Interpretation & hypotheses
The PDF includes your edits, notes, working hypothesis and current what-if scenario. The Excel file holds the cleaned data, every table and every chart's numbers.
Interpretation
A 1-in-200 claims month costs about USD 4.01M (95% interval USD 3.94M to 4.09M); the exact FFT calculation gives USD 3.99M, inside that interval. An average month costs USD 1.51M, so covering a 1-in-200 month needs about USD 2.5M of capital above the expected loss.
Claims split into two groups: about 86% are routine claims and 14% are large ones, and the two-component model reproduces the observed average claim (USD 27.2k against 27.3k in the data). Counts show no clustering and no trend over the 36 months, so the Poisson frequency is a fair description.
A layer of USD 200k excess of USD 50k per claim pays about USD 10.4k per claim on average, a pure premium of USD 578k a month; the simulation agrees (USD 578k).
With that capital and a 25% premium loading, the chance of running out of money within 12 months is about 3.1% (± 0.2%). The main caveat is the short history: 36 months cannot show how often a truly extreme claim arrives, so treat the far tail with care.
Your working hypothesis
What do you expect, or want to test, before acting on these numbers? Saved in this browser and printed in the PDF.
Hypotheses tested
Every formal test behind this dashboard. H₀ is the null hypothesis; it is rejected when p < 0.05.
| Null hypothesis (H₀) | Test | Statistic | p-value | Decision at 5% | What it means |
|---|---|---|---|---|---|
| There is no trend in events per M over time | Spearman rank correlation with time | ρ = 0.032 | 0.852 | ✓not rejected | The rate looks stable over the period. |
| Event counts are Poisson (no clustering) | Likelihood ratio vs negative binomial (boundary-corrected) | LR = 0.02 | 0.441 | ✓not rejected | No evidence of clustering beyond chance. |
| The loss amounts follow the fitted Lognormal mixture (2) | Kolmogorov-Smirnov, parametric bootstrap (199 refits, n = 2,000) | D = 0.0090 | 0.94 | ✓not rejected | The model is consistent with the data. |
| The loss amounts follow the fitted Lognormal mixture (2) (tail-sensitive) | Anderson-Darling, parametric bootstrap | A² = 0.170 | 0.785 | ✓not rejected | No evidence the tails deviate from the model. |
| The model's mean loss equals the data's | z-test against the sample mean | z = -0.06 | 0.953 | ✓not rejected | The model reproduces the observed average. |
Aggregate loss per period
Collective risk model S = X₁ + … + X_N, simulated 50,000 times and computed independently with the FFT of the probability generating function; the two must agree.
Distribution of total loss per period
Bars: simulation. Line: exact discretised compound distribution.
Data table
| bin from | bin to | density |
|---|---|---|
| USD 179k | USD 286k | 3.735e-09 |
| USD 286k | USD 394k | 2.559e-08 |
| USD 394k | USD 501k | 8.721e-08 |
| USD 501k | USD 608k | 1.952e-07 |
