Interesting thought from Nick Seaver.  If all else fails, perhaps he
can use Pierce Piano Atlas as meta-data, combined with an assessment
of the proportion of pianos with player actions for each manufacturer,
and by year.  I.e., the number of pianos per year times the proportion
with player actions equals the approximate result.

Oh, I dunno...  Sounds like a will-o'-the-wisp.

Cheers
Jim Marke