One thing this doesn't necessarily account for is factors like population growth (particularly the the baby-boom) and the explosion of popular culture and celebrity notoriety through mass media. Are there simply considerably more people considered "celebrities" since say the 1950's than before?
I'd expect celebrity to be more "concentrated" for people born from 1920-1980, since the nature of mass media distribution naturally limited how many famous people there were. With the internet, attention has been fractured onto many smaller celebrities, though pop music still creates some very recognizable and controversial stars (Bieber, Kardashians).
I think that is accounted for in his trend line that you are correct, there is an increase. But what he shows is that the increase is beyond what should be expected.
His trend line is a linear one, so I don't believe it is accounting for anything other than previous death counts. Population growth and the expansion of celebrity through mass media would expect a more exponential curve. Life expectancy increasing may have also lead to a delay of seeing a more exponential uptick before now, but death always catches up eventually.
Anyways I don't have an answer just postulating that unless you would need to have a distribution of current (and past) celebrities and their ages as well to account for that factor. Something you could probably do with that Wikipedia analysis actually.
it doesn't feel like life expectancy changes would make such an impact, since it 'feels' like many of the celebrities died before life expectancy. i say feels because I haven't seen an age breakdown.
would it make sense to make many random draws of non-celebrities who died and compared the age of that group against the celebrity group?
