The U.S. Constitution mandates a census every 10 years, not for taxation but for governance, apportionment for the House of Representatives.
There are two main parts to the process: counting every resident (citizens and noncitizens) and then tabulating those results.
The year 1890 marked the country’s 11th decennial census, “the broadest and most comprehensive” to date. It was also the first census to use punch cards (aka, punched cards, Hollerith cards or IBM cards) for detailed tabulation.
The business case
In 1880, the Bureau was overwhelmed with data; it took almost a decade to complete the census tabulation.
During the years before the  U.S. census, Hollerith considered many alternatives, including the use of paper tape and cards with holes punched in them to store information. He began conducting experiments during his spare time while teaching for one year at the Massachusetts Institute of Technology (MIT) and then while an assistant examiner in the U.S. Patent Office in Washington, D.C.
In 1884, Hollerith quit the patent office and filed his first patent for “an electromechanical device that could rapidly read information that had been encoded by punching holes on a paper tape or set of cards.” In 1887, he started the Tabulating Machine Company. The patent office granted patents 395781, 395782 and 395783 on 08 January 1889.
Recognizing that they needed a new (faster!) way to tabulate data, the Census Bureau held a competition in 1888 with the goal of awarding a contract for the 1890 census. The data set: 1880 census data from four areas in St Louis, MO.
There were only three participants, Hollerith; William C. Hunt; and Charles F. Pidgin.
- The first test: “capture” the data. The times: 72 hours, 27 minutes; 110, 56; and 144, 25.
- The second test: “prepare data for tabulation” by demographic categories. The times: 5 hours, 28 minutes; 44, 41; and 55, 22.
The winner of each test: Hollerith.
In 1890, the Census Bureau paid 46,804 “enumerators” to visit homes and institutions and document (on paper forms) the “name, age, sex, and color of all persons” that they counted. They began on 02 June 1890.
All that handwritten information had to be converted to information that Hollerith’s machine could tabulate. His innovation: one punch card for each person, where the holes (“punch”) represented data (e.g., male or female; age; Black or White; city and state).
The idea of using punch cards “to control the operation of machinery” dates from the early 19th century. Frenchman Joseph Marie Jacquard developed the first loom that could weave designs into cloth. His attachment used punch cards to create a pattern.
With a device to make punching straightforward, a Census Bureau clerk could create about 500 punch cards per day.
The punched card reader (tabulator) was manually-operated. A clerk placed a punch card between two horizontal plates and then closed them. The upper plate contained spring-loaded metal pins. The pins could pass through the data holes, allowing contact with a mercury bath which completed an electrical circuit.
The completed circuit advanced the pointer on a corresponding data dial. A ringing bell alerted the clerk that relevant dials had been updated. An experienced tabulator could process 80 punch cards per minute.
Although the population of the country increased 25% between 1880 and 1890, Hollerith’s machines took only two and a half years to tabulate demographic results. The total population computation took only six weeks of processing.
Hollerith’s tabulation system saved the government $5 million compared to projections. In today’s dollars, that’s more than $151 million in savings.
Rather than sell his tabulating equipment to the federal government, Hollerith built about 100 of his systems and rented them. His half of the bargain: “maintaining each system in good operating condition.” Renting equipment remained the basic industry business model until the 1950s; it’s still the case for many office machines, such as photocopiers.
The governments of Austria, Canada and Norway also used his system to process their 1891 censuses.
Hollerith took advantage of his monopoly position to raise the rent during the 1900 census. In response, Census Bureau used its own equipment in 1910. Census Bureau employees James Legrand Powers “developed a new electric tabulation machine” that included an automatic feeder and card sorter as well as printing capability. Hollerith unsuccessfully sued the government for patent infringement.
Hollerith recognized the nascent need of vital statistics in the business community. By 1907, his system could assist with business accounting.
His 45-column card would become the industry standard.
The Tabulating Machine Company was one of the three firms that comprised the Computer-Tabulating-Recording Company (CTR) that the “Father of Trusts” Charles R. Flint created in 1911. CTR hired Thomas J. Watson, Sr., as manager in 1914. In 1924 the company changed its name to International Business Machines Corporation (IBM).
Hollerith died in 1929.
In November 2000, the punch cards widely used in US voting machines entered public consciousness via hanging chads, an artifact of an incomplete punch. And yes, it was the same basic technology created more than 100 years earlier.
If you’ve ever reviewed data from the first five censuses (1790-1840), you know that the only recorded name is the head of a household. Other members were listed with a basic demographic accounting. Records from 1850 include names of all members of a household.
The Census Bureau used some form of Hollerith’s punch card tabulation until it installed the UNIVAC in 1951.
Note: According to the Computer History Museum, on 16 August 1890, the U.S. Census Bureau announced that the country had a population of 62,622,250. They do not provide a source for that date, and it is not accurate (archived from the original).
According to the Census Bureau, the last state to report data was Washington, on 21 August 1890. The Bureau announced Pacific coast states (California, Oregon and Washington) population on 20 October 1890.
The earliest Census report that I can find that discusses census results is dated 28 October 1890. This is the same date presented in The Development of Punch Card Tabulation in the Bureau of the Census (January 1965, Government Printing Office).
The figure 62,622,250 does not include “325,464 Indians and other persons in the Indian Territory and on Indian reservations and 32,052 persons in Alaska;” the entire population of the country was 62,979,766.