Herman Hollerith in 1884, at age 24. Prints and Photographs Division, Library of Congress.

The United States census wasn’t always the massive operation it is now. In March 1899, census director William Rush Merriam took office to find little more than “an old typewriter, much out of repair, a horse of doubtful age, a wagon practically useless, an old cart,” and stacks of undistributed reports. He later demanded the census office stop dissolving itself as soon as its work was complete, comparing the repeated closures to a textile manufacturer who demolished his mill, dismissed its staff, and destroyed all its records because of a temporary surplus of cloth.[1] A permanent census office was finally formed in 1902.

As the U.S. Census Bureau solidified into a permanent institution, technology also lent it power. The Library’s Manuscript Division holds the papers of Herman Hollerith, an inventor who helped expand the influence of the census at a moment when the government’s thirst for information was exploding. His inventions ultimately transformed global data processing and led to the founding of the International Business Machines Corporation (IBM). Hollerith’s papers offer a window into an unexpected story about war, governance, and the power of data: one that unfolded in the Philippines, more than 8,000 miles from Washington, D.C.

The first U.S. census was mostly conducted on horseback by officials who collected basic household-level data on around 4 million people and published their results in a slim 56-page report. In 1850, census takers began surveying individuals. By 1870, they were gathering information on nearly 40 million people, including basic demographic data and categories like literacy, voting status, value of real estate holdings, and whether one’s parents were foreign born.[2] Not surprisingly, errors and processing time mounted. Tabulations for the 1880 census took eight full years to complete.

This lightweight celluloid panel, which appears to be coded for the 1890 census, enabled the user of a pantograph punch to easily insert a stylus into the correct hole while the device punched a card at the back.

Hollerith’s inventions included a suite of punch card-based devices designed to speed processing and increase accuracy.  They were first used by the Bureau for the 1890 census, and together reduced tabulation time to just 2½ years. But that increased processing power also allowed the government to sort its data in new ways and to see Americans in ways they may not have anticipated. Discovering what percentage of African Americans were divorced or how many female foreign-born residents were juvenile delinquents just hadn’t been practical before. Suddenly it was.

The mechanization of data processing in Washington coincided with America’s global expansion, and with its territorial acquisitions in Puerto Rico, Cuba, Guam, and the Philippines during the Spanish-American War. In the Philippines, where America’s war against Spain had evolved into a second war against a new Philippine Republic, the creation of a new Census Bureau was made contingent on the end of hostilities. Therefore the launch of that census, which eventually produced the massive, four-volume Census of the Philippine Islands, became a way of declaring victory. That report then systematically counted, tabulated, and classified Filipinos in entirely new ways, and helped reveal the archipelago and its people to a new government.[3]

An unidentified coder uses a pantograph punch to record data. Prints and Photographs Division, Library of Congress.

What did that Philippine census look like? On the surface, not too different from the ones being conducted in the United States. Punch cards for the 1900 U.S. census were divided into regions that signified categories like geographic location, marital status, parents’ birthplaces, or race. Cards for the 1903 Philippine census also included categories like “name of tribe” and “material of house.”

The Philippine census cards were even processed in the same place, and Hollerith remained personally involved. Once census takers had reached as far into the archipelago as they could go, the seven million cards they punched were shipped off to Washington via San Francisco, finally coming to rest at the U.S. Census Bureau.[4] Once there, Hollerith went “out of my way to devise new connections, and furnish machines” with new counters and relays, to expedite tabulation.[5]

All that data was used to sort Filipinos into roughly 25 linguistic groups and five skin colors, among other categories. They were then sorted into just two broad groups: “civilized” and “wild,” terms the census used interchangeably with Christian and non-Christian. That binary was more nuanced than it appeared. Many Americans argued that the non-Christian Filipinos Spain had failed to conquer would be more open to U.S. influence, and were skeptical of the ties between Filipino Catholics and a so-called “corrupt” Spanish state.

A diagram of the punch cards used for the 1903 Philippine Census. Census of the Philippine Islands, vol. 2. (Washington: Government Printing Office, 1905), 12.

Even so, the categories dramatically oversimplified the nation’s ethnic and linguistic diversity. In part, because a punch card can only record a simple yes or no answer. Censuses conducted by punch card therefore helped intensify what media scholar Seb Franklin calls a “compression of social reality,” in which the binary logic of a machine actually hardens existing social categories, or creates new ones.[6]

Punch cards have been so dramatically superseded by newer computing devices that it can be difficult to image the revolutionary power they once had. But early punch cards were powerful, and at an important historical moment. In the Philippines, punch cards helped establish a territorial government that retained power, in some form or another, until World War II. In the United States, early census punch cards exercised their own kind of power. By the 1920s, Congress was circling back to the data generated by the 1890 census to create quotas that cut immigration by 80%, and mostly limited admissions to Northern and Western Europe. The nation maintained those quotas for more than four decades.

So maybe a punch card doesn’t look much like the “big data” of today. But if we look, we can find its influence in all sorts of unexpected places.

[1] William R. Merriam. “Need of a Permanent Census Office.” The North American Review 174, no. 542 (January 1902), 105, 107.

[2] Keith S. Reid-Green. “The History of Census Tabulation.” Scientific American 260, no. 2 (February 1989), 98.

[3] Vicente Rafael. White Love and Other Events in Filipino History. (Durham: Duke University Press, 2000), 25.

[4] “The Philippine Census: Difficult Work Accomplished by the American Officials with Gratifying Results.” Barbour County Index, November 4, 1903, 3.

[5] Herman Hollerith to Samuel G. Metcalf, September 20, 1904, Box 10, Folder 2, 2, Herman Hollerith Papers, Manuscript Division, Library of Congress, Washington, DC.

[6] Rafael, White Love, 32-33; Seb Franklin. Control: Digitality as Cultural Logic. (Cambridge: The MIT Press, 2015), 28.