How did the Instagram likes for Cincinnati Zoo change following Harambe? (using R and ggplot2)

Recently, I was given an interesting project to do with levels of Instagram engagement that a conservation organization has. While playing around with various graphical depictions, I started thinking, of all things, about Harambe. Poor, poor Harambe! That in turn led me to wonder how the Cincinnati Zoo (CZ) fared in its Instagram engagement (specifically, number of likes on Insta posts) following the tragedy. As you all must remember, the controversy generated from that incident was vicious. Whether you believe the zoo was at fault or not, I think we are all in agreement that it was a huge PR scandal for the zoo. So, my question is: Did they suffer because of it? Did people boycott the social media of the zoo, specifically their Instagram?

I decided to take a look at the Instagram of the CZ and make up a visualisation of the weeks prior (starting April 1^st, 2016), and the few months following Harambe’s death (July 30^th, 2016). FYI – for those who don’t have it seared in their memory, Harambe died on May 28^th, 2016. Because I was going through the Instagram feed by hand, I saved myself some time and only sampled every other day of posts (n = 55). As this data is publicly available, I’ve gone ahead and made the dataset created from it publicly available as well (you can find it here).

Here’s the code for the first visualization:

And here is the first visualization:

You’ll note I’ve included a “rough and dirty” smoothing line. The v-line intersecting the graph indicates the death of Harambe on May 28^th. The first interesting thing to note is that the likes actually appear to increase following Harambe’s death! In fact, if you take a look at the post CZ did memorialising Harambe’s death, it had a record 9,802 likes. (I didn’t include that post in my analysis because it was actually on my “skip” day, which is probably for the best as it would have been a HUGE outlier.)

Here are a couple of graphs highlighting the difference between the two groups (pre and post Harambe [the code for the circles is at the end of this post]):

If you’re a number cruncher and want to know the averages, the average number of likes in my sample, prior to Harambe, was 2,297 likes per post. The average number, post Harambe, was 3,380. A simple t-test performed between the two groups, pre and post, found p = 0.00000671. Obviously highly significant! (This is technically bad statistics- the data are not normal and as such I should really use a Mann-Whitney/Wilcoxon test. However, such a significant p-value tells me that a rank-sum test wouldn’t have a result that’s terribly different.)

I certainly don’t want the conclusion of this post to be that PR scandals help zoos out (at least in their social engagement). This is obviously situational, and other factors are at play. One factor I think may have had an effect is that CZ constructed a new hippo exhibit in the weeks following Harambe’s death, and those hippos are hugely popular (e.g. Baby Fiona [who has an article dedicated to her on People.com]). In fact, CZ’s social reach has, as of March 6th, 2018, quadrupled from where it was pre-Harambe; an average post now has 8,000+ likes!

I do think that Harambe’s death probably did have an indirect effect in that people noticed the zoo, and checked it out, out of morbid curiosity. Then, maybe they stayed liking and following the zoo on Instagram because they realized that CZ is a pretty normal, well-run zoo, that just happened to have a major, bizarre tragedy happen to it.

Leave a comment if you have aesthetic thoughts on the graphs, thoughts on the rough and dirty stats, thoughts on the R code, thoughts on Harambe or if you just have thoughts in general (I really love spongebob gifs, so comments in spongebob gif form will be appreciated).

All of the code can be found here.

* Update: Follow-up post here.