Município-level estimates of child mortality for Brazil: A new approach using Bayesian Statistics

Previous efforts to estimate child mortality levels in smaller geographical areas have been hampered by the relative rarity of child deaths which has often resulted in unstable and noisy estimates. However, with a spatial smoothing process based upon Bayesian Statistics it is possible to “borrow” information from neighboring areas in order to generate more stable estimates of mortality in smaller areas. The objective of this study is to use a spatial smoothing process to derive estimates of child mortality at the level of the município. Using data from the 2000 Brazil Census including mother's age, number of children born alive, and number of children living, we derive both Bayesian and non-Bayesian estimates of mortality for each of the 5.507 municípios. In this procedure, we worked with the proportion deceased among the children ever born to women age 20-29 in the município, which gives an estimate of the probability of dying before age 5. In comparing the smoothed and raw estimates of this parameter, we find that the use of Bayesian smoothed estimates yield a much clearer spatial pattern of child mortality. Likewise, the variance of the Bayesian estimates for the smaller municípios is much smaller and, we think, more likely to accurately reflect the true mortality situation of those municípios.