I understand your point about referrals really not being accounted for, but I am not sure if they can be. Given that most companies do not post information on referral jobs filled vs. new postings vs. turnover, we do what we can. I highly doubt that there is a significant difference in “job satisfaction” among the various programming languages, or that the data would really skew the statistics that much. The idea that this is the “approach of last resort” is true, but also very common. In many cases, jobs are posted at the same time as referral candidates are looked for.

The statistics are meant more as a barometer of the current conditions, and I also made no conclusions except for the direction the trends are taking.

As they say, “There are lies, damn lies and statistics.”

For instance, by now most COBOL programmers know each other, so they don’t often have to go hunting in public for qualified candidates. Also, to some degree while postings represent new jobs, they also represent “turnover”. So, the low COBOL numbers may just be showing that people in those positions are happier and stay longer, while the Java people might just be skipping around an awful lot. Since in some areas, putting out a job offer into the public is actually the “approach of last resort” for finding people, it may also mean that there are far more “undesirable” Java jobs than most other languages (they’re not getting filled by contacts).

Unfortunately, these types of stats aren’t really all that clear in what they are saying.