Because the fact that the things likely to last the longest into the future are the ones that have already lasted and not the newest, most-apparently-forward-looking is counterintuitive to many people.
Or because they recognize that, but recognize that its also likely that something new will do even better than the things that are already old but that the hard part is picking the right new thing, and they think they have unique insight on the right new thing.
Or because they realize they probably don't have unique insight, but gambling on new has the best payoff if you are right, whereas the upside of choosing COBOL because its still trucking since the 50s is...you have job security eternally working on a COBOL [0] code base.
[0] And this generalized to all the metaphorical COBOLs out there in various domains, as well as actual literal COBOL.