The Onion

Whether or not an unsecured human poses a direct danger to another human in a crash, there are negative externalities to self-harm. One is that family members may be deprived of an important source of income and emotional support if you are killed in a crash. But the most directly-measurable cost is hospital expenses. At first glance medical expenses are another factor that affect only yourself. But in reality in the US emergency room care is guaranteed regardless of ability to pay, so plenty of ER costs are paid by hospitals or by governments. Emergency and non-emergency healthcare costs may be covered by Medicaid or Medicare - in other words, paid for by taxpayers. Maybe you can afford whatever treatment you might need, but policy must take into account people who cannot. That means that a personal choice not to wear a seatbelt, in aggregate, puts measurable costs on people who are not you.

Societal medical costs are discussed most prominently in relation to smoking. A study from last year estimates that healthcare made necessary by smoking costs an average of $2700 per person per year. That's a major part of the justification to tax cigarette sales. Healthcare costs caused by not wearing a seatbelt aren't as high, but are still substantial. Here's a study that found that hospital costs are 84% higher for people injured while not wearing a seatbelt vs wearing a lap & shoulder seatbelt.

Cost savings from seatbelt requirements might be smaller than savings from reduced smoking. But on the other hand the measurable burden of wearing a seatbelt is tiny. Policy should be based on the measurable costs & benefits of its requirements, and seatbelt requirements are a very clear-cut example of a net-benefit analysis.

You made an argument about the violation of personal liberty. When thinking about cost vs benefit there are two ways to look at this:

1. Intrinsic value of liberty: I don't know of a measurement of harm from restricting personal liberty in the specific case of seatbelt use. We have to draw a line somewhere on where personal liberty must be restricted to prevent obvious, unacceptable harm like murders. Considering the data, and the low burden of compliance seatbelt requirements seem to me to be an obvious case where restricting liberty as a worthwhile cost of harm reduction. Until there is some metric that shows that restriction of liberty may be more harmful than cost savings in this case I have to say, that's like, your opinion man.
2. Individuals are the best judges of their specific situation: Policy needs to consider the inevitable outcome of people exercising their right to implement bad judgement. If this were a case where a reasonable analysis could conclude that not wearing a seatbelt in some situations is a good decision then it would be a different story. But it's not. Data overwhelmingly shows that seatbelt use is the right decision in every case while driving. And data also shows that a high proportion of people make the wrong choice, likely due to a highly-inflated sense of their own invulnerability. Here's an analysis of how seatbelt requirements influence good judgement in several states.

I don't want to dismiss personal liberty. I think it is important to be able to make our own decisions. But it's also important to prevent extraordinarily-problematic decisions in certain cases. With seatbelts (I'm assuming we're not debating the cost of building seatbelts into cars at the moment) the measurable cost as far as I'm aware is the time taken to put the seatbelt on, which is negligible. Maybe there is a real cost to one's self of individualism to be required to put that seatbelt on. If the cost is real, there must be some way to measure it. Maybe that could be evaluating happiness, or creativity, or lifetime earnings, or some such thing. If we want to take this factor into account in cost-benefit analyses we have to have a measurement. We can't apply some arbitrary value because some people are going to say "infinite", and others are going to say "zero", and every value in between. Not only are those subjective opinions - those are self-evaluation estimates which humans tend to be bad at. I'm going to speculate that self-evaluations of the importance of liberty in the abstract is one of those areas people tend to get wrong. We need some kind of objective metric.

Edited for clarity