An Innovative Methodology for Space-Time Analysis with an Application to the 1960–2000 Brazilian Mortality Transition

Abstract

This chapter analyzes space-time patterns in Brazilian regional childhood mortality levels between 1960 and 2000, using a modification of the Knox test. The Knox test, which detects space-time event clustering as possible evidence for contagion, is well established in epidemiology. However, its fundamental assumptions may not hold in demographic applications. The chapter demonstrates how demographers can adapt the Knox test to include space-time patterns in underlying covariates. Analysis of mortality change reveals that much of the contagion found by the standard Knox test is explained by differential timing in regional development. The new method also has an appealing feature: in addition to providing a more appropriate test for demographic contagion, it also identifies regions with unexpectedly early or late mortality declines relative to covariates.