Dr. Gordon Darroch was happy to fulfill some of the Canadian Census Research Infrastructure’s outreach mandate at a seminar at the University of Guelph this morning, but his main purpose was to discuss his experience with the 1871 Census. His work with this document in the 1970s created a sample of the first Census of the Dominion of Canada and spearheaded most of the other national microdata sampling projects that have cumulated with recent initiatives such as the CCRI/IRCS and the 1891 Census Project at Guelph.

He was trained as a sociologist and began to use the census around the same time as the ‘new social historians’ were reconsidering the value of this and other routinely generated sources. It was a time when historians were fascinated with the primary document, the ordinary anonymous ‘folk,’ and the bottom-up approach that came to define a discipline. A seminal work, according to Dr. Darroch, was Stephen Thernstrom’s Poverty and Progress (1969) which used censuses to study a small case study of labourers in a 19th century American town. It was considered a sort of academic wizardry in its time, a meticulous ante-computer tale of people’s fates.
With equally arduous research and nascent punch-card technology, Canadian scholars began to uncover the voices of ordinary people in places like Peel County (David Gagan) Saguenay (Gérard Bouchard) and Hamilton (Michael Katz). Their findings revealed, among other things, that Canadians were surprisingly transient, and they attacked the persistent image that the past was more stable than the present. Dr. Darroch wrestled with the major limitation of these case studies, their local or regional focus, and sought to use the census to generalize more broadly across the country.
His talk this morning dealt with some of the problems of creating a national microdata sample, especially attracting funding which was surprisingly difficult, even when the Canada Council and SSHRC funding agencies were created. But there was also the major problem of deciding how to form a sample and how to make it representative. He found that linking people between census years was impossible with a randomly sampled dataset. Choosing all the people with surnames beginning in T and R was actually very representative as French researchers found in the “TRA sample” of censuses in France. An even better approach was to sample names in certain phonetic pockets and compare them for accuracy to the 1881 census, fully indexed by the Church of the Latter Day Saints. Some groups were oversampled such as Atlantic Canada, urban areas, and certain ethnic groups, and the final project represented 60,000 records, or almost 2 percent of the total Canadian population in 1871. A final challenge was to make the sample usable, but because funding was inadequate for full documentation many potential scholars have been dissuaded and Dr. Darroch feels that more use could be made out of the dataset.