The Great Enumeration


Yet the pressure for more data consistently exceeded the capability of the technical and administrative machinery to provide and disgorge the answers. The encyclopedic censuses of 1880 and 1890 surpassed all previous efforts, employing more than two hundred schedules which asked an almost unbelievable thirteen thousand questions. They included, just for a few examples, the number of old soldiers and their widows still alive, the number of pounds of butter made, and data on fire losses, fire departments, garbage disposal, cemeteries, rail traffic, prisons, idiots, occupations, and unemployment. Besides this, the 1890 census attempted to include all Indians for the first time, which required a Solomonic decision as to who was an Indian. Biological, legal, and cultural considerations were taken into account. The Census Bureau decided that in addition to full-blooded Indians, persons of mixed blood were Indians if they were enrolled by a tribe or registered at an Indian agency, or if those who knew them regarded them as Indians. This definition undoubtedly admitted some Mexicans and whites who had dropped out of conventional American life. The information gathered included such fascinating facts as the number of Indian polygamists, the number of Indians killed during the year, including women and children, whether by soldiers or interested citizens, and the number of whisky sellers prosecuted.

Although the counts of 1880 and 1890 produced an acute case of statistical indigestion, they did profit from an exhaustive study of census operations conducted by a House Committee on the Ninth Census, headed by Ohio congressman James A. Garfield. The marshals were replaced by civilian supervisors, pay was liberalized, accuracy improved, and confidential information was protected by law. For the first time, women appeared, two hundred of them, among the army of fifty thousand census takers.

Up to and including the census of 1880, a multitude of clerks was employed in tabulating the census, shuffling the piles of schedules, making tally marks on ruled sheets, a task so slow and tedious that the summaries were out of date before they could be published. The outlook for the equally complex 1890 census was even more dismal when a fortunate meeting of two able minds led to the invention of tabulating machinery and the era of the punched card.

According to a generally accepted account, Dr. John Shaw Billings, a surgeon and managerial genius, then in charge of the Division of Vital Statistics of the 1880 Census, made a suggestion while walking through a room in which hundreds of clerks were bowed over their tables as they hand-tallied items, handling each schedule at least six times. Dr. Billings’ companion was Herman Hollerith, a young engineer who was preparing a report for the Census Bureau. Dr. Billings said to him, “There ought to be some mechanical way of doing this job, something on the principle of the Jacquard loom, perhaps, whereby holes on a card regulate the pattern.” Hollerith thought about the problem and ultimately produced a machine that could punch and sort cards on which information had been entered in code form. First used on a large scale in the census of 1890, the “Hollerith cards” immediately saved $5,000,000, two years in processing time, and set in train a sequence of developments which led to the modern computer.

Hollerith’s invention made possible the tabulation of complicated facts with such refinement that it became possible to identify (for example) a male, white, civil engineer, thirty-two years old, who had four years of college, who lived in Indianapolis (Tract No. 6, Enumeration District No. 1), was born in Pennsylvania, and had lived previously for five years in Ohio. The machine, moreover, could reject suspicious cards, such as the report of a person seven years old who was a physician and a widower.

As plans were made for the twelfth census, there was general appreciation of the fact that the last two had tried to do too much without time for careful preparation, since the organization was always dismantled after each census and had to be put together again every ten years. So this time the questions were greatly restricted in number, and many topics were covered by being transferred to a series of intercensal reports. The 1900 census is rated as one of the best, though one special agent who worked on manufacturing schedules in Philadelphia reported on many intractable problems in trying to decide who was a manufacturer, and in getting hard information from reluctant or naive small operators. He had to interrogate and interpret the answers of wage earners who moonlighted after supper repairing bicycles, housewives who took in sewing in September and October, and the owner of a factory who had to be cornered and asked, “How much would you have paid your sister, if she had been someone else’s sister?” About all that enumerators who recorded business figures could do around the turn of the century was reach an amicable agreement with the respondent on a set of figures that was highly speculative. But they did show reasonable internal relationships, item for item, and thus met “all the requirements of mathematical accuracy and statistical harmony,” if not of absolute truth.