A month ago I mentioned that in his forthcoming book,
Poverty and Progress, Mr. Seebohm Rowntree had recorded the results of an interesting test of the validity of " sample- surveys " (questioning five or ten per cent. of a given com- munity on the assumption that the results for -that proportion will hold good of the whole). The book is published this week, and some details of the test can be quoted. Mr. Rowntree's investigators in York dealt altogether with 16,362 families and averaged the condusions in such matters as family budgets and various special items of expenditure. Those conclusions were of- necessity mathematically accurate. Then Mr. Rown- tree tried the sampling-method, dealing with every tenth, and then with every twentieth and every fiftieth family, to see how far conclusions based on them tallied with those derived from the whole 16,362. In some cases, i.e., the number of families living below the poverty-line, the sampling-method proved remarkably reliable. In other cases, such as analysis of the causes of poverty, the result is very different. The con- clusions are tabulated in seven groups, and in two of the seven the sampling-method result is misleading by over 20 per cent. Mr. Rowntree goes into the matter in great detail, giving eleven pages of tables to tell their own story. He does not give his own verdict on the figures. Broadly speaking, they suggest that " sample-results " are usually within 15 per cent. of the truth—either way.