Contraceptive utilization is an inherently dynamic process of initiation, discontinuation, switching, and failure. Beyond these simple markovian dynamics, it seems likely that contraceptive experiences early in the life-cycle have impacts on later contraceptive practice. In addition, it seems likely that different family planning program efforts have differential effects on the different parts of the dynamic process and on experiences with specific methods. These dynamics imply that static evaluations of the importance of family planning programs and programs designs based on static models of program effects may be quite misleading. To explore the importance of this dynamic perspective, this project analyzes several complementary data sets. First, using the 5-year calendars from two Demographic and Health Surveys in Indonesia (1991 and 1994) with a combined sample of over 40,000 women (over 2 million months), the project will generalize existing models of the effect of family planning program effort on contraceptive prevalence and fertility to incorporate the initiation/continuation distinction. Following recent developments in the literature, these models incorporate village-specific fixed effects to control for village level unobserved heterogeneity and endogenous program placement. Second, the short span of the DHS calendars and the limited amount of pre-calendar data severely limits the types of contraceptive dynamics which can be considered; specifically excluded are many of the models which are indicated by our exploratory data analysis. To explore a richer class of models of contraceptive dynamics, the study will also analyze a subset of cases from the panel of Indonesian Family Life Surveys (IFLS-1 in 1993 and IFLS-2 in 1997). For about 1,500 women married after 1983 (87,000 months), IFLS-1 collected life-time contraceptive histories from marriage to the survey date. IFLS-2 will extend the histories for these women four more years, yielding contraceptive histories from marriage to the survey date for 2,100 women (165,000 months). For this subsample of women, this data collection strategy eliminates the initial conditions problem, allowing the investigators to explore the importance of long-term dependence in contraceptive utilization. Specifically, they will consider the importance of ever having used a method, ever having used a method """"""""successfully""""""""/""""""""unsuccessfully"""""""", having used a method in the previous parity, and having used a method in this pregnancy interval. The results from these two analyses are combined in two ways. First, the IFLS data are used to explore the biases induced into the DHS analyses of program effect due to its limited pre-calendar data and the consequent limitations on forms of contraceptive dynamics which can be included. Second, the study combines information from the two models to construct prototypical simulation models which explicitly incorporate the dependence of current contraceptive choices on past contraceptive choices and on past family planning program efforts. This simulation model is used to derive insights about the effects of considering contraceptive dynamics on the optimal allocation of family planning program effort.
