Measuring fertility with the 1901 Canadian census: a critical assessment.

n demography, the most sophisticated measures of fertility are usually based on the registration of births I combined with census data. But where the registration of vital statistics is nonexistent-as in Canada before 1921 'demographers have found other ways of measuring fertility. Censuses are one source that can be used for this purpose, at both the aggregate and household levels. Aggregate measures of fertility-also referred to as macrolevel measures-are the most commonly used to describe historical fertility trends. They are comparatively easy to establish, because they can usually be derived from published material available for specific geographic areas.2 But the analytical power of such measures is limited to associations between fertility indices and the other characteristics of a given territory. The high aggregate fertility in Quebec counties with large proportions of Francophones, for example, tends to support a link between linguistic features and fertility beha~ior.~ But this "ecological" association actually tells us nothing about how Francophone women in various counties actually behaved in comparison with their English-speaking contemporaries. Household-level measures differ from aggregate measures in that they are based on a direct count of each mother's own children. That observation can then be linked to the mother's personal characteristics and to those of the household in which she lived. For obvious reasons, these microlevel research tools provide much greater analytical power. However, they are not particularly common in the literature. Furthermore, when they are used, it is most often with relatively small samples because household-level measures are much more difficult to establish than aggregate fertility indices. Constructing household-level fertility measures involves looking beyond published census tables to the raw data pertaining to each individual and family that were collected by enumerators, in this case almost a century ago. We have recently done this kind of work using the Canadian manuscript census of 1901, both in the context of the Canadian Families Project (CFP) and our separate research project on the Quebec fertility de~ l ine .~ The careful transcription of nominative data from enumerators' lists has allowed us to develop more detailed household-level measures of fertility. It thereby opened new opportunities for testing specific hypotheses about individual reproductive behavior, at a time when fertility rates in Quebec were showing some signs of a decline but when differentials with other parts of Canada (where fertility change had begun earlier) were still substantial. Our focus in this article on the Province of Quebec is of interest from the perspective of ethnic, religious, and linguistic variation. The Province of Quebec-where the quality of parish registers, at least for the Catholic majority, was high-also offers interesting possibilities for the independent testing of the fertility data derived from the 1901 census. It is well established that fertility levels in Quebec remained substantially above those in the rest of Canada until as recently as the 1 9 6 0 ~ ~ Researchers can use various measures to trace the overall pattern of fertility decline in the province (table 1). Crude birthrates show almost no sign of change before the period 1921-1931. More refined indices, however, suggest a very slight decrease in marital fertility toward the end of the nineteenth century, more precisely between 1871 and 1891. The sharper decline that took place between 1921 and 1941 (about 30 percent) was followed by the baby boom in the forties, an interlude that, in the case of Quebec, did not interrupt the downward trend in marital fertility. In Quebec, the higher birthrates of the