The connection between divorce rates and women's labour force participation rates in industrialized societies is well established in the literature. What is not clear is the causal order of the connection. Does work outside the home make it easier for women to leave marriage, or do women seek employment as insurance against possible future divorce?
Such a connection may be understood in terms of changes in the functions of the family to individuals and society (Goode, 1977; Wrigley, 1977). Particularly important are structural differentiation and social transformations that have created opportunity structures which allow women to find work outside the home and thus play divergent roles from their traditional ones of home-making and nurturing (Review of Demography, 1989). Such opportunities have expanded especially since World War I and, concomitantly, the family has become less a place where "people (need) each other for survival" to one where members seek emotional gratification from each other, a centre "of nurture and affection, providing individuals with emotional support as they retreat from the achievement-oriented struggles of the outside world" (Beaujot, 1991, p. 20).
Two major difficulties usually confront social scientists wishing to establish causal connections between social phenomena. At the aggregate level one faces the possibility of co-causality or even multicausality among variables of interest. Moreover, an observed correlation may be spurious because the phenomena of interest may be sub-processes of other underlying factor(s). For instance, in Easterlin's relative income hypothesis linking women's employment patterns and the likelihood of divorce in the United States in the 1960s and 1970s, tight economic conditions is the underlying factor accounting for both women's employment and divorce patterns (Cherlin, 1981).
At the individual level the difficulty arises at the point of generalization. In discussing the empirical connection between women's employment and fertility, Bhrolchain (Bhrolchain, 1993) argues that it cannot be assumed that the temporal order of the two events establishes causality. We can make a similar argument with regard to women's employment and divorce. A woman's employment before divorce implies no necessary causality since employment may offer a range of gratifications to a woman which are not financial and which may not be directly relevant to her marital happiness; similarly divorce may be a response to a bad marriage rather than a quest for independence.
In this analysis we examine the causal connection between women's employment and divorce in Canada at both the individual and aggregate levels of measurement. We apply an econometric method - the Granger-Hsiao model - to time series data to identify the form and direction of the causal order between these events and attempt to predict the pattern of change over time. We make no claim to nonspuriousness of the observed correlations at this point because of the difficulty of controlling for possible rival explanations at such an aggregate level of analysis. However, using retrospective survey data, the analysis is extended with statistical controls introduced for demographic, socio-economic, and attitudinal characteristics measured at the individual level with the view to testing whether the patterns of causal order observed at the aggregate level are evident at the individual level.
Theoretical Concerns and Related Studies
Even though the causal connection between female labour force participation and divorce remains undetermined in the literature, two theoretical tendencies are evident. The first tendency attributes the rise in the divorce rate to increases in the numbers of women entering the work force especially since World War I. Summarizing studies done in the United States in 1960s and 1970s, Cherlin (1981) suggested that the increased labour force participation of young married women was the most important stimulus to the initial rise in divorce after 1960. …