Sponsored academies have improved more than other schools, but only by 0.7% when compared to similar schools.
The DfE press release yesterday reported that that results in sponsored academies had increased by 3.1% compared to 0.6% in other schools. This claim was repeated amongst others by The Guardian but there is a simple reason for this figure: Sponsor academies are disproportionately concentrated amongst lower achieving schools and rarer amongst high achieving schools. Low achieving schools (particularly sponsored academies) are therefore likely to increase more and become ‘more average’ whilst the higher achievers (mainly not sponsored academies) are likely to reduce more in order to become ‘more average’ – so called ‘regression to the mean’.
It is therefore much more useful to compare the performance of similar schools. So, I divided schools into 10 parts ranked using their 2011 results, counted how many sponsored academies (and sponsored special schools academies) there were in each and randomly selected an equivalent group of non-sponsored academies from each decile. Once I’d cleared out schools for which there was incomplete information (no school type registered or missing data from 2011 or 2012) I got a matched group of 520 schools (260 sponsored academies and 260 others). Sponsored academies had improved by 3.16%, others by 2.48%.
I was also able to compare changes in results amongst all schools within each decile to the sponsored academies within it (I’ve since discovered that Henry Stewart at the Local Schools Network has done something similar here). A word of caution- amongst the top and bottom deciles, numbers of schools are so small that the results are less useful (for example there are only 3 sponsored academies amongst the top and bottom deciles).
We can see from this that sponsor academies mainly just mirror the performance of other schools in their decile with a few exceptions, particularly in the second.
So yes, sponsor academies do do a little better. But really not that much!
(I’m no Becky Allen or Chris Cook so do bear with me if my maths isn’t 100% spot on- you’re more than welcome to check my workings on the attached spreadsheet!)
P.S. I was concerned that the average for my matched sample of “other” schools might just be a consequence of which schools made it into my sample so I also tried matching each academy to the average for a non academy in its decile. The result was 2.45% yielding the same 0.7% gap as my original method.
A lot of people have asked me if the difference is significant. I’m sure you can guess. But no it’s not. Somehow I don’t think a DfE press release which said “Sponsor academies *may possibly* increase results by *almost* 1% more than other schools” would have sounded quite so good.
*amended 18.25 25/01/13 to correct figure on change in non-academies from 2.52% to 2.48% and thus total difference from 0.6% to 0.7%