I'm going to say something very cynical here - that multi-generational upper middle class Americans have realized that getting
really good at math provides diminishing returns, career wise. It's worth getting good enough at math to major in economics, but beyond that, the additional costs may outweigh the benefits.
To get an MBA, you need to be quite good at math, but hardly exceptional. To go to and elite law school, you don't need to be good at math though of course it helps. To go to medical school, you need to be good enough at math to take first year calculus, often (in many US universities) offered as an easier track separate from the one that most STEM majors take. That's probably the hardest math route taken by large numbers of upper middle class Americans, and it's still considerably easier than what you need to do to get a STEM degree even at the undergrad level.
There are certainly a few jobs requiring advanced math that pay exceptionally well, especially in quantitative finance, though there are plenty of those high paying jobs in finance for people who are "pretty good" at math.
Beyond that... engineering is a fine career choice, but let's face it, from an American (and perhaps British) class-based perspective, the american upper middle class places it below the professions (note that this often isn't the case in many other cultures). Law isn't a great place to be at the middle tier, but the upper middle class americans, the ones who can consistently hit 95%ile and up on standardized tests, don't go to those middle tier schools. They experience a different law grad market.
Beyond that, there's great work to be done in science, but in terms of pay and career stability, a PhD in Science is vastly more difficult than professional school and ultimately doesn't add nearly enough to justify the cost and time. And keep in mind, attrition rates in STEM PhD programs range from 35%-50% (including in elite programs), whereas the attrition rate from elite law and medical programs is typically below one half of one percent.
The incentive just isn't there, and the US has decided to fill these STEM jobs by 1) restricting general immigration to levels much lower than demand, and 2) creating specialized visas that allow people into the US on the condition that they get these degrees and work in this field. While that will fill the positions with high talent people from overseas, it further reduces the incentive for people who already have citizenship to go into these fields.
Being good at math is all benefit, of course, but there are trade-offs. As the effort increases and the marginal value decreases, it turns out there is an optimal point, though that optimal point will differ based on whether you need a specialized visa to access the US labor market.
For those with existing citizenship or residency rights, you maximize your earning potential by getting good enough at math to major in economics, but not by getting good enough at it to major in physics. If you don't have citizenship or a path to it (such as through family reunification), then the marginal benefits to getting really good at math remain positive, as this will gain you a limited right access the labor market in certain specialized fields experiencing a "shortage" of labor (albeit to a lesser degree than what US citizens already possess).