Geographic inequalities in mortality in the United States: looking beyond socio-economic and racial differentials

Magali Barbieri, Institut National d'Études Démographiques (INED) and University of California, Berkeley

Geographic variations in mortality have tended to decline in the United States up to around 1980 but have increased thereafter, whether measured at the state or at the county level. Because of large differences in the racial and socio-economic composition of the population between geographic areas, most of the geographic variation has been attributed to differential deprivation (factors related to poverty and education). The relationship between mortality and socio-economic factors has been extensively documented. The challenge is to go beyond the socio-economic divide to identify the other factors, hopefully more amenable to intervention than these individual characteristics, contributing to geographic disparities in mortality. A first step in identifying such factors is to measure how much of the inter-state differential in mortality in the U.S. is not due to racial and socio-economic characteristics, which is the purpose of this paper. In the absence of publicly accessible data combining information on both mortality and detailed socio-economic characteristics at the state level, we implement an approach derived from the classic demographic method to standardise mortality rates by age. The basic principle of this approach is to compare two series of mortality rates by sex and age in each state of the U.S. in the year 2000. The first series is directly estimated using vital statistics information (which does not include the desired socio-economic categories) for the year 2000 following standard demographic methods. The second series is produced indirectly using data from the 1998-2002 National Health Interview Survey Linked Mortality Files, with information on survival and socio-economic characteristics but not on geographic location, and from the 2000 Census in a two-stage procedure. The difference between the two series indicates how much of the inter-state differences in mortality is not explained by the direct effect of the socio-economic structure of the population at the state-level.