Ethnicity and infant mortality in Texas counties

An ecological analysis of infant mortality rates is conducted using data on Texas counties for the year 1968-72. It is found that when counties along the Texas-Mexico border (which have unreasonably low infant death rates and which are economically less advantaged than the remainder of the counties) are excluded from the analysis, the associations between socioeconomic status variables and infant mortality rates improve considerably. A significant negative correlation between neonatal mortality and per cent Spanish surname is retained and does not disappear when the effects of other relevant variables (per cent Negro, per cent urban, and per cent below poverty) are controlled using partial correlation and regression analysis. Possible explanations for this unexpected finding are suggested, and implications for further research are discussed.