Search390 reports the results of a Micro Focus survey on mainframes and COBOL. According to the survey, COBOL is still the dominant language on mainframe computers and the median age of COBOL programmers is 45-59. This leads to the question: where will the next crop of COBOL programmers come from?
An important data point and one that I didn't see in the report was how many new COBOL applications are being written? There's reported to be 180-200 billion lines of COBOL code out there, but let's be frank--no one really wants to prepare for a job maintaining 20 years old code.
When I was CIO for Utah, we had old COBOL applications running on the mainframe. For many agencies, these were some of the most mission critical and important applications. On the other hand, we were also actively working to replace them as we could. It wasn't the case that we were replacing them just because they were in COBOL, but they were old and not meeting functional needs. No one thought we ought to do major refactoring in COBOL to add feature functionality.
One of the points of the Search390 article was that we should encourage CS students to learn COBOL. I wouldn't recommend it to my students and certainly not to my children. I try to move my students toward knowledge, skills, and jobs that are somewhat protected from offshoring. COBOL maintenance jobs don't fit that category as far as I can see.