| USD 608k | USD 715k | 3.141e-07 |
| USD 715k | USD 822k | 4.523e-07 |
| USD 822k | USD 929k | 5.786e-07 |
| USD 929k | USD 1.04M | 6.613e-07 |
| USD 1.04M | USD 1.14M | 6.777e-07 |
| USD 1.14M | USD 1.25M | 7.072e-07 |
| USD 1.25M | USD 1.36M | 6.981e-07 |
| USD 1.36M | USD 1.46M | 6.742e-07 |
| USD 1.46M | USD 1.57M | 6.217e-07 |
| USD 1.57M | USD 1.68M | 5.605e-07 |
| USD 1.68M | USD 1.79M | 4.945e-07 |
| USD 1.79M | USD 1.89M | 4.241e-07 |
| USD 1.89M | USD 2M | 3.649e-07 |
| USD 2M | USD 2.11M | 3.126e-07 |
| USD 2.11M | USD 2.21M | 2.672e-07 |
| USD 2.21M | USD 2.32M | 2.123e-07 |
| USD 2.32M | USD 2.43M | 1.899e-07 |
| USD 2.43M | USD 2.54M | 1.419e-07 |
| USD 2.54M | USD 2.64M | 1.285e-07 |
| USD 2.64M | USD 2.75M | 9.842e-08 |
| USD 2.75M | USD 2.86M | 7.526e-08 |
| USD 2.86M | USD 2.96M | 6.761e-08 |
| USD 2.96M | USD 3.07M | 5.752e-08 |
| USD 3.07M | USD 3.18M | 4.501e-08 |
| USD 3.18M | USD 3.29M | 3.81e-08 |
| USD 3.29M | USD 3.39M | 3.007e-08 |
| USD 3.39M | USD 3.5M | 2.222e-08 |
| USD 3.5M | USD 3.61M | 1.662e-08 |
| USD 3.61M | USD 3.71M | 1.961e-08 |
| USD 3.71M | USD 3.82M | 1.065e-08 |
| USD 3.82M | USD 3.93M | 9.525e-09 |
| USD 3.93M | USD 4.03M | 7.657e-09 |
| USD 4.03M | USD 4.14M | 9.151e-09 |
| USD 4.14M | USD 4.25M | 5.976e-09 |
| USD 4.25M | USD 4.36M | 5.603e-09 |
| USD 4.36M | USD 4.46M | 4.482e-09 |
| USD 4.46M | USD 4.57M | 4.109e-09 |
| USD 4.57M | USD 4.68M | 2.428e-09 |
| USD 4.68M | USD 4.78M | 1.494e-09 |
| USD 4.78M | USD 4.89M | 1.494e-09 |
| USD 4.89M | USD 5M | 1.868e-09 |
Return-period loss curve
The total loss exceeded on average once every N periods.
Data table
| x | loss |
|---|---|
| 1-in-2 | USD 1.4M |
| 1-in-5 | USD 1.98M |
| 1-in-10 | USD 2.36M |
| 1-in-20 | USD 2.73M |
| 1-in-50 | USD 3.22M |
| 1-in-100 | USD 3.6M |
| 1-in-200 | USD 4.01M |
| 1-in-500 | USD 4.5M |
| 1-in-1000 | USD 5M |
Risk measures
| Level | Return period | VaR · simulation | 95% CI | VaR · lattice | TVaR · simulation | TVaR · lattice |
|---|---|---|---|---|---|---|
| 90% | 1-in-10 | USD 2.36M | USD 2.35M – USD 2.37M | USD 2.37M | USD 2.9M | USD 2.91M |
| 95% | 1-in-20 | USD 2.73M | USD 2.7M – USD 2.74M | USD 2.73M | USD 3.28M | USD 3.28M |
| 99% | 1-in-100 | USD 3.6M | USD 3.55M – USD 3.66M | USD 3.6M | USD 4.2M | USD 4.21M |
| 99.5% | 1-in-200 | USD 4.01M | USD 3.94M – USD 4.09M | USD 3.99M | USD 4.64M | USD 4.65M |
Cross-checks (analytic vs simulated vs exact)
Differences beyond the Monte Carlo error mean something is wrong.
| Quantity | Analytic / exact | Simulated | Difference |
|---|---|---|---|
| Mean E[S] = E[N]·E[X] | USD 1.51M | USD 1.51M ± USD 5.78k | -0.07% |
| Std dev √(E[N]Var[X] + Var[N]E[X]²) | USD 663k | USD 659k | -0.62% |
| 99% VaR · lattice vs simulation | USD 3.6M | USD 3.6M | +0.13% |
| Probability mass captured by the FFT of the probability generating function | 100% | 100% |
Frequency
Events per M
Complete periods only; empty periods count as zero.
