630 THE AMERICAN JOURNAL OF SOCIOLOGY
eliminated are the ones which it required the most time to answer correctly from the books. The fact is that it was quite impos- sible for the ordinary special agent, at the censuses of 1900 and 1905, to fill the schedules completely from actual bookkeeping figures within the limit of time which he was allowed. He was practically forced to resort to estimates with respect to certain of the questions, and this naturally tempted him to accept estimates for other questions which could have been more readily answered from actual bookkeeping data.
These condensations in the manufactures schedule are prin- cipally in the following three respects :
I. At the censuses of 1900 and 1905 the schedule called for the average number of men, the average number of women, and the average number of children under 16 employed during each month of the year. In the office these monthly averages were combined to give the annual average. Inasmuch as pay- rolls do not distinguish age and do not in all cases distinguish sex, the mere segregation by age and sex involved great difficulty and practically necessitated estimates. Moreover, to obtain the average number of employees in any single month would have required the examination of all the pay-rolls for that month, giving regard to the number of days on which each person was employed, and adjusting the cases where a single pay-roll ex- tended from one month into another. All this was quite im- possible in any reasonable length of time. At the present census we are asking for the distinction between men, women, and chil- dren only for one single date in the year. The distribution as between men, women, and children for that date will be assumed, as it safely can, to be fairly typical of the distribution for the year as a whole. With a view to ascertaining the annual average number of employees of all classes combined, we ask, not the average number for each month, but the number employed on the fifteenth day of each month. The average for these twelve days will give substantially the same result as an average calcu- lated from twelve monthly averages ; and the data can be obtained from the actual pay-rolls in but a small fraction of the time which would have been required to calculate monthly averages.
