A short venture into correlation and causation


As a budding data scientist talking about correlation and causation is a rite of passage, so I would almost be ashamed of myself if I did not have at least one, albeit short, article about correlation and causation. So as 2018 is coming to a close I decided to mine away on Google trends for 2018 to see what I was able to dig up.

The structure of the below is pretty simple, I use the compare function on Google Trends and see what turns up, and I have only compared within the United States as to take out geographical trends.

First lets start with a classic, ice cream and summer, a snippet of virtually every economist’s favorite example of correlation. This is usually portrayed as the correlation between warm weather and ice cream sales, because people eat more ice cream when it is warmer, in this case proxied by the comparison of the phrases “Ice cream” and “Summer”. The causality here is pretty obvious summer (warm weather) causes people to want ice cream increasing the amount of searches on google for it, likely in relation to buying/making it. Colder weather would very likely mean lower searches for ice cream.

Next up is correlation without causation, illustrated through the comparison of apple pie and stuffing.

The underlying causation here should be pretty clear, Thanksgiving. It is not like stuffing goes in apple pies or apple pies goes in stuffing, but both are part of a traditional Thanksgiving meal in most of the US and people will therefore tend to search for recipes etc. for both around that time. Thus, they both correlate with the underlying theme of Thanksgiving and not each other, i.e. if people would stop eating apple pie on Thanksgiving that would not cause searches for stuffing to drop.

Lastly, let’s look at two examples of what at least I found to be semi-odd correlations.

As I doubt people have a tendency to poison their friends and families at barbecues during the summer, my suspicion is that people tend to have barbecues in the summer and people are also more likely to get bitten by snakes/scratched by poison ivy in the summer, meaning that both of them essentially correlate with the season of summer. At least I highly doubt that either of them are direct causes of one another.

Slippers and hot chocolate I suspect is similar albeit for a different season. People will tend to stay in-doors in the winter and put on their slippers and drink warm chocolate, topping out around Christmas time. This will cause people to search for these two specific phrases as they will e.g. want to buy slippers.

I’d be interested to hear if you have any other/better explanations for the above, in which case please let me know at contact@theoutliar.com


Google trends (trends.google.com)