Indeed is used often to show that language a has more adoption than language b. a and b being languages in the non- or pre-commercial domain. I'm only arguing that statements like "haskell overtakes prolog, statical typing > declarative programming" cannot be inferred from indeed data.
As for the "reasonable indicator": if the average java/php job points out that "haskell/erlang/prolog is a pre" , job ads might only be a reasonable indicator for the languages we already know to be at the top.