Multivariate Analysis of the Total Fertility Rate and Its Components

该应用提出了一种新的总生育率（TFR）及其多变量分析方法

This application proposes a new method of multivariate analysis of the total fertility rate (TFR) and its components, applied to individual-level survey or census data. A discrete-time multivariate survival model, the complementary log-log (CLL) model, is used to model parity progression. (The parity of a woman is the number of children that she has ever born.) A separate CLL model is estimated for each parity transition. Predictor variables must include woman's age at starting parity (except for the birth-to-first-marriage transition, where everyone starts at the same age) and may include additional socioeconomic variables (broadly conceived to also include health, demographic, environmental, and other variables). The CLL models yield, for specified values of the socioeconomic variables, discrete-time failure probabilities by woman's age, parity, and duration in parity. These failure probabilities enable calculation of a life table that follows women by age, parity, and duration in parity between the ages of 15 and 50. Because this life table is multivariate, all measures calculated from it are also multivariate. The measures considered include parity progression ratios (PPRs), mean and median ages at first marriage, mean and median closed birth intervals, age-specific fertility rates, mean and median ages at childbearing (both overall and by birth order), total marital fertility rate (TMFR), and TFR. The methodology allows calculation of "adjusted" values of these measures by categories of one socioeconomic variable while holding other socioeconomic variables constant at their mean values. The methodology also enables multivariate analysis of trends in these measures over two or more surveys. The analysis can and will be done for both period measures and cohort measures of the TFR and its components. The methodology requires birth histories. The methodology will be tested and refined using the 1993, 1998, and 2003 Philippines Demographic and Health Surveys (DHS), which include birth histories. In the case of census data, which do not include birth histories, the methodology is applied to "reconstructed birth histories." By way of illustration, the new methodology will be applied using reconstructed birth histories to estimate the effect of international migration on the TFR and each of its components in the United States, based on a 1 percent Public Use Microdata Sample (PUMS) from the 1990 U. S. census. An application of the methodology to analyze the effect of child mortality on fertility in India has been dropped, because it is now the major part of Retherford’s research assistant’s Ph.D. dissertation. The new methodology involves major extensions of methodology developed during an earlier R03 grant (“Components of Nuptiality and Fertility Change,” 5R03HD45508-2, PI: Retherford) that expired November 30, 2007.