the colors! Also, these graph are live. So the text I'm writing here is valid as of May 23rd!

Doug Ford is well ahead. That can be a good thing or not. For instance the main search associated with Doug Ford is "Doug Ford greenbelt" related to the infamous "secret" discussions that Ford had with developers and promising them to open the greenbelt to them.

As for Kathleen Wynne, things don't improve as the main search is "how old is Kathleen Wynne"!

What to take from this graph? Mostly two things. 1. Ford is by far the most searched leader. This is impressive since usually incumbents have an edge. The divisive Ford creates controversies and interest. 2. Horwath is still third. One of her issues is that she is simply less known. In the polls, she is the leader with the most "neutral" or even "don't know" answers when asked if they have a positive or negative opinion of her. You'd think that would have changed with the campaign and the polls showing her in (possibly) a winning position. But this hasn't been the case (yet maybe?). It also didn't really change for John Horgan last year in BC. At least the searches increased for Horwath right as the campaign started.

If instead we look at searches for the parties, the NDP does much better. This is similar to what was happening last year in BC with Christy Clark crushing it with her popularity but the NDP doing much better with the party.


Interestingly here, it seems we can pick up the recent decline of the Conservatives and the rise of the NDP. It'll be interesting to monitor this for the next two weeks. The rise is even more pronounced if we use the last 7 days:


So, any useful information for predictions? I think yes, if you aren't too picky. These two graph to me confirm the Liberals are not in the race. They are the incumbent and main (historical) party and are quite low in Google searches. That isn't normal or a good sign for them.

After, does it show the NDP or PC ahead? I'm tempted to do an average of the two measures as a vote is simultaneously influenced by the leader and the party. If you do that and normalize to 100, you get the following "estimates" for the last 30 days:

OLP: 24%
PC: 50%
NDP: 26%

If you only use the very last 7 days:

OLP: 22%
PC: 45%
NDP: 33%

Not bad! Keeping in mind the possible overestimation of the PC due to the "celebrity status" of Ford (or putting more weight on the party searches instead than the leader ones), you get a pretty good estimate of the polling average.