The 27 Club
Mark Honeychurch - 10th October 2022
In preparation for our next podcast episode, where we plan to talk about Bronwyn’s articles on Dr Bronner, I watched a documentary the other day called Dr Bronner’s Magic Soapbox. The documentary used footage from another, much older documentary called Rainbow Bridge - a weird piece of experimental film about a hippie community in Hawaii. The end of the documentary contains footage of a live gig Jimi Hendrix performed in Hawaii soon before his death.
It was obvious from another part of the documentary that Jimi, shown with a bottle of spirits in hand and talking with other “visionaries” such as an ex-wizard, was off his face. In fact everyone in the film was talking nonsense about mysterious planets, aliens, levels of consciousness and other silliness. Seeing Jimi in this way led me to his Wikipedia page, where I wanted to learn more about his death at a young age. And I wasn’t overly surprised when I got there to read about his issues with substance abuse, including a long-standing problem with alcohol. One thing that stood out for me, though, was a section of the Wikipedia page that mentioned that Jimi is a “member” of the 27 Club, a group of musicians who have died at the age of 27.
The idea goes that there is a spike of musician deaths at age 27, including big names such as Janis Joplin, Jim Morrison, Kurt Cobain and, more recently, Amy Winehouse. The Wikipedia page for the 27 Club even had a quote from a famous music biographer, Charles Cross, who asserted:
“The number of musicians who died at 27 is truly remarkable by any standard. [Although] humans die regularly at all ages, there is a statistical spike for musicians who die at 27.”
Thankfully there’s a more skeptical section of the page lower down that’s titled Scientific studies, and it goes on to say:
A study by university academics published in the British Medical Journal in December 2011 concluded that there was no increase in the risk of death for musicians at the age of 27, stating that there were equally small increases at ages 25 and 32
However, this section also mentioned a possible limitation of this study:
The selection criteria for the musicians included in the study, based on having scored a UK No. 1 album between 1956 and 2007, excluded several notable members of the 27 Club, including Hendrix, Joplin, Morrison, Pete Ham, and Ron “Pigpen” McKernan.
I figured it probably wouldn’t be too hard to see what data is out there to further support these claims. So off I went, trying to see if I could prove or disprove the idea that musicians tend to die more often at age 27 than other ages. I figured that rather than trying to decide what constitutes a musician worthy of being on an exclusive list, I would just include all musicians I could find where both birth and death dates are available.
For projects like this I’m a big fan of structured data, where pieces of digital information are properly categorised - and I rue the fact that Wikipedia does not make effective use of structured data in its pages. Thankfully there’s a sister project, WikiData, which does store encyclopaedic data in a structured format, with semantic markup. This is extremely useful, as data like this can be easily processed by computer software.
I looked around for the best way to query WikiData, and found that there’s an online service called the WikiData Query Service where you can write queries in a web page and have the page output the data in a handy format like JSON or CSV (a type of simple spreadsheet). The service uses a query language called SparQL (pronounced sparkle), which is similar to the popular SQL database query language but made for graph databases. (Data that is stored in a graph database is optimised for queries related to the connections between those pieces of data, which is useful for items where there are lots of threads joining them. I’ve used graph databases before in projects like a Six Degrees of Kevin Bacon calculator.)
While poking around the code examples provided by the WikiData Query Service, I found one that was almost perfect for what I needed - titled “‘27 club’ - musicians who died at age 27”. The query listed all “musicians” (a category of “occupation”) who died at the age of 27. Thankfully it didn’t take long to change the code from this existing example to request musicians who died at all ages, and then sum the number of them who died at each age (in years). My resulting query looked like this:
_SELECT ?age (COUNT(?age) AS ?count) where {{
SELECT (xsd:integer(?age_) as ?age) where {{
SELECT DISTINCT ?person ?age_
where {
?person wdt:P31 wd:Q5;
wdt:P106/wdt:P279* wd:Q639669;
p:P569 ?dob_st;
p:P570 ?dod_st.
?dob_st psv:P569 [ wikibase:timeValue ?dob; wikibase:timePrecision ?dob_prec ];
a wikibase:BestRank.
?dod_st psv:P570 [ wikibase:timeValue ?dod; wikibase:timePrecision ?dod_prec ];
a wikibase:BestRank.
FILTER(?dob_prec >= 11 && ?dod_prec >= 11)
bind((?dod - ?dob) / 365.2564 as ?age_)
}
}
SERVICE wikibase:label { bd:serviceParam wikibase:language “[AUTO_LANGUAGE],en”. }
}}}
GROUP BY ?age ORDER BY ?age_
Don’t worry if this looks like nonsense to you, some of it does to me too! Thankfully I could make enough sense of it to add the part at the very top and bottom, that tells the query to summarise (GROUP) the data by age (?age), and I removed a line that limited the results to just people who died at age 27 (that line was “filter(?age >= 27.0 && ?age_ < 28.0)”_). If you’re interested in seeing the results yourself, you can load my modified query into the service using this handy link, and click the play button on the left to run it.
One thing that was interesting straight away was that a couple of negative ages showed up! I’m finding it hard to believe that any musician died at the age of -25 or -21. Similarly there were 8 musicians who died at age 0, and 1 each at ages 1, 2, 5, 6 and 8. I can only assume that these are a result of bad data, and as a good citizen I will probably figure out which musicians these are and fix the data when I have time (it’s always good to give back to projects like WikiData where possible).
The rest of the data looked good, though, with no obvious issues, and so I exported it all as a CSV file and saved it to my desktop (I couldn’t be bothered filtering out the anomalous ages). Here’s a sample of some of the rows in the spreadsheet, from ages 20 to 30:
Age Count
20 48
21 67
22 74
23 86
24 116
25 124
26 141
27 173
28 206
29 229
30 250
A quick glance at those numbers suggests that, at least when we consider all musicians recorded on WikiData (which is likely to be a fairly comprehensive data set), there doesn’t seem to be a spike at age 27. Still, to be sure, and to see if there are any other anomalies, I figured plotting the full set of numbers on a graph would help. So I searched online for a service that would graph the data for me, if I just uploaded the file I had. A quick google search led me to csvplot.com, which was simple and did exactly what I needed:
And there we have it. I guess it’s pretty much what we’d expect to see if there’s no special age where artists die more often. There’s a little noise in the line, with the trend starting to go up slowly from around age 15 or 16 (around the earliest you’d expect someone to be considered a musician). The line increases to a peak, maybe something we could even call a plateau, for the entire decade of musicians in their 70s, which sounds about right for the time that a natural death of old age would occur. The graph then declines again, as fewer and fewer people manage to live a long life. (I’m guessing there might end up being a tiny spike at 120 when Keith Richards finally dies!).
A quick search on google images finds similar graphs for mortality of the general population (like the one below for women, found here), but the curve for musicians appears to be a little steeper early on (maybe a product of musicians’ tendency to chase a lifestyle of sex, drugs and rock n’ roll?), and the musician graph doesn’t have the same high infant mortality spike (for obvious reasons).
But, the important thing here is that, for this data set at least, we see absolutely no spike at the age of 27. It’s not a thing, at least when we consider all musicians on WikiData. Maybe I could refine the list to only include musicians who have had hit singles, released best selling albums, played at Wembley, made over a million dollars, or been on Top of the Pops. But for now I’m happy that my 30 minute foray into the data shows that this idea of a 27 club appears to be nothing more than a myth, one that will likely be reinforced whenever a musician dies at the age of 27 - when every media outlet around the world will dust off their tired old story about the 27 Club, and reprint it with another name added to the list.