Case Files: Herman Hollerith

Introduction

A Chief Special Agent to the Census Bureau as well as an instructor in mechanical engineering at the Massachusetts Institute of Technology, Herman Hollerith sought to devise a method to process the expanding population data for the 1890 U.S. census. His observation of a railroad conductor punching riders' tickets contributed a key idea to Hollerith's system.

Who was Herman Hollerith, and what was his Electric Tabulating Device? How did it work to process information?

An Impressive Student

Herman Hollerith, the son of German immigrants, was born on February 29, 1860, in Buffalo, New York. Not suited to conventional schooling, Hollerith left at the age of nine and was privately tutored until entering the City College of New York in 1875, then graduating with distinction in engineering from the Columbia School of Mines in 1879. He impressed his instructor, Professor W. P. Trowbridge, who immediately hired him as an assistant, and continued to employ him as a statistician when he moved to become Chief Special Agent to the Census Bureau. There, analyzing the 1880 data, Hollerith discovered the inefficiencies and errors in information collection and handling.

Breaking Point

In 1882, at the age of 22, Hollerith moved from the Census Bureau to become an instructor in mechanical engineering at the Massachusetts Institute of Technology (MIT), where he continued to think about the problem posed at the Census Bureau—how to devise a mechanical method to process the rapidly expanding data from the next U.S. census.

Improvements were needed to handle the wider question list and expanding population (over 60 million) anticipated. By June 1, 1890, the population count and distribution would reflect the fourfold effect of industrialization, Civil War survivors, westward migration, and immigration, in addition to the larger population. The manual system was close to breaking point.

Two incidents contributed to Hollerith's solution: conversations with Census Bureau colleague, Dr. John Shaw Billings, about count mechanization and the Jacquard loom card system, and observations of a railroad conductor punching riders' tickets for identification purposes.

The Jacquard loom uses a series of punched pasteboard cards to mechanically control the warp threads in a loom against the weft threads, automatically creating duplicated patterns in the finished fabric. A railroad conductor punches holes around the edge of each passenger's ticket to signify details of the travel and traveler.

A Tabulating System

Through experimentation, Hollerith discarded perforated paper tape input and pneumatic signal handling. He worked on a system in which mechanical counting and sorting of cards by detection of a certain punched hole location and electrically driven signals would yield the required statistics from a very large collection of information.

Hollerith left MIT in 1884 to work at the U.S. Patent Office and on September 29 of that year patented his first invention: a "method, system, and apparatus for compiling statistics", otherwise known as the Hollerith Electric Tabulating System. 30 U.S. patents and many foreign patents on improved systems would follow.

Competition

Hollerith set up his Tabulating Machine Company in 1886 and the Hollerith System was successfully tested on mortality data for insurance companies and freight bills for railroads in New York and Baltimore in 1887. It was entered in a Census Bureau competition against two others to determine the system to be used for the 1890 census. Other contestants were statisticians Charles F. Pidgin and William C. Hunt, whose systems involved transcription of the data to colored slips or cards respectively. The Hollerith System won and production of punching machines and card sorters was set in motion.

Twenty Months

On June 1, 1890, the census count was made and the data began to arrive for counting in September. On December 12, processing was completed on the United States population which was measured at 62,622,250. Some 20 months had been cut from the time that a manual count would have taken.

On September 15, in the midst of this landmark achievement, Herman Hollerith married Lucia Beverley Talcott, the daughter of a mining engineer working in Vera Cruz, Mexico. Their family grew to include three sons and three daughters. In the same year, he was awarded a Ph.D. by Columbia University for his thesis on the Tabulating System.

The tabulating system won worldwide acceptance and use in census counts throughout the world.

High Price

The 1900 U.S. census used leased Hollerith equipment and cards, at a noticeably high price since the company now dominated the field. The cost prompted the Census Bureau (made a permanent division of the Department of Commerce in 1903) to cancel the Hollerith arrangement and use the slower Pidgin equipment while developing alternative equipment in time for the 1910 count.

