From the U.S. Constitution to IBM
The 1890 Census, Hermann Hollerith, and the origins of IBM
Historically there have been two forces which have driven the development
of the electronic computer. The first was the need for fast and accurate scientific
computation. The second was the need for better "data processing" which coincided
with the industrial revolution and the rise of government sponsored social
programs. This paper is an exploration of one facet of the second force where
the rise of IBM can be indirectly traced to the U.S. Constitution and the
Crisis of the 1880 Census.
The Census and the U.S. Constitution
The U.S. Constitution mandated a census for the purpose of apportioning members
to the House of Representatives.
"The actual enumeration shall be made within three years after the first
meeting of the Congress of the United States and within every subsequent term
of 10 years, in such manner as they by law direct" (U.S. Constitution - Article
1, Section 2, Paragraph 3)
A census was required so that the proper number of congressional representatives
could be elected from each state.
The Crisis of the 1880 Census
During the late 19th century the population of the United States rose dramatically.
Year Census Population
Unfortunately, the 1880 Census took 9 years to complete at a cost of 5.8
million dollars! There was concern that the 1890 census would not be completed
before the 1900 Census.
1850 23.191 million
1860 31.442 million
1870 38.558 million
1880 50.198 million <- 9 years to complete census!
1890 62.979 million
1900 76.211 million
Taking the Census required a number of steps, all of them labor intensive.
However, the problem was not in the collection of the data; it was with the
processing of the data.
Born in Buffalo N.Y and educated at Columbia University School of Mines (1879)
Hermann Hollerith worked on the 1880 Census where he met Dr. John Shaw Billings,
head of the division of vital statistics. During a dinner conversation Billings
casually suggested to Hollerith the idea of mechanically tabulating Census
results using cards with punches on the edges. Hollerith thought he could
work the details for such a solution but Billings indicated no interest in
pursuing the idea further.
- Raw Data for each household was taken by Census enumerators. This was
written down on large sheets of paper called "schedules".
- The schedules were sent to Washington where the raw data was manually
transferred to "tally sheets" using "slashes" (5 to a box) to mark different
categories. There were six major tallies to complete. For example the first
broke down the population by sex, race, and birthplace.
- Finally the tally sheets were transferred to consolidation sheets
whose figures were combined to yield final population totals.
Hollerith first worked on an electro-mechanical paper tape device but problems
with the design caused him to abandon this approach (1882). He then switched
to the idea of using a Punch Card system which used a 3/16 inch hole located
in 1/4 inch squares to record information. The presence or absence of a hole
would indicate the existence or non-existence of some characteristic or item
Hollerith claimed he got this idea from "punch photograph cards" used by
rail road officials. Used to prevent the theft of railroad tickets from passengers,
conductors would "record" the physical characteristics of the ticket owner
(e.g. eye color, hair color) by punching specially marked areas on the edge
of the card. Hollerith used holes punched through the card, not on the edge
of the card.
Hollerith also invented equipment to punch, read, and sort punch
cards. Specifically he invented
machines were ready for 1890 Census
- a manual "pantograph" punch to transfer data to cards
- an electo-mechanical tabulator device which read a card. This device
employed an array of pins which when pressed down on the card went through
the holes to make contact with small cups of mercury underneath. This completed
a circuit which registered on a dial.
- a sorter which used the same pin array /mercury cup mechanism to open
a chute into which the card was manually placed.
Hollerith's Tabulator Machines and the 1890 Census
After first beating out two other systems in a contest arranged by Census bureau,
Hollerith rented his tabulator machines to the U.S. Census Bureau for the
1890 Census. Work began in July of 1890 and by August 16, an "unofficial tally"
of the United States population was obtained (officially announced in October).
Since Census clerks were able to handle between 7000 and 8000 cards per day
the greater speed of processing data permitted a more comprehensive census.
In the end, the 1890 Census was completed in 7 years at a cost of 11.5
million dollars. While this might seem like a long time (and it certainly
cost more money), it should be noted that a more complete census was
actually done. Even
in the 1890's there was a temptation to use "technology" to the limit.
Hollerith was paid $750,000 for the rent of his machines
From The Tabulating Machine Company to IBM
Hollerith's company, the Tabulating Machine Company, rented out machines
to other customers both government (Austria, Canada, France, Russia) and private
(New York Central RR, Marshall Fields, Penn Steel). The company was profitable
but due to its policy of renting instead of selling machines, it had cash
flow problems. The Tabulating Machine Company merged in 1911 with 3 other
companies, International Time Recording Co, Bundy Manufacturing, and Computing
Scale of America to become Computing Tabulating Recording Company. In 1924
its name was changed to International Business Machines - IBM.
Information for this Web page was obtained in part from Bit by Bit:
An Illustrated History of Computers by Stan Augarten; Tichnor & Fields;
1984. This is an excellent book and well worth reading for more information
on the history of computers.
Return to Math/CS
© 2007 Wittenberg University