The goal of this project is to analyze two longitudinal data sets--one from the U.S. and one from Sweden--using integrated multistate models that control for time varying variables and unobserved heterogeneity to systematically investigate the socioeconomic determinants of the timing and spacing of births over the life cycle. The two longitudinal data sets we plan to analyze are: (1) the Panel Study of Income Dynamics (PSID) collected by the Survey Research Center at the University of Michigan and (2) the Survey of Family and Work of Swedish Women (BAS81)--a survey collected by Statistics Sweden in 1981. Each data set is contains unusually rich information regarding the evolution of fertility, education, labor force participation, marriage and home leaving decisions over the life cycle and both provide a rich set of information about the background of the interviewees. Using these data we shall use multistate tools to investigate dynamic timing and spacing phenomena. A major goal of our research is to uncover stylized facts about such behavior using a variety of methods to examine the robustness of these """"""""facts"""""""" to alternative estimation strategies that account for time varying variables and unobservables. By pinpointing specific regularities we hope to shape the direction of theoretical explorations of dynamic fertility phenomena. Our statistical investigations will be guided by the qualitative predictions produced by the available theory from the fields sociology, demography and economics. Our research will also serve to illuminate and evaluate the piecemeal approach to estimating these dynamic relationships that have been pursued in the literature. Piecemeal analysis strategies that (a) condition on other life cycle events and (b) ignore the history of the process generating events such as births may produce misleading inference. By analyzing data using both piecemeal and ingegrated approaches, we will explore the empirical importance of these considerations. We also plan to compare and evaluate dynamic longitudinal procedures for uncovering """"""""causal"""""""" or """"""""structural"""""""" relationships with conventional simultaneous equations procedures used by sociologists and economists. We will also address questions of how sampling plans may affect one's ability to estimate behavioral relationships and development methods to deal with various forms of sample selection which frequently arise in event history data.
