Page:American Journal of Sociology Volume 3.djvu/643

 OFFICIAL STATISTICS 629

suit the workman himself and ascertain from him the amount of wages he actually received. When you have obtained this from all the workmen in the country, the total wages may be divided by the total number and the true average earnings ascertained. It is only under such conditions that the total number can be correctly used as a divisor. A similar division may, of course, be made for a representative number, but it will only be a rep- resentative average wage. This representative average is the only average that it is practicable to obtain, and it is given in all statistical works on wages.

A serious error is frequently made in the use of census statis- tics by blending data reported in answer to apparently similar questions, but applied to entirely different lines of investigation. For instance, one writer states : " There is another important factor in this problem entirely overlooked by Colonel Wright. The census reports as adults males over sixteen and females over fifteen, classifying the remainder as children. In ascertaining the age, however, the question asked at the last census called for 1 age nearest birthday,' which would include as children males under sixteen and a half, females under fifteen and a half. At the preceding census 'age last birthday' was called for. which would include males to their seventeenth and females to their sixteenth birthday." The census reports that classify as adults males over sixteen and females over fifteen relate entirely to the statistics of manufactures as reported by establishments. The latter portion of the quotation refers to questions contained in the general population schedule. The two have no connection whatever. Colonel Wright had reference entirely to the statistics as reported by manufacturing establishments, and in them the question as to "age nearest birthday" or at "last birthday" did not appear at either census. The writer quoted was endeavoring to criticise the statistics of population, which for 1890 show the number of children in occupations arranged in age-groups of ten to fourteen years, inclusive, instead of ten to fifteen, as at prior censuses. While I was not connected with the compilation of the statistics of population, it appears that the quinquennia < riod was