And if this mystical top 5% can't spare the ten minutes to run through what is essentially an interviewing captcha - for a job they were obviously interested in enough to apply for in the first place - then how do I know they won't consider themselves too good to do their actual work if I do end up hiring then? I'm happy to hire the next 5% down the list of it means they have a good attitude and a willingness to do their job.
And what's the alternative? I waste my team's time setting up screening phone interviews for every halfway decent resume that comes down the pipe? With the number of applications that we get, we'd be losing days of time every time we look to hire, before we even got to the actual interviewing.
Every coding test I've been asked to take is a 2-5 hour job. I'll do it for a company I really want to work for... But not for a "maybe".
E.g. If I know you are paying 25% over market then 5 hours becomes worth it.
Now 10 minutes? I can spend 10 minutes on a maybe.
You're basically saying you want to lop off the bottom 60% of an imagined Normal distribution. Except you're really only lopping off some of the bottom 60%, and some fraction of that bottom 60% gets through, and you're paying the cost of also lopping off the top 5% who think you're a joke of an employer for being way too worried about the costs for you to more substantively evaluate those bottom 60% folks -- especially since you, as an employer, are probably in the bottom 60% of employers anyway, yet are cargo culting to try to act like you only hire the top 0.000001% or something.
You get what you pay for. If you go cheap on candidate evaluation (e.g. lazy commodity HackerRank), you get the McDonald's version of a developer, all while acting like you're being extremely selective.