If you sorted H1B applications by salary and only let in the top XX highest paid, then the allocation would simply skew heavily towards tech in Silicon Valley,

If you sorted H1B applications by salary and only let in the top XX highest paid, then the allocation would simply skew heavily towards tech in Silicon Valley, TX, WA, MA and finance in NYC, and almost none would be allocated to Kansas, Alabama, Mississippi, etc., and not much in lower-paid sectors. And if you only compared to median wages by state (not county, or metro area), then lower cost-of-living areas like Folsom CA would get hugely penalized for being stuck in the same bucket as SF/SV. (Also by the way now employers would have to compete against each other on raw salary, not stock grants, so you're removing the incentive component for early-stage startups, and raising their effective tax rate, which breaks how they're set up.) Arguably instead H1B should enforce some reasonable minimum salary, based on metro area. (don't confuse that with minimum wage). So if you want to reform things you need to construct something less simplistic which can't easily be gamed. (What you're describing is like Canada's Provincial Nominee Program, which is ~35% of their economic admissions, but it grants permanent residence, not just a work visa.) Anyway a less-discussed backstop solution to prevent abuse in H1B is to shorten (legislate) the maximum time to acquire (employment-based) Green Card - it originally used to be <6mths, now it can easily be 10+ years (from the date the employee first arrived on F1/H1B/L1, not the date the GC petition was finally raised, which can itself be 5+ years after that if the employer drags things out, which happens).