◷ Computing history

Herman Hollerith and the 1890 census: data processing is born

How Herman Hollerith's tabulating machines processed the 1890 US census using punched cards, and how his company later became one of the roots of IBM.

If Jacquard had used punched cards to control a machine and Babbage to program one, Herman Hollerith used them for a different and very modern task: counting and classifying large amounts of data. His system, adopted for the 1890 United States census, is considered the beginning of automatic data processing and, through a series of corporate mergers, lies at the origin of IBM.

A problem of scale

The United States Constitution requires a census every ten years, because the apportionment of seats in the House of Representatives depends on the population of the states. As the country grew, the work became ever heavier.

The 1880 census was processed almost entirely by hand: clerks read the questionnaires and made tally marks on large counting sheets. The count took several years. It was clear that, with a fast-growing population, the next census risked not being finished before the one after it was due. A new method was needed.

Hollerith's idea

Hollerith, a young engineer who had worked for the Census Office, developed an electromechanical system during the 1880s. He patented it in 1889 and won the competition to process the 1890 census.

The three parts of the system

  1. The punch. An operator translated each person's answers into holes in a card. Each position had a meaning: sex, age, marital status, place of birth and so on.
  2. The tabulator. The card was placed in a press. A set of metal pins came down onto the card: where a pin found a hole, it passed through and touched a small cup of mercury underneath, closing an electrical circuit. Each closed circuit advanced a dial counter by one.
  3. The sorting box. The same electrical pulse could open the lid of a compartment, into which the operator placed the card. In this way the cards were divided by category, ready for further counts.

A concrete example

Suppose we want to know how many foreign-born married women live in a county. With the manual method, every questionnaire had to be read again. With Hollerith's system, you could first sort the cards by sex, then run only the women's cards for marital status, and so on. Or the circuits could be wired so that a counter advanced only when several holes were present at the same time. That is, in electromechanical form, a logical AND condition.

The result

The 1890 census was processed much faster than the previous one, with a considerable saving in manual labour. The system also made it possible to produce statistical cross-tabulations that would previously have been too expensive to compute. The success was such that Hollerith's machines were adopted for censuses in other countries and for commercial applications, such as railway company accounting.

From tabulators to IBM

The corporate history is straightforward and well documented:

Year Event
1896 Hollerith founds the Tabulating Machine Company
1911 The company merges with others to form the Computing-Tabulating-Recording Company (CTR)
1924 Under the leadership of Thomas J. Watson, CTR takes the name International Business Machines (IBM)

For decades punched-card machines were the heart of data processing in businesses and government. When electronic computers arrived, cards remained the main input medium for programs and data until the 1970s.

Why it is a decisive milestone

Hollerith's system was not a computer. It did not run programs and did not perform general calculations: it counted, added and sorted. Yet it introduced concepts that anyone who works with data will recognise immediately.

  • The record. Each card represented one person, with fields in fixed positions. It is the concept of a table row.
  • Encoding. Information was translated into a standard, machine-readable format, separating the data from its original medium.
  • Queries. Counting the cards that meet certain conditions is, in essence, what a query with a filter clause and a count does.
  • Grouping. Sorting cards by category and counting them anticipates the grouping operations of modern databases.

In SQL, the question from the earlier example would be written like this:

SELECT COUNT(*)
FROM census
WHERE sex = 'F'
  AND marital_status = 'married'
  AND foreign_born = 1;

More than a century separates the two technologies, but the logic is identical.

Something not to forget

The history of tabulating machines also reminds us that the ability to classify people on a large scale is a powerful tool, which can serve to distribute services better as well as for purposes of control. It is one of the reasons why the protection of personal data is today considered an integral part of designing any computer system.

In short

  • The 1880 US census, processed by hand, had taken several years of work.
  • Hollerith devised a system of punch, electric tabulator with mercury contacts and sorting box, used for the 1890 census.
  • His Tabulating Machine Company became part of CTR in 1911, which in 1924 became IBM.
  • Records, encoding, filters and grouping, concepts at the core of databases, are already present in his system.
MM
Marco Marzaduri

Marco Marzaduri works in cybersecurity, digital forensics and web development. With degrees in computer science and psychology, he combines technical analysis with an understanding of the human factor.

Read also