The unobservable factors that underpin the infant mortality gap between blacks and whites have persisted for more than 20 years and now appear to play a larger role than the observable factors, according to a new study by Michigan State University researchers

The study is the first to investigate long-term changes in both sets of factors, providing vital context of the scope of the problem as medical experts around the world try to solve the mystery behind the gap. "What's surprising about our findings is that when we take out all the factors we can observe - including mother's age, education level, marital status and state of residence - the difference in the rate in which black and white infants die remained absolutely stable for two decades," said Steven Haider, professor of economics. "We made no progress in shrinking that part of the gap." Haider and fellow economists Todd Elder and John Goddeeris and epidemiologist Nigel Paneth analyzed the birth certificate data from about 50 million births in the United States between 1983 and 2004. The factors they used to predict mortality differences were then drawn from the information available on the birth certificates. For blacks, the rate of deaths within the first year of life dropped from 18.6 per 100,000 live births in 1983 to 12.3 per 100,000 in 2004. The rate of white infant deaths dropped from 9 to 5.3. Overall, the infant mortality rate gap between the races declined by 2.6 deaths per 100,000 live births. While U.S. health officials have said the narrowing of the gap is encouraging, they still point to the fact that the difference remains large and difficult to shrink significantly. The study separated the gap into the part that could be predicted by differences in observable factors and the part that remained after taking those factors into account. The researchers found that the part of the gap that could not be predicted remained remarkably constant over the two decades. By 2004, about 74% of the black-white infant mortality gap was not accounted for by differences in these observable factors. In other words, even if policymakers could wave a magic wand and eliminate disparities in education, marital status and all the other factors the researchers accounted for, it would only reduce the gap by one quarter, Haider said. While the MSU study doesn't explore what the remaining factors may be, it does, for the first time, establish how large and long-standing the problem is.