Teacher effectiveness

Massachusetts has been doing some interesting work on teacher effectiveness. Findings so far:

• There are large differences in teacher effectiveness. A teacher at the 75th percentile, relative to the median, improves student achievement by the equivalent of 13 to 15 weeks of learning;
• The cumulative effects of more effective teachers are large. Getting a teacher even only at the 60th percentile rather than the median over fourth through eighth grades provides the equivalent of an extra two months' learning over those five years. Gaps between exemplary teachers and unsatisfactory ones are even larger;
• Low income students are much more likely to get teachers rated as unsatisfactory or as needing improvement;
• Inequities in access to effective teachers increase gaps in achievement between poorer and richer kids.

Wouldn't it be nice if it were possible to do the same work here? There are substantial achievement gaps between low-decile and high-decile schools. Some of it would come down to family effects, whether genetic or environmental. But some of it would also come down to differential effectiveness of schools: recall that there is high variance in outcomes even among low decile schools. 

What would it take?

• It is possible to build contextualised achievement measures for high schools. Grades on externally invigilated NCEA standards can be used to build an outcome measure [or whatever other outcomes you like], and IDI background data can be used as conditioning variables to see which schools over- or under-perform relative to student background characteristics.
• It is currently impossible to link that back to teachers to get a measure of teacher effectiveness. Nothing links students to teachers in the background administrative data, so you don't know which students map to which teachers. And nothing links teachers to schools - or at least if there's a way of linking it, we've not seen it. 
• It is possible to build a contextualised achievement measure for primary schools, but only at one step removed: you infer primary school achievement by outcomes for those schools' students once they get to secondary school. You could try building it out of National Standards data, but it would likely be a bad idea. Students' progress towards National Standards is evaluated by their teachers and you don't want them gaming that. And it looks like Labour's going to scrap National Standards anyway. 

I doubt we get better measures of teacher effectiveness or school performance without strong parent demand for it. I'd be pretty surprised if there weren't similar problems here in getting great teachers into our more challenging schools (on average).