Skip to contents

Checks the chain ladder against the claims of the latest calendar period: the chain ladder of the triangle without its latest diagonal predicts how much each origin develops in that period, which is compared with what was reported.

Usage

actual_vs_expected(triangle, ...)

Arguments

triangle

A triangle of cumulative claims, as for chain_ladder.

...

Arguments passed to chain_ladder() for the earlier triangle, such as average or periods.

Value

An object of class netsimr_reserve: a list with by_origin, a data frame with each origin's claims at its previous age, the expected and the actual development, their difference, the ratio of actual to expected, the standard error of the expected development and the standardised difference; total, the same for all origins; and chain_ladder, the chain ladder of the earlier triangle.

Details

The triangle without its latest diagonal (each origin's latest value left out) gives development factors \(f_k\) and Mack's \(\sigma_k\) and factor standard errors. An origin with \(C\) claims at its previous age \(k\) was expected to develop by \(C (f_k - 1)\), with a standard error \(\sqrt{C^{2 - \alpha} \sigma_k^2 + C^2 se(f_k)^2}\) (process and parameter error, \(\alpha\) as in chain_ladder). The standardised difference (actual less expected, over its standard error) shows which origins moved more than the method allows for: beyond about 2 in either direction is worth a look. The total's standard error adds that the origins at the same age share a factor. An origin that has a single value, or whose previous age has no development factor in the earlier triangle (the oldest origin, at the end of the triangle), has no expected development and is left out of the total.

See also

chain_ladder; merz_wuthrich for the uncertainty of the next period's development.

Examples

ave <- actual_vs_expected(raa_triangle)
ave
#> Actual vs expected development of the latest calendar period 
#> 
#>  origin previous expected actual difference actual_to_expected     se      z
#>    1981   18,662             172                                            
#>    1982   16,169       47    535        488             11.402    120  4.050
#>    1983   22,863      868    603       -265              0.695    182 -1.456
#>    1984   26,083    1,147    984       -163              0.858  1,535 -0.106
#>    1985   25,955    3,958    225     -3,733              0.057  1,016 -3.675
#>    1986   12,935    2,111  2,917        806              1.382    992  0.813
#>    1987   10,946    3,203  1,368     -1,835              0.427  3,124 -0.587
#>    1988    6,947    4,092  6,165      2,073              1.507  3,073  0.675
#>    1989    3,133    6,935  2,262     -4,673              0.326 10,651 -0.439
#>    1990                                                                     
#>   Total  125,031   22,360 15,059     -7,301              0.673 11,708 -0.624
ave$by_origin[, c("origin", "expected", "actual", "z")]
#>    origin   expected actual          z
#> 1    1981         NA    172         NA
#> 2    1982   46.92208    535  4.0504517
#> 3    1983  867.98197    603 -1.4557689
#> 4    1984 1146.81339    984 -0.1060883
#> 5    1985 3958.19260    225 -3.6751055
#> 6    1986 2110.82325   2917  0.8128652
#> 7    1987 3203.06059   1368 -0.5873377
#> 8    1988 4091.89296   6165  0.6746367
#> 9    1989 6934.62805   2262 -0.4386906
#> 10   1990         NA     NA         NA