The statistic is accurate, at least according to
this Census Bureau report from 1995. What it doesn't contain is evidence of this conjecture that "welfare mothers" drive the disparity in premarital birthrates.
They even offer an interesting, rudimentary logistic regression analysis at the end of the article. I'm baffled, however, by a number of methodological shortcomings in their analysis.
First, they use a 90-percent confidence level for statistical analysis when a 95-percent confidence level is standard for social science work. Second, they exclude income as an independent variable when I feel it probably explains the greatest amount of variance of anything else they included. Income is probably important enough to bias race results by its exclusion.
-- Todd/Leviathn