A Former President of the Institute once remarked that whereas the graduation of a mortality table might be considered a specialized technique not required by the practising actuary, the efficient testing of any graduation submitted to him should be an integral part of his actuarial ability. The student anxious to equip himself in this respect encounters considerable divergence of opinion regarding the best method of testing a graduation. There is complete agreement as to the duality of purpose of such a test: smoothness and fidelity to the data are essential factors. But whilst the criterion of smoothness is easily formulated—the run of the second or third differences of qx must show no awkward breaks—the views relating to the tests of adherence to the data are divided between the application of a “mean deviation” rule at each age or group of ages and a purely “practical” judgment based on that elusive virtue, experience.

The recent publication of the Registrar-General's Report on Occupational Mortality in England and Wales in connexion with the 1931 Census has suggested that a Note on this Report and its recent predecessors, with reference mainly to the methods employed in the investigation, may be of use to actuarial students.

Occupational Mortality has been investigated in connexion with every census since 1851. It may at first sight seem a simple matter to take, in each age group, the number exposed to risk in a given occupation from the census schedules and the number of deaths from the death registers, and from these figures to find the rates of mortality. In practice there are many pitfalls. The history of the successive investigations records various attempts to improve the statement and classification of the particulars and the presentation of the results, although there are still a number of known or suspected sources of error, the effect of which can only be estimated roughly, if at all.