or How the Gendered Wage Gap and Child Labour Killed the British Computer Industry before it Even Started*
All I ever want to do these days is talk about late-Victorian offices (much to the chagrin of my partner and students, I’m sure), but I have already laid bare my soul on the subject of vertical files and press books on this blog, so thought I might stray a bit into a frank eye-to-eye chat about the American influence on late 19thC British census taking and the history of computing (it’s these frank eye-to-eye chats on the history punchcards, ink wells, and clerk’s stools that, I think, try my students so much, but bear with me. So far they have shown me great latitude and patience and I hope you will too. I will let you know in future posts if they ever reach their breaking point). I have a keen interest in the social history of computing, and get real pleasure from ferreting out the points where computing might have taken a left turn or a right one, giving us some other version of the colonial, gendered, or racialized state of computing that we live with today.
One of these pinch points was the development and commercialization of census taking technology. The British were census-takers extraordinaire. The Royal Statistical Society was formed in 1834 and they had a centralized General Register Office by 1837, led by novelist Thomas Henry Lister, who, alongside statistician William Farr, guided the 1841 census. The British continued to use Farr’s labour intensive system until 1911 when the British census was mechanized. The mechanization process was not a British affair (Campbell-Kelly). In 1894 former superintendent of the US Census Office, Robert P. Porter and his former employee Herman Hollerith, had read a paper, “The Electrical Tabulating Machine” before the Royal Statistical Society, outlining how the Americans had cut the time it took to process their census data by three quarters through mechanization. In 1911 British census taking was mechanized by Porter who by then was running a British affiliate of Hollerith’s Tabulating Machine Company in London. Why weren’t the British the first to mechanize census taking and tabulation? Why did they lag behind the Americans by more than 20 years?
The American 1880 census had taken almost a decade to process. Recognizing that the 1890 census would likely be more onerous than the 1880 census, the US Census Office put out a call for proposals to streamline their data collecting and processing procedures. Three schemes were put forward. The first two used coloured paper or coloured ink to make data recognition easier for the clerks who would have to read them. Herman Hollerith, a Census Office clerk at the time and great fan of William Farr, proposed what would be the winning design: a punchcard that could be processed by a purpose-built machine. Each card was the size of a dollar bill, and contained 288 possible punch points to record age, marital status, employment status, profession etc. Once punched, the cards were read by small electrified pins in Hollerith’s patent machine. When a pin passed through a hole in the card it closed a circuit, triggering the machine to sort the card into a specific compartment for counting. The cards could be sorted and resorted for tabulation again and again, to find, for example, how many married and unmarried woman lived in Ohio, or how many farmers rented rather than owned their farms below the Mason-Dixon line, all based on whichever closed circuits the machine was set to respond to. The data from a 1880 census had taken nearly a decade to process; the data from the 1890 census was processed so rapidly that the Census Office started to release national population statistics after six weeks.
Hollerith left the Census Office and started the Tabulating Machine Company in 1896, which provided cards and machinery for the 1900 census. The Tabulating Machine Company was one of the three companies that merged to form Computing-Tabulating-Recording Company in 1911, later renamed International Business Machines (IBM). In 1902, in exchange for a 25% royalty fee Robert Porter bought the rights to sell Hollerith’s machines in England, where no similar company existed, calling his company The Tabulator Limited.
Later the company was renamed the British Tabulating Machine Company, and is most famous now, perhaps, for the bombes machines they made to decrypt Engima messages in World War II. I say “most famous now” because we’ve just come through one of our Great Man years (doesn’t everyone miss the Charles Darwin’s bicentennial birthday year?) in celebration of what would have been Alan Turing’s 100th year. If the bombe machines at Bletchley Park hadn’t been destroyed, some have argued, England might have been the country famous for commercial and consumer computing. There is, I’d say, a greater mystery here –even if the bombes had become the basis for a nascent computing industry, the parent of the company that made them was American after all– why didn’t the British pull ahead in the data processing game in the 1890s? They had a well-established statistical society after all. But by 1910, while Austria had 10 tabulating machines and France had 25 (with Germany and Denmark boasting 9 and 1 respectively by 1914), Britain had a scant 4 tabulating machines (Knight 597).
As it turns out, it wasn’t a lack of mechanical or statistical aptitude that kept the British back. Martin Campbell-Kelly argues that there was resistance to American “excesses” — the American census comprised more information and cost almost 20 times as much as the American one. Campbell-Kelly doesn’t stress the next point, which arises from his excellent study of British taking, but it is a point that invites further investigation: the General Register Office didn’t need to mechanize to save money. The Farr method hadn’t changed since 1841, but the cost per capita of the British census went down between 1871 and 1901, thanks not to mechanization, but because of the General Registrar’s Office’s willingness to hire cheap child and female labour. Office boys were particularly inexpensive, since, even though they often had to pass entrance exams to get their posts, they could remain unpaid for weeks as they trained for their positions. in the 1870s and 1880s junior clerks made between £35 and £45 per year —which, as the late Sally Mitchell pointed out, was less than a working-class men’s average income— and women office workers made as little as £25 (68). Which leaves me to ask, why didn’t the Americans use the same strategy to keep their census-processing costs down. Was their talent pool too small? Were the “excesses” of their data collection and processing too intricate for women and child workers? Were they just a machine-obsessed country?
For more, see
Campbell-Kelly, Martin. “Information Technology and Organizational Change in the British Census, 1801-1911.” Information Systems Research 7.1 (1996): 22–36. Print.
Hollerith, Herman. “The Electrical Tabulating Machine.” Journal of the Royal Statistical Society 57.4 (1894): 678–689. JSTOR. Web.
Knight, F. H. “Mechanical Devices in European Statistical Work.” Publications of the American Statistical Association 14.110 (1915): 596–599. Print.
Mitchell, Sally. Daily Life in Victorian England. Westport, Conn: Greenwood P, 1996. Print.
* I love a good hyperbolic subtitle