A Census Bureau engineer, James Powers, patented the resulting machine that won most of the 1910 census work. In 1911, he formed the Powers Tabulating Machine Company, an effective competitor to Hollerith's company.

At that time, Hollerith's Tabulating Company was involved in various disputes over patent infringement, which led to lawsuits and final judgment by the Supreme Court. Hollerith's challenges were unsuccessful.

Merger

The fortunes of the Tabulating Machine Company faded and it widened its range by merging with the Computing Scale Company of America and the International Time Recording Company to form the Computer Tabulating Recording Company (CTR). Herman Hollerith continued as a consultant with the company. His firm business approach had always been to let the company's achievements speak for themselves. Never employing a sales force, he had succeeded on recommendations from satisfied customers to power sales expansion.

Thomas J. Watson, an accomplished salesman and manager, came into the company in 1918 from National Cash Register. CTR's leadership in the market was restored through product research and improvements including the introduction of printers for input and output. Hollerith took little interest in the changes and retired in 1921. CTR's name was changed in 1924 to International Business Machines Corporation (IBM).

The Powers company also went through a procession of mergers, becoming Remington Rand, Inc. in 1927 and much later, after multiple mergers, reaching the company now known as Unisys.

Herman Hollerith retired to his farm on Chesapeake Bay and died of a heart attack on November 17, 1929. He is buried in the Oak Hill Cemetery in Georgetown, Virginia.

Stated Aims

When given the task of tallying and sorting large amounts of information, individuals will devise a system which involves an original design stage (counting hierarchy), a repetitive operation stage (actual counting) and a final "results gathering in" (register of tally) stage.

Herman Hollerith's stated aims in creating his tabulating system were to "reduce labor, increase rapidity of tallies and insure accuracy." He accomplished this by providing an unambiguous input profile, mechanizing the repetitive sorting stage, and including an indicator of output tally results.

After unsuccessful experiments with punched paper tape inputs, Hollerith fixed on a punched card system, echoing their previous similar use in Jacquard looms and Babbage's imagined analytical engine. The manila cards were durable and non-conductors of electricity.

Punch Cards

For 1890 census data, an operator, using a pantograph, transcribed information items dictated from report sheets to punch holes in specified row locations on a card—one card for each person. The card contained 40 columns.

An operator then slipped each card into the tabulating machine, pressed down the detector pins which then "read" the card and registered an increment on a corresponding dial for each hole detected. Thus, 40 characteristics were measured in a single pass of the card.

Mercury Cups

The mechanical/electrical detection device for each column consisted of a hard rubber bedplate containing an arrangement of small cups of mercury which coincided with each possible hole location on the card. After placing the card over this array the operator pulled down a "pin-box" of spring-loaded pins with exactly the same arrangement as the mercury cups. Pins which passed through a hole dipped into the mercury cups and completed the circuit to an electromagnet, which advanced the corresponding counter dial by one unit. The operator hand copied the tabulation totals and reset the dials to zero at the end of each run.

Improvements

After a card had been processed, a slot in the sorting box was popped open by electrical signal and the operator dropped the card into the slot. The card becomes the permanent record for an individual. The 1890 census measured the characteristics of over 65 million people.

Improvements in later models included a card punch machine built along the lines of a typewriter, an adjustable wiring feature to enable counts of column combinations, an automatic card feeder, and connection to adding machines to add multi-digit numbers. A redesigned card of 80 columns and 14 rows extended the recording capability to the alphabet plus certain punctuation marks. A printer of card data improved the distribution of information.

Hollerith's card sorter and counter remained in use by businesses through the 1960s. An operator with ingenuity could manipulate column contents and sorting sequences to produce sophisticated statistical models.

Acknowledgement

The 1890 Elliott Cresson Medal was awarded by The Franklin Institute to Herman Hollerith for his Electric Tabulating Device.

Other awards to Hollerith include the 1893 Medaile D'Or of the Paris Exhibition and the 1893 Bronze Medal of the World's Fair. The British Royal Statistical Society also established the annual Hollerith Award.