Comparing Instagram Likes Between Cincinnati Zoo, San Diego Zoo and Wildlife Reserves Singapore Before and After Harambe (R and ggplot2)

The fantastic Diogo Veríssimo brought to my attention that my previous post didn't have a counterfactual element to it. He is absolutely right that it was bad science in that respect. Not including a controlled comparison point to the Harambe incident could lead to some faulty conclusions.

To that end, I updated my initial analyses with data from the Instagrams for the San Diego Zoo and Wildlife Reserves Singapore. Why did I choose those particular two? San Diego Zoo I chose because I'm biased towards it, and also because it is another US zoo, and thus probably has a similar online (and perhaps physical) audience to the Cincinnati Zoo. I then chose to compare with Singapore Zoo (and affiliates) because they, in theory, wouldn't have been affected by the Harambe scandal.

Here's the script for the graphs:

	#Compared to SDZ
	ggplot(zoo, aes(x=Date2, y=Likes, colour = factor(Zoo))) +
	geom_point() +
	#geom_line() +
	geom_smooth(method="loess", se=F) +
	#xlim(c(0, 0.1)) +
	ylim(c(0, 20000)) +
	labs(subtitle="(taking note of the death of Harambe)",
	y="Number of Likes",
	x="Month",
	title="Number of Instagram likes over time (April 1st, 2016 - July 30th, 2016)",
	caption = "Source: Cincinnati and San Diego Zoo, and Wildlife Reserve Singapore Instagrams")+
	theme_light() +
	scale_color_manual(values=c("#E69F00", "#56B4E9", "maroon")) +
	theme(text = element_text(size = 20), axis.text.x = element_text(angle=65, vjust=0.6)) +
	geom_vline(data=zoo,aes(xintercept=as.numeric(as.Date("2016-05-28"))), color = "skyblue")
	#Looking at just Cincinnati and Singapore
	newdata <- subset(zoo, Zoo == 'Cincinnati'| Zoo == 'WRSingapore')
	ggplot(newdata, aes(x=Date2, y=Likes, colour = factor(Zoo))) +
	geom_point() +
	#geom_line() +
	geom_smooth(method="loess", se=F) +
	#xlim(c(0, 0.1)) +
	ylim(c(0, 6000)) +
	labs(subtitle="(taking note of the death of Harambe)",
	y="Number of Likes",
	x="Month",
	title="Number of Instagram likes over time (April 1st, 2016 - July 30th, 2016)",
	caption = "Source: Cincinnati Zoo and Wildlife Reserve Singapore Instagrams")+
	theme_light() +
	scale_color_manual(values=c("#56B4E9", "maroon")) +
	theme(text = element_text(size = 20), axis.text.x = element_text(angle=65, vjust=0.6)) +
	geom_vline(data=zoo,aes(xintercept=as.numeric(as.Date("2016-05-28"))), color = "skyblue")

view raw CincinnatiZooComparedtoSDZooandSingapore hosted with ❤ by GitHub

And here is the first graph, comparing all three:

There are a few things to notice about this graph. The first is that poor Singapore is pretty minor compared to the other two zoos in its level of engagement. The second is that the San Diego Zoo is a little bit more chaotic in its engagement compared to Singapore and Cincinnati, at least when looking at the superficially drawn local regression line. But what's interesting is that the San Diego Zoo's level of engagement also appeared to go up following Harambe.

So, the questions are: a) did Harambe have an effect on other zoos? If so, b) was the effect only seen in US zoos?

A bit of statistics helps, I think, to answer this question. You'll remember from my previous post that before and after Harambe the amount of likes Cincinnati Zoo got on Instagram did change significantly (you'll also remember I noted it's hard to tell whether this is because rubber-neckers stayed around, or whether people got more engaged because Cincinnati built their very popular hippo exhibit). Looking at the before and after for San Diego Zoo (SDZ) and Wildlife Reserves Singapore (WRS), I got the following p-values: 0.282 and 0.000000002!!

WRS's Instagram account got significantly more likes before and after Harambe. Now, they didn't mention Harambe at all in their posts. But did Harambe have an indirect effect on a zoo thousands of miles away? I'm not sure this is a question that can adequately be answered, but my instinct is that this is one of those strange instances where a completely unrelated event seems unconnected, like one of Tyler Vigen's "spurious correlations".

Back to Harambe. Did he (or the hippos) affect the Cincinnati Zoo's Instagram engagement after all, if we compare fellow American zoo the San Diego Zoo? The p-values seem to say yes, but as we all know, a lot of factors contribute to things like this.

*Disclaimer to those who check the data out: some of the dates don't quite match up because, unsurprisingly, the three zoos I looked at didn't coordinate their Instagram posting!