Coordination games and public goods games (which arguably model insurance) work best when people don’t adversely self-select, but coordinate around the social optimum (for insurance, when the risk pool is as large as possible). Whatever can orchestrate such coordination adds value. If people do it on their own, great, but some problems have characteristics like time horizons such that the coordination doesn’t happen without an authority. Yes, this brings in other public choice problems, but the trade-off is not necessarily bad.