> One needs to take that spread into account (i.e. realize that all people will not behave as you do) to get an optimal solution. The question to ask is "What behavior/policy will get the best outcomes given the differing behaviors you will see in society?"

The categorical imperative isn't about "is it likely that everybody will act like me" but "would it be good if everybody acted like me". You'll likely not end humanity if you start murdering your neighbors, because most people will not follow your example. Your murder might not even get noticed, so even the statistics won't change - and yet you'll probably agree that there's an ethical problem involved in murdering your neighbors.

I do agree with you that we shouldn't use Kant as the one-size-fits-all solution. Don't rely on telling people (or people understanding!) not to throw their trash everywhere, just put up a trash can so they can throw it in there. It is a solid framework for judging acts, but has it's weaknesses of course. The whole "the murderer is coming for your friend who is hiding in your bedroom, are you allowed to lie to him to save your friend" thing is something where I don't agree for example.

> The US is making the message clear: "We will not provide you with socialized healthcare. You are free to pursue whatever medical care as you wish." Which includes going abroad.

Sure, externalizing costs will provide advantages.