Where is the evidence that learning COBOL pays and is in demand?

I've worked with companies that provide COBOL contractors and no, it does not pay well. The vast amount of fees collected from clients go to consultancies that train-up interns in South Asia (have huge staff turnover) and these interns-cum-employees A: Stick with it in the hopes of being directly employed by their client (which happens, a F500 job in South Asia remains a big deal for many), or B: See those that did other technologies get paid much better after only a couple of years but get locked-in to their vendor through wider skills.

Niche technology skills can become suddenly in-demand, and in Y2k COBOL certainly was. Is there another Y2k event hitting COBOL? The higher paying jobs are only in architecture roles which means intimately knowing a vast code-base (product-specific) rather than COBOL itself.

This is my anecdotal observation having worked with COBOL/CICS/DB2 teams over a few years.

I also agree with the comments from smt88. Haskell would take a tad longer to get proficient in but once so you could sneeze in the direction of a large company that uses it like, I don't know, StanChart, and probably get a good offer. COBOL's not like that. Unless you're particularly experienced on a particular code base, it's easy to cycle-in a replacement (for any large employer).

I have one anecdote. It may not be enlightening for you at all.

My high school CS teacher made ~$50k per year doing quick contract jobs working on COBOL, mostly for factories in the Ohio area.

These jobs were never posted online. Once the companies found a contractor to fix the systems, they just kept using the same one, not unlike a homeowner who has a favorite plumber.

It's very likely that the majority of the COBOL community is hidden from online searches because churn is low and/or they don't use reddit, StackOverflow, etc.

The things I've read about Haskell remind me of what people are saying about COBOL now. Very few companies use Haskell, but the ones that are using it can sometimes be desperate to find a Haskell expert and will pay a lot for one.

COBOL doesn't have a lot of jobs; the systems that are still running are old, stable, and get rare and mostly minor changes.

There is little and shrinking steady-state demand and that's mostly met by the a dwindling pool of old-timers and internal staff that get trained by their workplace to assist or replace the old-timers.

On the other hand, occasionally one of those big enterprises (public or private) encounters an urgent, immediate need for some new work on one of those systems, and it's those moments where there is an intense premium for COBOL programmers because there's little idle capacity to take up a temporary demand surge. (E.g., right now with some state unemployment systems suddenly needing substantial rules changes and modifications to handle order-of-magnitude volume increases over baseline capacity.)

Banking software developer here. Cobol is specialized and mainly for maintenance. In banks. Has been mainly phased out, if it can be.

It is an investment in a dead/dying language. Why is that worth your time and effort? Have you tried Cobol programming? Do you like it? I dropped out of my Cobol class in college. Was mind-numbingly boring and archaic to work with.

Are you trying to find a viable alternative to JavaScript? That pays 6 figures? You could pivot your javascript skills into something like react-native... mobile app development.