Ed Week's Diplomas Count report is out today, as is a warning from four distinguished academics that its figures are "exceedingly inaccurate." And having read the two papers below, I have to agree. The following is the statement issued by Jim Heckman, Paul LaFontaine, Larry Mishel, and Joydeep Roy:

In our examination of the data and methodologies available to estimate high school graduation rates we have found that insights can be gained from household surveys and from administrative data on student enrollment and diplomas granted. However, we find the measures of graduation rates in Education Week’s Diploma Counts project, computed from diploma and enrollment data, to be exceedingly inaccurate. The main problem is the assumption that the number of students enrolled in 9th grade is the same as the number of students entering high school. This assumption artificially lowers the estimates of current graduation rates, especially for minorities who are more likely to be retained (repeat 9th grade). This measure also artificially reduces the growth of the graduation rate over time because the practice of grade retention has grown over time, again, especially among minorities.

The resulting errors are sufficiently large to artificially lower the graduation rate by 9 percentage points overall and by 14 percentage points for minorities. Grade retention also differs sharply across states and localities, distorting geographic comparisons. Last, these measures do not reflect the ultimate graduation rates of a cohort of students because the data do not capture diplomas provided by adult education and other sources than schools.

You can find Mishel and Roy's paper, published today in the Education Policy Analysis Archives, here and Heckman and LaFontaine's paper here.