Internal Model Parameter Change and Regulatory Capital Strain

The finding is material: the copula parameter is a key driver of the 99.5% capital requirement, yet it is estimated from only eight years of data. That is a small sample for a dependence parameter, and dependence is difficult to estimate even with more data. The proposed increase from 0.45 to 0.60 may be prudent, but 'prudent' is not the same as 'better'. A higher parameter increases the modelled joint tail and the SCR by £35m; if the true dependence is lower, the insurer would hold excess capital and could be uncompetitive. If the true dependence is higher, the current model understates risk. The validation should test sensitivity, backtesting and expert judgement, and should consider whether the data period includes relevant stress events.

I would not submit the change solely on the grounds of prudence. I would first quantify parameter uncertainty, compare the estimate with industry benchmarks and alternative copulas, and assess the effect on the SCR, risk appetite and capital plan. If the evidence supports a range, the model could use a conservative point within that range, but the choice must be documented and approved by the model governance committee. The PRA will expect a clear rationale, not an arbitrary increase.

I would recommend a phased approach rather than an immediate submission of 0.60. The insurer should first run the model with a range of parameters, say 0.45 to 0.60, and present the results to the board with the capital impact and the evidence. If the board's risk appetite requires a higher confidence level, it could hold a management buffer or a capital overlay while the parameter evidence is strengthened. Alternatively, the insurer could improve the data set by extending the history, using external data, or applying a more formal Bayesian approach. A partial internal model change application is costly and time-consuming, so it should be made only when the evidence is robust.

If the PRA is likely to challenge the current parameter, delaying is not neutral: it leaves the model potentially understated. I would engage the PRA early with the validation findings, propose a timeline, and if necessary submit a change for a smaller increase with a clear sensitivity analysis. The CFO's concern about £35m is valid, but the cost of understating capital may be higher. The recommendation should balance model risk, regulatory credibility and capital strain.

I would communicate early and transparently, before submitting the formal change. The discussion should explain the validation finding, why the parameter matters, the data limitations, and the range of plausible estimates. I would provide the sensitivity analysis, the effect on the SCR and the capital plan, and the governance route for the decision. The PRA will want to see that the insurer is not selecting a parameter to minimise capital or to appear prudent without evidence; it will expect a documented expert-judgement process, independent validation, and board oversight.

I would also set out the alternatives considered, including retaining the current parameter with a capital buffer, and the rationale for the preferred option. If the insurer proposes a phased increase, the PRA should see the milestones and the data improvements. I would ask for feedback on the evidence required and keep a record of the interaction. The tone should be cooperative, not adversarial. The aim is to demonstrate that the model is credible and that the insurer understands its limitations, rather than to win an argument about a single number.