Actual versus expected analysis (A vs E)

At December 31, the 2022 accident year of general liability was 60% developed according to the selected pattern. By the following June 30, the same pattern led one to expect 1.2 million euros of newly reported losses, and 2.1 million came in. That is an actual versus expected analysis. For each accident year, the expected amount comes from the prior valuation: the selected ultimate times the difference between the percent developed at the new date and at the old one. That forces an annual pattern to be interpolated to the quarter or half year. It is run on paid and on incurred, on amounts and on counts, line by line, and it is repeated at every interim close, when nobody rebuilds the triangles. An A/E ratio persistently above 100% across several accident years at once points to adverse development. The question then becomes how much of it to carry into the ultimate: in a Bornhuetter-Ferguson reading the excess is added one for one, in a chain ladder reading it is multiplied by the remaining factor. The tool breaks on three fronts. A single large loss is enough to blow up the ratio of a small book without saying anything about the trend. A case reserve strengthening campaign makes incurred A/E jump while paid stays on pattern, hence the habit of reading the two side by side. And interpolation assumes reporting is even through the year, which the seasonality of claims handlers' closings often contradicts. Not to be confused with the claims development result, which measures the change in ultimate from one valuation to the next, or with the calendar year effect.