It's only second among the (up-and-coming) languages he chose. Scheme takes only 0.05% of the overall job market. Witness:

http://www.indeed.com/jobtrends?q=java%2C+C%2C+Python%2C+sch...

Do the actual search, and look at the results:

http://www.indeed.com/jobs?q=scheme&l=

Not so convincing anymore, right? Obviously, 'scheme' is a generic word, giving many vacancies not related to the programming language Scheme.

There's no good way to easily gauge scheme with this site. At first I typed in "scheme programming language" and came up with 250 results, "scheme lisp" provided something slightly more relevant, but with only 45. I didn't scroll through, but even then they were only asking for people with experience in scheme (among other languages that aren't de jour, like smalltalk).

Sure, at least he showed this percentage on the Y-axis. That still means there are five times as many Scheme jobs than for Scala, Groovy, Clojure, Erlang. That's not the case (although I wish it were), just search for "scheme" on indeed.com.

What did he do wrong?

Search for the string "scheme", as his trends link shows. That's a pretty common term, while "scala", "lua" etc. have no meaning other than the language, at least as far as a job site is concerned.

www.indeed.com redirects to www.indeed.ch here, but there I get nine hits for "scala" and 153 for "scheme". The scheme hits are all "Creates validation scheme with manufacturing site" etc.

Scheme is a niche language, and it's niche is education. It's almost never used in industry because it lacks a single popular standard library adds enough functionality to make it practical to use in the real world. In general, you don't write a practical scheme application, you write a practical [insert implementation]-scheme application.

Erlang will never have the same appeal to the masses as Python, not will it be as efficient at everyday scripting tasks.