A simple Google Trends search predicted the outcome of the past 4 elections

[Graham Osborn]

It's funny how given all the polls and meta-polls conducted prior to the election, noone ran a simple Google Trends search of "Trump" vs "Clinton." The outcome was clear for months in advance (see attachment 1).

 

This pattern also held for "Obama" vs "Romney" in 2012 (see attachment 2) and for "Obama" vs "McCain" in 2008 (attachment 3) and "Bush" vs "Kerry" in 2004 (attachment 4).  The last comparison, I concede, could be confounded by other meanings of the word "bush."

[eclecticvalue]

Well why didn't I think of that. However, you still need to label your lines so we know who's who.

[Travis Wiedower]

Not convinced this means much. Four is a tiny sample size and it's easy to retroactively find indicators like this that look nice in a rear view mirror.

[DooDiligence]

I may have skewed the bush results a bit...

[Graham Osborn]

Not convinced this means much. Four is a tiny sample size and it's easy to retroactively find indicators like this that look nice in a rear view mirror.

 

Not if you only test 1 indicator (as opposed to data mining).  In this case the probability of the indicator agreeing with the results by chance is about 6%.  On the face of it one would think that this should be more correlated with the popular than the electoral vote, but it's actually more nuanced than that if you start thinking about the many drivers behind search selection and search intensity.  While electoral models are more complex/ nuanced they also tend to be riddled with false assumptions.  For fun, we'll give it another try on some other upcoming international elections.

[rkbabang]

This is interesting, but I'm glad that I didn't know the results of the election in advance.  I thought the market would crash if Trump won, so had I known he was going to win I might have sold off a lot of my portfolio before the election.  You don't only need to know the outcome of the election, but you also need to know what it means for the markets.

 

 

[Graham Osborn]

This is interesting, but I'm glad that I didn't know the results of the election in advance.  I thought the market would crash if Trump won, so had I known he was going to win I might have sold off a lot of my portfolio before the election.  You don't only need to know the outcome of the election, but you also need to know what it means for the markets.

 

Aye, therein lies the rub.  Even if you could/ can predict the outcome of elections, I'm not sure what the relevance would be to market participants.  By reflexivity, the only absolute truths are unactionable (at least from a profitability standpoint).  This is one reason I think there might be predictive value to SOME test, even if not this one.  Clearly, using such a test as a basis for trading would have been poorly advised if you simultaneously subscribed to the second incorrect prognostication, i.e. that Trump would crack the indices.