31 JULY 1941, Page 12

SAMPLING PUBLIC OPINION

Sm,—Your contributor " Janus " expresses some doubt on the weight which " should be attached to inquests like the Gallup Survey' Perhaps we may be allowed to submit some material points which will help in this respect. The method of investigation adopted by the B.I.P.O. consists in recoraing the reactions in the matters nods inquiry of a balanced sample of the population. The sample used based on careful statistical and social analyses of the population and due weight is given to the principal groupings into which all member, of the population fall iv one category or another. The degree if probable accuracy to be expected by interviewing different number of people may be accurately defined statistically. If, say, nine hun- dred people are questioned and their opinions divide evenly on a given subject, the mathematical chances are 997 in a thousand du the results obtained will be correct for the whole sample within t margin of error not greater than 5 per cent. Should opinion divide unevenly, then the margin of error will be even smaller. The sample used by the B.I.P.O. is such that the greatest margin of error expected on our results is between 3 per cent. and 4 per cent. with the leas favourable division of opinion.

This much said, the question remains to what extent may its sample tested by the B.I.P.O. be considered as representative of the British public. I venture to submit that the following results, checked by experience, permit us to claim with some justification that, indeed our results are representative.

(t) Using an identical technique, the American Institute of Public Opinion correctly estimated the results of the American Presidential Election within 2 5 per cent of accuracy. In no region was the margin of error greater than 4 per cent. This compares favourably with any other poll and certainly no other estimate was so uniformly correct throughout the country.

(2) In February, 1941, the British Institute asked when people interviewed had joined fire-parties. Our results indicated that 6o pa cent. had joined since Mr. Herbert Morrison's appeal at the beginning of January. Some weeks later Mr. Morrison, in answer to a parlia- mentary question, stated that in the London Civil Defence regiot the number of fire-bomb fighters increased from about 2oo,000 at the end of December to just over 7oo,000 in the middle of February. In 43 towns selected at random from all over the country, the nun' bets increased in the same period from about 190,000 to about 430,000. If these figures are weighted in accordance with proportion in the total population, a final figure indicates that about 59.5 pa cent. joined since the Home Secretary's appeal, as against our figure of 6o per cent Similar examples could be quoted for many of the outstanding issues of the day.

Trusting that we have made our point clear.—We are, yours faith-