Data table
| x | events |
|---|---|
| 2022-01 | 53 |
| 2022-02 | 52 |
| 2022-03 | 75 |
| 2022-04 | 55 |
| 2022-05 | 65 |
| 2022-06 | 59 |
| 2022-07 | 56 |
| 2022-08 | 52 |
| 2022-09 | 52 |
| 2022-10 | 59 |
| 2022-11 | 45 |
| 2022-12 | 63 |
| 2023-01 | 49 |
| 2023-02 | 64 |
| 2023-03 | 64 |
| 2023-04 | 57 |
| 2023-05 | 46 |
| 2023-06 | 44 |
| 2023-07 | 44 |
| 2023-08 | 44 |
| 2023-09 | 60 |
| 2023-10 | 61 |
| 2023-11 | 50 |
| 2023-12 | 48 |
| 2024-01 | 48 |
| 2024-02 | 51 |
| 2024-03 | 62 |
| 2024-04 | 51 |
| 2024-05 | 46 |
| 2024-06 | 57 |
| 2024-07 | 66 |
| 2024-08 | 65 |
| 2024-09 | 53 |
| 2024-10 | 64 |
| 2024-11 | 59 |
| 2024-12 | 61 |
Trend test
| Spearman ρ (counts vs time) | p-value | first 12 mean | last 12 mean | ratio |
|---|---|---|---|---|
| 0.032 | 8.52e-01 | 57.2 | 56.9 | 1.00× |
Frequency model
Observed vs fitted count probabilities
Share of periods with exactly k events.
Data table
| x | observed | Poisson | Negative binomial |
|---|---|---|---|
| 0 | 0% | 0% | 0% |
| 1 | 0% | 0% | 0% |
| 2 | 0% | 0% | 0% |
| 3 | 0% | 0% | 0% |
| 5 | 0% | 0% | 0% |
| 6 | 0% | 0% | 0% |
| 7 | 0% | 0% | 0% |
| 8 | 0% | 0% | 0% |
| 10 | 0% | 0% | 0% |
| 11 | 0% | 0% | 0% |
| 12 | 0% | 0% | 0% |
| 13 | 0% | 0% | 0% |
| 15 | 0% | 0% | 0% |
| 16 | 0% | 0% | 0% |
| 17 | 0% | 0% | 0% |
| 18 | 0% | 0% | 0% |
| 20 | 0% | 0% | 0% |
| 21 | 0% | 0% | 0% |
| 22 | 0% | 0% | 0% |
| 23 | 0% | 0% | 0% |
| 25 | 0% | 0% | 0% |
| 26 | 0% | 0% | 0.001% |
| 27 | 0% | 0.001% | 0.001% |
| 28 | 0% | 0.002% | 0.002% |
| 30 | 0% | 0.006% | 0.007% |
| 31 | 0% | 0.011% | 0.013% |
| 32 | 0% | 0.019% | 0.023% |
| 33 | 0% | 0.032% | 0.037% |
| 35 | 0% | 0.084% | 0.094% |
| 36 | 0% | 0.13% | 0.14% |
| 37 | 0% | 0.19% | 0.21% |
| 38 | 0% | 0.28% | 0.31% |
| 40 | 0% | 0.56% | 0.6% |
| 41 | 0% | 0.76% | 0.8% |
| 42 | 0% | 1% | 1.1% |
| 43 | 0% | 1.3% | 1.3% |
| 45 | 2.8% | 2% | 2.1% |
| 46 | 5.6% | 2.5% | 2.5% |
| 47 | 0% | 2.9% | 2.9% |
| 48 | 5.6% | 3.4% | 3.4% |
| 50 | 2.8% | 4.2% | 4.2% |
| 51 | 5.6% | 4.6% | 4.6% |
| 52 | 8.3% | 4.9% | 4.9% |
| 53 | 5.6% | 5.2% | 5.1% |
| 55 | 2.8% | 5.4% | 5.3% |
| 56 | 2.8% | 5.3% | 5.2% |
| 57 | 5.6% | 5.2% | 5.1% |
| 58 | 0% | 5% | 4.9% |
| 60 | 2.8% | 4.3% | 4.3% |
| 61 | 5.6% | 3.9% | 3.9% |
| 62 | 2.8% | 3.5% | 3.5% |
| 63 | 2.8% | 3.1% | 3.1% |
| 65 | 5.6% | 2.3% | 2.3% |
| 66 | 2.8% | 1.9% | 2% |
| 67 | 0% | 1.6% | 1.6% |
| 68 | 0% | 1.3% | 1.4% |
| 70 | 0% | 0.84% | 0.88% |
| 71 | 0% | 0.66% | 0.69% |
| 72 | 0% | 0.51% | 0.54% |
| 73 | 0% | 0.39% | 0.41% |
| 74 | 0% | 0.29% | 0.31% |
Count models
| Model | MLE parameters | log-lik | AIC | ΔAIC | χ² | df | χ² p |
|---|---|---|---|---|---|---|---|
| Poisson | λ = 55.56 | -123.8 | 249.7 | 0.0 | 3.70 | 4 | 0.4475 |
| Negative binomial | r (size) = 1584, p = 0.9661 | -123.8 | 251.7 | 2.0 | 0.89 | 3 | 0.8273 |
| Geometric (from 0) | p = 0.01768 | -180.9 | 363.9 | 114.2 | 134.65 | 4 | 0.0000 |
Severity
Severity distribution
Bars: individual losses. Lines: fitted models.
