Quasicohort and period measures are weak proxies for the new federal graduation rate

If you were wondering where I was going last week with the release of Florida graduation rates using the new federal definition, it’s this: an analysis of proxy high school graduation measures. It’s short, but I wanted to make it available quickly both here and on the Social Science Research Network because the point is sobering: all the ways researchers have tried to use administrative data turn out to be mediocre at best. Ironically, at a first glance the best performing measure out of the common ones used at the district or school level turns out to be the crude completion rate (diplomas divided by ninth grade enrollment three and two-thirds years before), if one looks at both bias and correlation. But that is still mediocre, accounting for just a little over half of the variation in the new federal rate using data for graduation years 2003-2009.