Quick, think back to 2016! What an awful year. Even if we could erase Brexit and Trump from history, it’d still be remembered as a miserable time because of all of the beloved celebrities we lost: Bowie! Prince! Carrie Fisher! And, umm, Fidel Castro.
So last year - 2017 - must have been much better in terms of the celebrity death count, right? Sure, we lost Brucie and Adam West but it didn’t feel like the previous year, when it felt as if the Grim Reaper was swinging his scythe with no concerns about who it might hit.
Surprisingly then, today Gizmodo UK can reveal that in 2017 almost as many famous faces kicked the bucket... the only difference is that we didn’t really notice. And that’s because of one crucial difference: Of all of the famous people who died… they weren’t really as famous as their colleagues who died in 2016.
Crunching The Numbers
You might remember that around this time last year, we took a look back at 2016’s celebrity deaths and concluded that - yes - it really was a bad year for celebrities dying. We’ve taken the same approach with this year’s analysis.
To do this we first had to figure how to measure what it means to be called a celebrity - and secondly we needed to find a dataset of celebrities who we can analyse. This is where Wikipedia provided an excellent baseline. In order to have a Wikipedia page, you need to be declared “notable” by the community. This is a necessarily fuzzy definition but what it does mean is that the decisions on which celebrities to include in our dataset are not being made by us, or any other single individual. Instead, a list of notable people arises from thousands of individual decisions, which seems like the fairest way to it when you have to compare politicians with film stars, and so on. So in our model, if you are notable enough to be on Wikipedia, you’re considered a celebrity.
This also solves the problem of getting a list of celebrities - brilliantly, Wikipedia curates a list of notable deaths, instantly giving us a dataset to work with. We’ve gone back to 2010 - as if we included years much further prior, there would be a risk that lists would be incomplete, as they would have been filled in retroactively rather than in real time with the news.
Obviously this doesn’t make it completely foolproof: Wikipedia has its own biases, such as being biased towards the English-speaking world - but in this case at least, we think that works to our model’s advantage, as this probably better reflects our English-speaking expectations of who celebrities are.
So with this in mind, according to our data 6713 notable people died in 2017 - compared with 6737 in 2016. Though in fairness, 2016 was a leap year so there’s an extra day of data. If we only count the first 365 days of both years - that means we’re comparing 6713 deaths in 2017 to 6726 deaths in 2016. In other words, just 0.2% fewer famous faces bit the dust in 2017 as in 2016.
Here are the cumulative deaths for each year plotted on a chart:
Why Nobody Noticed
So… why didn’t we notice? Why didn’t we all tweet “God damnit, 2017!” every time we got a push notification announcing another beloved actor, sportsperson or politician had been lost, like we had the previous year?
The difference can be explained in terms of cultural impact. But how can we measure this? Is is really possible to compare, say, the relative fame of Roger Moore and Chris Cornell?
Using the Wikipedia data and borrowing a trick from Google’s page ranking system we’ve found a way: One of the ways that Google chooses to order search results is by counting how many other websites are linking to to a given page: If one relevant page has 10 other sites linking to it, and another has 500, Google assumes that the one linked to by 500 other places has a better answer so lists it higher up.
We can use this same logic by measuring how many other Wikipedia pages link to the dead celeb’s page as a proxy for this impact: For example, Roger Moore’s wiki page is linked to on 1239 other wiki pages - and Chris Cornell’s is linked to on 957. So by this measure, we can conclude that Moore was the bigger name, and therefore probably had the bigger cultural impact. Let’s call this the notability score.
This isn’t completely without its flaws: Wikipedia will once again have built in biases - for example, sports people could conceivably be overrated for example as there are lots of pages of statistics. But hey, when you’re comparing apples with oranges, this is the best metric we have.
So what happens when 2016 and 2017 are compared using notability? This is where we start to see a difference. Here’s a chart showing what happens as you go through each year (starting at day 1, then day 2, etc) and add up the notoriety scores of everyone who dies:
Note how 2016 still has - by a bit of a distance - the highest cumulative notability score. Crucially, imagine if this graph didn't contain 2017 (the lime green line) - the jump from 2015's deaths to 2016's must have felt pretty massive. By contrast, 2017's deaths are somewhere in between - so would feel less remarkable coming hot on the heels of 2016.
Seeing the notability totals for each year also really makes this point:
The 5 biggest celebrities to die in 2017 according to their notability score were:
- Chuck Berry (2173) (Musician)
- Glen Campbell (1706) (Singer)
- Jim Bunning (1365) (Baseball Player)
- Kenneth Arrow (1204) (Economist)
- Jana Novotna (1123) (Former Wimbledon Champion)
What’s notable about them is perhaps how relatively obscure they are - at least compared to 2016’s crop of dead celebrities. Bowie was the biggest name to die, with a score of 5728, then Prince (3742), Fidel Castro (2939 - yes he counts!), George Michael (2308), and Muhammed Ali (2251). All of these 2016 deaths scored more than Chuck Berry - 2017's highest entrant.
So though roughly the same number of people died, the ones who did die in 2016 were way more famous.
(Incidentally by using the notability score it explains why this piece is seemingly all about dead men. Skimming through the list there are not many women - especially women who this author has heard of. Perhaps all of the old celebrities from earlier eras dying off shows just how even more male-dominated the world was in the past?).
This 2016/17 difference is even more pronounced if you filter the data to only include celebrities who have a notability score of over 500 - to cut out anyone who might only be borderline famous.
As you can see above - here the sheer abnormality of 2016 with its litany of high profile deaths is even more notable. 2017, by comparison, was just another fairly normal year - more or less the same number of famous people died, but in aggregate, all made less of a cultural impact. So that’s why we didn’t actually notice.
James O’Malley is Interim Editor of Gizmodo UK and tweets as @Psythor.
Update (15th Jan): We've updated our figures after it was pointed out that our initial analysis included non-article Wikipedia pages factoring into the notability score (eg. talk pages and user profiles). Brilliantly, having re-done the calculations, it hasn't undermined our conclusions - which are still the same. Hurrah!