Constructing census families and classifying households: "relationship to head of family or household" in the 1901 census of Canada.

In this article, I first consider some interpretive issues regarding the central place of the notion of “family” in the construction of the 1901 Canadian census. I then discuss the meanings of the key variable “relationship to head of family or household” and some implications of the conceptualization and classification of dwellings, families, and households in the Canadian Families Project (CFP). Following international census practices from the middle of the nineteenth century, the Canadian population was nominatively recorded in the context of family dwellings or normal places of abode. Before 1901, the central organizing notion of the census was the “census family,” considered to be any person living alone or any number of persons living together under one roof and having their food provided together. “Census families” were first defined in the first national census of 187 1 and carried over to 189 1, although the general idea was more or less taken for granted in the nominative colonial censuses of 1851 and 1861. In 1901, however, rather different and more specific definitions were introduced. Presumably, the seemingly obvious and “natural” allocations of individuals to dwellings and families in the earlier censuses had been deemed inappropriate or problematic. No clear rationale for the changes exists, either in the instructions to enumerators or in surviving archival evidence. Clearly, however, the new definitions in 1901 aimed to make explicit the classification of the Canadian population in three ways-by dwelling units, by households, and by families. Dwelling houses were numbered in “order of visitation” in the first columns of Schedule 1 of the enumeration.’ The instructions to enumerators went on to distinguish families from households, referring to the former in a quite restricted sense, as consisting of parents and sons and daughters “. . . united in a living and housekeeping community. . . .” But the instructions allowed that families could include relatives and servants. Households, in contrast, included “. . . all persons in a housekeeping community, whether related by ties of blood or not, but usually with one of their number occupying the position of head.” Thus, the 1901 census definitions corresponded more or less to the distinctions often made in family history between forms of “extension” and “augmentation,” with the exception that servants were to be included among extended family members. Finally, the new census definitions classified single persons living alone and carrying on their own housekeeping as “households,” whereas they had been included as a type of “census family” in prior censuses.2 Beyond these distinctions, a key variable was included, the putative relationship of each member of a family or household to its head. In the CFP’s 5 percent national sample, there are a total of 332 different recorded entries of such relationships, including single entries for missing and illegible ones.3 Many of these entries are simple variations on others: for example, “head,” “chief,” “head widow,” “widow head,” and so on. Among those of substantive interest, the entries ranged from the obvious, common ones of heads and wives and sons and daughters of heads, through stepchildren and grandchildren and relatives, to a very diverse array of visitors, boarders and lodgers, domestic and “nondomestic” employees, and institutional inmates. A basic principle of the project has been to transcribe this full array to the file in alphanumeric form and to assign separate numeric codes to the individual mentions.“ Thus a researcher can, in principle, begin with the original enumeration and create a coding scheme tailored to any analytic purpose. I will comment further on this prospect and the implications for analysis of a common coding scheme for dwelling units based on the relationship to head that the project adopted. First, a more general consideration of the meaning of the construction of census families and households is in order. In her current research for the CFP, Annalee Golz (1998) reflects on the implications of the changes and specifica-