methods.cape_cod()
Cape Cod: Bornhuetter-Ferguson with the loss ratio estimated from the triangle.
Usage
methods.cape_cod(
cells,
*,
premium,
decay=1.0,
trend=0.0,
n_iters=1,
dev_grain_months=12,
average="volume",
history_periods=None,
drop_high=False,
drop_low=False,
preserve=1,
drop_above=None,
drop_below=None,
exclude=(),
exclude_valuations=(),
trim_ties="volume",
unsupported_factor="raise",
exhausted_exclusions="keep",
zero_cells="missing",
tail=None,
tail_factor=None,
tail_decay=None,
tail_attach_lag=None,
tail_fit_lags=None,
tail_steps=None,
tail_rows=None
)This is Gluck’s generalized Cape Cod. Each origin’s expected loss ratio is a weighted ratio of reported losses to “used-up” premium (premium times pct_reported) over all origins, where an origin k periods away is weighted by decay ** k. decay=1 (the default) weights every origin equally, which is the classic Cape Cod and one loss ratio for the whole triangle; decay=0 uses each origin’s own experience only, which is the chain ladder.
trend is an annual rate, such as 0.05 for 5% a year (0, the default, compares amounts as they are). Each origin’s losses are first brought to the valuation date’s level, multiplied by trend_factor = (1 + trend) ** (m / 12), where m is the whole months from the end of the origin period to as_of. The weighted ratio of those to used-up premium is the trended_loss_ratio, the loss ratio at the valuation date’s level (chainladder-python’s apriori_). Each origin’s expected_loss_ratio, applied to its own premium, is that divided by its own trend_factor (chainladder-python’s detrended_apriori_). With decay=1 the trended loss ratio is one number for the whole triangle; with decay=0 the trend cancels and the answer is the chain ladder’s.
n_iters iterates the result as benktander() does, with each origin’s premium * expected_loss_ratio as the first a priori ultimate; 1 (the default) is Cape Cod itself, and 0 and counts above 10,000 are refused.
cells, the development options and the tail options are as in chain_ladder() (with a tail, the used-up premium includes it), and premium as in bornhuetter_ferguson(). The loss ratios are in origins.expected_loss_ratio and origins.trended_loss_ratio, and origins.trend_factor is each origin’s factor (1.0 at trend 0).