Data table
| bin from | bin to | density |
|---|---|---|
| USD 466 | USD 574 | 0.02774 |
| USD 574 | USD 706 | 0.03884 |
| USD 706 | USD 869 | 0.04439 |
| USD 869 | USD 1.07k | 0.1054 |
| USD 1.07k | USD 1.32k | 0.172 |
| USD 1.32k | USD 1.62k | 0.3551 |
| USD 1.62k | USD 1.99k | 0.4883 |
| USD 1.99k | USD 2.45k | 0.5271 |
| USD 2.45k | USD 3.02k | 0.6825 |
| USD 3.02k | USD 3.71k | 0.9044 |
| USD 3.71k | USD 4.57k | 1.043 |
| USD 4.57k | USD 5.62k | 0.9932 |
| USD 5.62k | USD 6.92k | 0.91 |
| USD 6.92k | USD 8.52k | 0.8434 |
| USD 8.52k | USD 10.5k | 0.7768 |
| USD 10.5k | USD 12.9k | 0.5992 |
| USD 12.9k | USD 15.9k | 0.4494 |
| USD 15.9k | USD 19.5k | 0.2996 |
| USD 19.5k | USD 24k | 0.1997 |
| USD 24k | USD 29.6k | 0.1054 |
| USD 29.6k | USD 36.4k | 0.1165 |
| USD 36.4k | USD 44.8k | 0.09433 |
| USD 44.8k | USD 55.1k | 0.2108 |
| USD 55.1k | USD 67.8k | 0.1165 |
| USD 67.8k | USD 83.5k | 0.1221 |
| USD 83.5k | USD 103k | 0.1387 |
| USD 103k | USD 126k | 0.09433 |
| USD 126k | USD 156k | 0.1165 |
| USD 156k | USD 191k | 0.1609 |
| USD 191k | USD 236k | 0.111 |
| USD 236k | USD 290k | 0.07213 |
| USD 290k | USD 357k | 0.05549 |
| USD 357k | USD 439k | 0.03329 |
| USD 439k | USD 540k | 0.03329 |
| USD 540k | USD 665k | 0.005549 |
| USD 665k | USD 818k | 0.02774 |
| USD 818k | USD 1.01M | 0.01665 |
| USD 1.01M | USD 1.24M | 0 |
| USD 1.24M | USD 1.52M | 0 |
| USD 1.52M | USD 1.88M | 0.005549 |
Q-Q plot · Lognormal mixture (2)
Points on the dashed line mean the fitted model reproduces the observed quantiles. Bending away at the right end = the tail is mis-modelled.
