Actuarial risk & insurance pricing
Collection
Open-data pricing and reinsurance models behind the Café Poliçe articles: GLM/GBM frequency, EVT large losses, XL priority. • 4 items • Updated
Danish fire losses: excess-of-loss priority (EVT)
Results of the Café Poliçe article "Reasürans Dilimini Nereye Koymalı?". Code: 000MAK/cafepolice-reinsurance-priority.
This repo holds the fitted tail model and the priority table, not the data. Data: danishuni (CASdatasets, GPL ≥ 2), 2,167 fire losses over 1 mDKK, Copenhagen Re 1980–1990, 1985 values.
| GPD above 10 mDKK (109 losses) | ξ = 0.50 (bootstrap 95% CI 0.13–0.79), σ = 6.98 |
| Lowest-cost priority (θ = 30% margin, 10% cost of capital) | 60 mDKK; 37.5–100 within 5% of the minimum |
| Cost vs no reinsurance | 34.2 vs 65.1 mDKK a year |
| Hold-out 1988–1990, priority 10 | expected 107, observed 213 mDKK a year (3.9% of simulations as bad) |
Files: metrics.json (all numbers), priority_table.csv (ceded loss, burning cost, SCR and cost per priority), 01_kuyruk_ve_muafiyet.png.
Assumptions: unlimited layer, Poisson frequency without trend, margin and cost of capital assumed. A method illustration, not a reinsurance recommendation.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support