Econometrics has always had better hiring stats than other econ subdisciplines, and historically, most Econ adjacent jobs were largely data cleansing, analysis, and benchmarking.
On the other hand, math, statistics, and coding is hard, and a large portion of Econ undergrads (primarily at non-target programs) lacked the wherewithal or ability to tackle Baby Rudin or a data structures class. In a lot of cases, their professors themselves lacked significant mathematical or econometrics sophistication.
What ended up happening was a glut of Econ majors who lacked the skills that were in demand in the Econ space (anything econometrics adjacent).
Now that Data Science as an undergrad disciple has arisen, a number of jobs that historically would have hired an Econ major even if they lacked some technical sophistication have switched to hiring Data Science majors. Additionally, Econ majors who had significant maths/stats/cs backgrounds usually ended up double majoring or minoring in those disciplines already.
And finally, at the top Econ undergrad programs, departments had already been extremely adamant about requiring their students to build maths/stats/cs sophistication beyond being a Stata monkey.
The fact that even undergrad Government majors at Harvard are expected to do statistical modelling at the statistics major level, and often take CS109 as well highlights how deeply computational the social science has become, and a large portion of undergrads just aren't cut out for that.
To paraphrase my friend: If you as an Econ undergrad cannot tackle Baby Rudin by your senior year, you have no employable future aside maybe law school, which itself is an increasingly losing proposition. You most likely won't do real analysis at the Baby Rudin level on a daily basis (if ever), but it's a positive signal about your mathematical and computational ability.
That said, CS/CE/EECS grads shouldn't be high and mighty either. I've seen programs in the US increasingly make core classes like circuits, computer architecture, data structures, OS, systems programming, and even algorithms beyond discrete structures optional.
The ideal of democratizing education is great, but should not come at the expense of foundational knowledge.