Data table
| model quantile | observed quantile |
|---|---|
| 344 | 466 |
| 940 | 931 |
| 1.24k | 1.25k |
| 1.47k | 1.49k |
| 1.62k | 1.59k |
| 1.8k | 1.76k |
| 1.97k | 1.91k |
| 2.09k | 2.05k |
| 2.25k | 2.22k |
| 2.41k | 2.37k |
| 2.52k | 2.5k |
| 2.67k | 2.69k |
| 2.82k | 2.87k |
| 2.94k | 2.96k |
| 3.09k | 3.11k |
| 3.25k | 3.28k |
| 3.36k | 3.38k |
| 3.52k | 3.54k |
| 3.68k | 3.71k |
| 3.81k | 3.81k |
| 3.98k | 3.97k |
| 4.15k | 4.16k |
| 4.29k | 4.32k |
| 4.47k | 4.43k |
| 4.66k | 4.62k |
| 4.8k | 4.76k |
| 5.01k | 4.98k |
| 5.22k | 5.16k |
| 5.39k | 5.41k |
| 5.61k | 5.62k |
| 5.86k | 5.91k |
| 6.04k | 6.08k |
| 6.31k | 6.38k |
| 6.59k | 6.63k |
| 6.81k | 6.89k |
| 7.13k | 7.16k |
| 7.47k | 7.52k |
| 7.75k | 7.8k |
| 8.14k | 8.28k |
| 8.58k | 8.75k |
| 8.92k | 9.1k |
| 9.45k | 9.58k |
| 10k | 10.1k |
| 10.5k | 10.6k |
| 11.3k | 11.4k |
| 12.1k | 12.3k |
| 12.9k | 12.9k |
| 14.1k | 13.9k |
| 15.8k | 15.5k |
| 17.4k | 17.1k |
| 20.4k | 19.8k |
| 25.4k | 24.5k |
| 31.8k | 33.3k |
| 45k | 47.1k |
| 63.4k | 60.2k |
| 79.9k | 74.7k |
| 108k | 105k |
| 148k | 158k |
| 196k | 192k |
| 323k | 322k |
| 1.51M | 1.88M |
Severity models
| Model | MLE parameters | log-lik | AIC | ΔAIC | KS D | Akaike weight |
|---|---|---|---|---|---|---|
| Lognormal mixture (2) | w1 = 0.8569, μ1 = 8.507, σ1 = 0.775, w2 = 0.1431, μ2 = 11.47, σ2 = 0.9444 | -20,936.9 | 41,883.8 | 0.0 | 0.0090 | 100% |
| Burr XII | c = 2.88, d = 0.2936, scale = 2665 | -20,999.8 | 42,005.7 | 121.9 | 0.0323 | 0% |
| Inverse Gaussian | mean μ = 2.73e+04, shape λ = 5016 | -21,111.8 | 42,227.6 | 343.8 | 0.1386 | 0% |
| Log-logistic | shape β = 1.444, scale α = 6478 | -21,171.0 | 42,345.9 | 462.1 | 0.0617 | 0% |
| Lognormal | μ (log-mean) = 8.931, σ (log-sd) = 1.312 | -21,243.2 | 42,490.4 | 606.6 | 0.1171 | 0% |
| Pareto II (Lomax) | tail index α = 1.325, scale θ = 9726 | -21,311.0 | 42,626.0 | 742.2 | 0.1233 | 0% |
| Weibull | shape k = 0.6211, scale λ = 1.554e+04 | -21,709.4 | 43,422.7 | 1,538.9 | 0.1788 | 0% |
| Gamma | shape α = 0.4957, scale θ = 5.506e+04 | -21,983.8 | 43,971.5 | 2,087.7 | 0.2591 | 0% |
| Exponential | mean θ = 2.73e+04 | -22,429.0 | 44,859.9 | 2,976.1 | 0.3964 | 0% |
| Lognormal mixture (3) | w1 = 0.3477, μ1 = 7.988, σ1 = 0.6399, w2 = 0.5002, μ2 = 8.839, σ2 = 0.6277, w3 = 0.152, μ3 = 11.39, σ3 = 0.9825 | -20,935.9 | 41,887.8 | 3.9 | 0.0100 | not selected: BIC prefers Lognormal mixture (2) |
| Lognormal mixture (4) | w1 = 0.346, μ1 = 7.915, σ1 = 0.6052, w2 = 0.2259, μ2 = 8.455, σ2 = 0.3775, w3 = 0.2656, μ3 = 9.212, σ3 = 0.4316, w4 = 0.1625, μ4 = 11.3, σ4 = 1.027 | -20,934.2 | 41,890.3 | 6.5 | 0.0065 | not selected: BIC prefers Lognormal mixture (2) |
Model checks · Lognormal mixture (2)
A good fit reproduces the observed mean, spread and centre. Large gaps mean conclusions depend on the model's tail more than on the data.
