FAMILY AND SOCIODEMOGRAPHIC INFLUENCES ON PATTERNS OF LEAVING HOME IN POSTWAR BRITAIN- MIKE MURPHY AND DUOLAOWANG We identify child-level and parent-level characteristics asso- ciated with children s patterns of leaving home. We use a multi- level discrete-time hazards model to examine the impact offamily and demographic factors at both levels, and utilize the Alternating Conditional Expectation algorithm optimally to transform the de- pendent and independent variables. We find that measured vari- ables at both the child and the parent level have important influ- ences, as do period and cohort factors. However, unmeasured par- ent-level factors have an influence on the departure ofchildren that is broadly similar in magnitude to measured factors. Children'S leaving home is an important event for chil- dren and their parents. For children, it is as a key stage in the transition from childhood to adulthood and it is often associ- ated with other major life-course events, such as forming a partnership and entering the workforce or higher education. For most parents, the event of last child leaving home marks their entry into the "empty-nest" stage, with consequent ad- justments in their life styles (White and Edwards 1990). Leaving home is therefore an aspect of intergenerational re- lations in which the interests of both the parents and the chil- dren are important. Whether a child leaves the parents' home will depend on (1) the relative preferences of both parties for joint rather than separate living arrangements; and (2) the feasibility of establishing separate households, which is influenced by wider socioeconomic factors such as current and likely fu- ture economic prospects. Although researchers have noted the importance of the joint nature of the decision (Gold- scheider and Goldscheider 1993; Young 1989), analyses of leaving home have usually been approached from the per- spective of either the child or the parent: The former ap- proach tends to be concerned with factors associated with departure, especially the characteristics of the child, and the latter approach focuses on the consequences of the event for 'Mike Murphy and Duolao Wang, Population Studies, London School of Economics, Houghton Street, London WC2A 2AE, United Kingdom; e- mail: M.Murphy@lse.ac.uk. Thanks are due to the Economic and Social Re- search Council, which partially funded this work as part of a project, Formal Modelling of Household, Families and Kinship (Grant L315 253 017). The data used from the British Household Panel Study were made available through The Data Archive. They were originally collected by the Economic and Social Research Council Centre on Micro-Social Change at the Univer- sity of Essex. Neither the original collectors of the data nor the archive bear any responsibility for the analyses or interpretation presented here. We grate- fully acknowledge the valuable comments of anonymous referees. Demography, Volume 3S-Number 3, August 1998: 293-30S the parents. This dualism has been identified as a major weakness of studies in this area (Aquilino 1990). The historical importance of societal norms about the patterns of living arrangements of parents and children has been emphasized (Hajnal 1965), but these patterns are sensi- tive to the current economic situation. In contemporary soci- eties, although poorer economic prospects may inhibit home- leaving, the increased tension that young unemployed people experience in the family may raise the emotional costs to the extent that they leave home earlier than those who are em- ployed (Murphy and Sullivan 1986). Thus, leaving home is influenced by economic, cultural, institutional, and demo- graphic factors that act at the macro level, the family level, and the child level. Empirical findings using such variables are discussed in the next section. Some relevant variables may be difficult or impossible to collect in large-scale surveys, and they will generally be subsumed into the residual term. Examples of unmeasured family-level effects are conflict or abuse in the family, intergenerational continuities in behavior, and imita- tive behavior (e.g., mimicking the home-leaving of a sib- ling). Such factors will lead to clustering of home-leaving behavior within particular families. Event-history data are required to model home-leaving. One approach is to collect information from the child, but typically with information available on only one child per family.An alternative is to collect data on the home-leaving of respondents' children, including information on the patterns of departure of all siblings. In the latter case, the data are at two levels-ehildren (at the first level) within fathers or moth- ers (at the second level)-permitting us to investigate tenden- cies for particular patterns of leaving home to cluster within families. We employ a multilevel discrete-time hazards model to estimate both measured and unmeasured effects by adding random effects that vary across different levels of such data (Allison 1987; Davies, Elias, and Penn 1992; Steele, Dia- mond, and Wang 1996). This random effect not only estimates the magnitude of the correlation between the home-leaving of children in the same family but also controls for unobserved factors that may affect a child's risk of leaving home. REVIEW OF PREVIOUS RESEARCH AND HYPOTHESES Family Size and Birth Order Large family size has generally been found to be associated with earlier home-leaving (Bianchi 1987; Goldscheider 293