Simulates the claims of a portfolio once and puts them through each of several reinsurance programmes (a surplus treaty, a quota share, an excess of loss layer on each claim and an aggregate stop loss, alone or together), to compare what each costs with the capital it saves: the expected ceded losses, the reinsurance premiums, the margin paid to the reinsurers, the net result, its value at risk and the capital it needs.
Usage
compare_reinsurance(
frequency,
frequency_params,
severity,
severity_params,
programmes,
premium,
n = 10000,
level = 0.995,
loading = 0.25,
coc_rate = 0.1,
seed = NULL,
pareto_thresholds = NULL,
pareto_alphas = NULL
)Arguments
- frequency, frequency_params
The claim count distribution and its parameters, as in
simulate_claims, such as "Poisson" (lambda) or "Negative Binomial" (r, beta).- severity, severity_params
The claim size distribution and its parameters, as in
simulate_claims, such as "LogNormal" (meanlog, sdlog) or "Pareto" (alpha, x_m).- programmes
A named list of programmes, each a list with any of:
surplus, a list with theretention(a line), thelinesceded above it,sum_insured(the sum insured of each risk, or of each band of risks), and optionallyweights(the chance a claim falls on each risk and its share of the premium, such as its premium; equal by default),premium(the ceded premium) andcommission(the ceding commission, a share of the ceded premium);quota_share(the share ceded) andcommission(the ceding commission, a share of the ceded premium);xl, a list with thedeductible,limit,reinstatements(Inf, the default, for unlimited and free),reinstatement_rate(1 by default) and optionallypremium; andstop_loss, a list with theattachment,limitand optionallypremium. "No reinsurance" is added as the first programme.The subject premium of the year.
- n
The number of simulated years.
- level
The level of the value at risk, such as 0.995.
- loading
The share of a priced layer's or stop loss's premium for the reinsurer's expenses and profit.
- coc_rate
The cost of capital a year, for the value of the capital relief.
- seed
A whole number for a reproducible run, or
NULL. The stream of the R session is restored after a run with a seed.- pareto_thresholds, pareto_alphas
Pareto slices of the claim size, as in
simulate_claims: increasing claim sizes above which the tail of the severity is a Pareto, and the alpha of each.NULL(the default) for none.
Value
An object of class netsimr_programmes, a list with summary, a
data frame with a row per programme (the expected ceded loss, the reinsurance
premium and margin, the net mean result and its standard deviation, the
probability of a loss-making year, the value at risk of the net result, the
capital, the capital relief and its value), results, the simulated net
result of each programme (a column each), and the settings.
Details
Every programme sees the same simulated claims. In each, the surplus treaty cedes its share of every claim first, then the quota share its share of what is left; the excess of loss layer then recovers from each retained claim the part above its deductible, up to its limit, and at most \((1 + \textrm{reinstatements})\) limits a year; the stop loss last recovers the part of the year's retained claims above its attachment, up to its limit.
A surplus treaty cedes of each risk the part of its sum insured above the
retention, up to lines retentions, and that share of the risk's
premium and of each of its claims: nothing of a risk within the retention, and
\(\min(S - R, \textrm{lines} \times R) / S\) of a risk of sum insured \(S\)
above the retention \(R\). Each simulated claim falls on one of the risks of
sum_insured, drawn with the chances weights (the same risk in every
programme with the same risks), and its size is drawn as every other claim's,
whatever the sum insured of its risk. The treaty's premium is given or, when it is
not, the subject premium times the ceded share of the risks, weighted by
weights, so give each risk's premium as its weight.
The premium of a quota share is its share of the subject premium (of what the
surplus treaty leaves of it), less its ceding commission. A layer's or stop loss's
premium is given or, when it is not, priced at
the expected ceded loss with loading as the share of the premium for the
reinsurer's expenses and profit (technical_premium, with the
reinstatements the simulations use). The excess of loss layer's reinstatement
premiums are paid pro rata to the cover reinstated.
The net result of a year is the subject premium less the reinsurance premiums paid,
plus the ceding commission, less the retained claims (before other expenses). The
capital a programme needs is the mean net result less its \(1 - \textrm{level}\)
quantile: what a bad year at that level costs against an average one. The capital
relief is the capital without reinsurance less the programme's, and its value is
coc_rate of the relief less the margin paid for it: positive when the
programme saves more in the cost of capital than it costs.
See also
simulate_claims, technical_premium and
run_shiny_reinsurance_pricing_tool, which compares programmes on its
Programmes tab.
Examples
programmes <- list(
`Quota share 30%` = list(quota_share = 0.3, commission = 0.25),
`XL 4m xs 1m` = list(xl = list(deductible = 1e6, limit = 4e6, reinstatements = 1)),
`Stop loss 30m xs 45m` = list(stop_loss = list(attachment = 45e6, limit = 30e6)),
# four lines above a retention of 1m, on bands of sum insured weighted by premium
`Surplus 4 lines` = list(surplus = list(retention = 1e6, lines = 4, commission = 0.3,
sum_insured = c(0.5e6, 1e6, 2e6, 5e6, 10e6),
weights = c(20, 15, 12, 8, 5)))
)
cmp <- compare_reinsurance("Poisson", 200, "LogNormal", c(11, 1.5), programmes,
premium = 60e6, n = 5000, seed = 1)
cmp
#> Reinsurance programmes, 5,000 simulated years, capital at 99.5%
#>
#> programme ceded margin mean_result capital relief
#> No reinsurance 0 0 23,166,213 30,032,694 0
#> Quota share 30% 11,050,136 2,449,864 20,716,349 21,022,886 9,009,808
#> XL 4m xs 1m 5,145,233 1,715,078 21,451,135 27,923,811 2,108,883
#> Stop loss 30m xs 45m 785,492 261,831 22,904,382 8,951,704 21,080,989
#> Surplus 4 lines 8,841,471 1,238,529 21,927,684 24,996,490 5,036,204
#> value
#> 0
#> -1,548,883
#> -1,504,190
#> 1,846,268
#> -734,908
