Besides the JVM part, this is my take on it: as a language, Racket defines a vast possibility space, which is a superset of that defined with Clojure. You could implement Clojure (barring the JVM) quite naturally in Racket, while the other way around would be impractical. Now, Clojure is a pretty sweet spot in that design space, and someone else has done the immense work to carve it out, define it, implement it, and foster a sizable community around it.