| Quantity | Model | Data | Verdict |
|---|---|---|---|
| Mean | USD 27.2k | USD 27.3k ± USD 3.76k | ✓ within the 95% interval |
| Standard deviation | USD 84.7k | USD 85.9k | |
| Median | USD 5.83k | USD 5.87k | |
| Expected count above the largest observation n·S(max) | 0.24 | ≈ 1 or less is consistent | ✓ |
Excess-of-loss layer
Per-event layer USD 200k xs USD 50k: the reinsurer pays min((X − R)⁺, L) on every event.
Layer pricing
| Quantity | Value |
|---|---|
| Expected ceded per event ∫ S(x) dx over the layer | USD 10.4k |
| Pure premium per period = E[N] × expected ceded per event | USD 578k |
| Simulated ceded per period (check) | USD 578k ± USD 2.55k |
| Layer share of expected loss | 38.3% |
| Probability the layer is hit in a period | 99.8% |
| 99.5% VaR of ceded loss | USD 1.47M |
Solvency (Cramér-Lundberg ruin model)
Surplus U(t) = u + c·t − S(t) with Poisson arrivals at the fitted rate. Ruin = surplus below zero at any claim.
Ruin analysis
| Quantity | Value |
|---|---|
| Initial capital u | USD 2.5M |
| Premium loading θ | 25% |
| Premium income per M c = (1+θ)λE[X] | USD 1.89M |
| Ruin probability within 12 M (simulated, 20,000 paths) | 3.1% ± 0.24% |
| Diffusion approximation, within 12 M (reliable only for many, light-tailed claims; the simulation is the reference) | 1.2% |
| Diffusion approximation, infinite horizon exp(−2θλE[X]u / λE[X²]) | 1.4% |
| Single-large-claim asymptotic, infinite horizon (valid only for very large u) | 0% |
Method
Collective risk model. S = Σᵢ₌₁ᴺ Xᵢ, N independent of the i.i.d. Xᵢ. E[S] = E[N]E[X]; Var[S] = E[N]Var[X] + Var[N]E[X]².
Exact compound distribution. Severity discretised on a lattice of span h by the rounding method fₖ = F((k+½)h) − F((k−½)h). Panjer: f_S(x) = [1/(1 − a f₀)] Σ_{y=1}^{x} (a + by/x) f_y f_S(x−y), with (a,b) = (0, λ) for Poisson and (1−p, (r−1)(1−p)) for NB. Large E[N]: f_S = IFFT(P_N(FFT f)) with zero padding.
Simulation. N drawn from the fitted count model, severities from the fitted severity model; quantile CIs from order statistics.
Layer pricing. E[min((X − R)⁺, L)] = ∫_R^{R+L} S_X(x) dx; pure premium = E[N] × that.
Ruin. Cramér-Lundberg U(t) = u + ct − S(t), c = (1+θ)λE[X]. Lundberg: λ + cR = λM_X(R), ψ(u) ≤ e^(−Ru); exponential claims: ψ(u) = e^(−Ru)/(1+θ). Subexponential claims: ψ(u) ~ (1/(θE[X])) ∫_u^∞ S_X(x) dx.
Assumptions and limitations
- Frequency and severity are assumed independent, and events independent of each other. Contagion or common shocks make the real tail heavier than shown.
- Severity parameters are fitted on all history; inflation or a change in the mix of events is not trended. Trend the amounts first if the period is long.
- The ruin model uses Poisson arrivals at the mean rate; with over-dispersed counts (dispersion 1.07) the true ruin probability is higher.
- Monte Carlo figures carry sampling error; intervals are shown where it matters.