I wonder if Perl's age disproportionately skews the pay data. My thinking is that most Perl developers are mid- to senior-level engineers due to younger developers primarily using languages like Python or ruby. This heteroskedasticity would make using averages (and by definition also variance and standard deviation) ineffective, and require some correction to more accurately determine the true Perl developer pay distribution.