They used to talk about anti-selection (a/k/a "adverse selection"). For example, if a retail insurer has no reinsurance, it has to incorporate both the probability of claims and the distribution of payout sizes.
But if they have reinsurance, the payouts are 'clipped' at their deductible. This means that they have an incentive to take on clients that are less likely to have any sort of accident, even if when they do have an accident, it has catastrophic consequences.
This means that a reinsurance company cannot rely on the overall statistics for claims, because the insurance companies that buy reinsurance price their products with reinsurance in mind. The people unlikely to run into massive claims will end up in pools where the retail insurance company doesn't buy reinsurance.
So they have to carefully price the reinsurance to account for the fact that the insurance company is packaging their most reinsurance-sensitive pool of customers together.
But if they have reinsurance, the payouts are 'clipped' at their deductible. This means that they have an incentive to take on clients that are less likely to have any sort of accident, even if when they do have an accident, it has catastrophic consequences.
This means that a reinsurance company cannot rely on the overall statistics for claims, because the insurance companies that buy reinsurance price their products with reinsurance in mind. The people unlikely to run into massive claims will end up in pools where the retail insurance company doesn't buy reinsurance.
So they have to carefully price the reinsurance to account for the fact that the insurance company is packaging their most reinsurance-sensitive pool of customers together.