For most of the last two decades an Indian insurer held capital against a formula: a fixed percentage of reserves and sum at risk, with a 150% solvency ratio as the floor. The formula did not care whether your book was a diversified set of term policies or a concentrated coastal property portfolio. Capital was a fixed cost of doing business, and pricing could be done in one department while solvency was managed in another.
Risk-based capital ends that separation. When the capital you hold depends on the tail of your own loss distribution, three things that used to be separate become one decision.
Pricing becomes a capital decision. A catastrophe layer that attaches at a 1-in-10-year loss and is exhausted by a 1-in-50 has an expected loss of a few percent of limit and a 99.5% tail equal to the whole limit. Under RBC the capital charge on that layer is the difference, and the cost of that capital is most of the premium. The pricing lab shows this directly: for the reference layers the premium is 2.7 to 3.3 times the expected loss, and for earthquake more than ten times, with the capital charge doing nearly all the work. A pricing actuary who cannot see the tail cannot price the layer.
Modelling becomes a regulatory conversation. If the tail drives capital, the regulator will want to know where the tail came from. A model that produces a number is not enough; it must show the threshold diagnostics, the fitted tail index with its uncertainty, the benchmark it was compared against, and the data it did not touch. This is why every diagnostic in the lab sits on the same screen as the price. A generative model that cannot be audited will not be allowed to reduce capital, and the actuaries I interviewed for my research were unanimous on that point.
Diversification becomes visible. Under a formula, writing a flood treaty in Bihar and a drought cover in Maharashtra costs the same capital as writing two of either. Under RBC the correlation between them is what sets the combined charge, and a reinsurer that can measure it has a structural advantage over one that cannot. The 21-dimensional drought model in the lab exists for exactly this reason: a model that treats sub-divisions as independent under-prices the aggregate layer by about a sixth.
None of this is an argument for lower capital. It is an argument that in an RBC regime the reinsurer’s product is its model, and the model must be good enough to show to someone who does not want to believe it.