Guest Post: Giving Thanks for Data Science

Photo of former Dallas Cowboys head coach Tom Landry in his trademark gray felt hat
Not actually Aaron

Guest Post!
One of the joys of being alive is having smart, curious friends to talk with – or to write guest posts for your blog. I’d love to see more of these, especially from friends with perspectives and opinions different from my own – email me your ideas!

Today it’s awesome friend Aaron Price, who is also the Director of Research and Evaluation at the Museum of Science and Industry in Chicago.

Aaron talks about an important question in today’s age of COVID-19: are people actually following the CDC’s advice to avoid large holiday gatherings?

Because I’m a nerd who works in data for a living, I wanted to see what evidence we have for people staying home this Thanksgiving.

In a widely reported prediction, AAA forecasted a 50% drop in air travel but only a 10% drop in overall travel. But, of course, we have to remember it’s the AMERICAN AUTOMOBILE Association that is saying that AMERICANS are robustly driving AUTOMOBILES this holiday. They don’t cite the source for their forecast, other than to say it was based on mid-October “models.” Even if those unexplained “models” are reliable, they still don’t include the impact of the last six weeks of the fall COVID-19 surge – or the whole prediction could be just a bunch of marketers throwing darts at a board.

Anyway, I looked at Google Trends for the word “traffic” a few days before the 2019 Thanksgiving vs. the last few days. Days on which more people search for the word “traffic” presumably are days on which more people are driving their cars. I graphed the results on the same relative y-axis scale, to place them on the same zeropoint – the vertical axis goes from zero percent (the number of Google searches for “traffic” on the day in which the fewest people searched for “traffic”) to 100 percent (the number the day the most people searched for “traffic.”). The graph is below.

Google searches for the word “traffic,” relative to the number of searches on December 1, 2019

The graph does indeed show an almost 50% drop year-over-year.

In the same article, AAA predicts the peak travel time to be the afternoon of the day before Thanksgiving. This is supported by last year’s Google Trends search data, which shows that searches for the word “traffic” did indeed peak on the day before Thanksgiving, which in 2019 came on Wednesday, November 27th.

Relative searches for the word “traffic” in the days before Thanksgiving 2019

Indeed, looking at the traffic over the days leading up to Thanksgiving 2019 does indeed show a jump of about 50% from Tuesday to Wednesday. We could look at the same trends in 2020 – except that at the moment, Google’s data only goes to up to 1 AM ET on Wednesday November 25th. So we’ll have to wait until at least tonight to see how many people searched for “traffic” yesterday (Wednesday). I’ll post an update here later for those of you on the edge of your seats.

I also checked Google Trends for recent searches for “turkey recipe,” with the assumption that in 2020, fewer large group gatherings will mean less turkey being cooked, because people don’t want to cook a turkey for a smaller group – although this could be debated, since four families staying home instead of going to grandma’s could mean four dead turkeys instead of one). The graph below shows the number of Google searches within the U.S. for “turkey recipe” from just before Thanksgiving 2019 to just before Thanksgiving 2020.

Google searches for “turkey recipe” by day, using a relative scale from zero to 100

Indeed, Google Trends shows that turkey recipe searches are down 25% over this time last year. Interestingly, the graph below shows U.S.-based Google searches for “honeybaked ham,” again between just before Thanksgiving 2019 and just before Thanksgiving 2020, again on a relative scale.

Searches for “honeybaked ham” from Thanksgiving 2019 to Thanksgiving 2020 on a relative scale

Searches for “honeybaked ham” are up almost exactly 25% over last year! Does this mean people prefer ham with smaller gatherings? Or that, when no one is looking, we all actually prefer ham over turkey? Count me in that group.

I also looked at searches for “flight status,” and those searches are down about 75% over this time last year.

My prediction

How will these trends hold up over the next few days? And – more importantly – what impact will this week end up having on COVID-19 cases in the U.S. over the coming weeks?

My prediction: overall, automobile travel will be down about 25% over last year and flights down about 50-75% over last year. So, about a third of our country is planning to do less for Thanksgiving than they would typically do. Which would make sense given our political makeup. About 1/3 are very liberal and take COVID-19 super seriously. About 1/3 are very conservative and believe it’s all fake. And 1/3 believe it is real and are concerned enough to make small to medium changes in lifestyle, but not enough to make major changes in lifestyle such as skipping a traditional Thanksgiving gathering.

Thus, we should see a surge in cases starting around December 6th. If I’m right, it may start out at about a 30% jump over this week’s pre-Thanksgiving rates… before the increase goes exponential.